id	sid	tid	token	lemma	pos
fcis-17186	1	1	frontiers	frontier	NOUN
fcis-17186	1	2	in	in	ADP
fcis-17186	1	3	computing	computing	NOUN
fcis-17186	1	4	and	and	CCONJ
fcis-17186	1	5	intelligent	intelligent	ADJ
fcis-17186	1	6	systems	system	NOUN
fcis-17186	1	7	issn	issn	VERB
fcis-17186	1	8	:	:	PUNCT
fcis-17186	1	9	2832	2832	NUM
fcis-17186	1	10	-	-	SYM
fcis-17186	1	11	6024	6024	NUM
fcis-17186	1	12	|	|	NOUN
fcis-17186	1	13	vol	vol	NOUN
fcis-17186	1	14	.	.	PROPN
fcis-17186	2	1	7	7	NUM
fcis-17186	2	2	,	,	PUNCT
fcis-17186	2	3	no	no	INTJ
fcis-17186	2	4	.	.	NOUN
fcis-17186	2	5	1	1	NUM
fcis-17186	2	6	,	,	PUNCT
fcis-17186	2	7	2024	2024	NUM
fcis-17186	2	8	69	69	NUM
fcis-17186	2	9	entropy	entropy	NOUN
fcis-17186	2	10	error‐based	error‐base	VERB
fcis-17186	2	11	convolutional	convolutional	ADJ
fcis-17186	2	12	neural	neural	ADJ
fcis-17186	2	13	network	network	NOUN
fcis-17186	2	14	sparse	sparse	ADJ
fcis-17186	2	15	optimization	optimization	NOUN
fcis-17186	2	16	and	and	CCONJ
fcis-17186	2	17	application	application	NOUN
fcis-17186	2	18	kaifang	kaifang	PROPN
fcis-17186	2	19	yang	yang	PROPN
fcis-17186	2	20	*	*	PROPN
fcis-17186	2	21	,	,	PUNCT
fcis-17186	2	22	dewei	dewei	PROPN
fcis-17186	2	23	yang	yang	PROPN
fcis-17186	2	24	,	,	PUNCT
fcis-17186	2	25	fen	fen	PROPN
fcis-17186	2	26	wang	wang	PROPN
fcis-17186	2	27	,	,	PUNCT
fcis-17186	2	28	jing	jing	PROPN
fcis-17186	2	29	zhao	zhao	PROPN
fcis-17186	2	30	,	,	PUNCT
fcis-17186	2	31	qinwei	qinwei	VERB
fcis-17186	2	32	fan	fan	PROPN
fcis-17186	2	33	xi’an	xi’an	PROPN
fcis-17186	2	34	polytechnic	polytechnic	PROPN
fcis-17186	2	35	university	university	PROPN
fcis-17186	2	36	,	,	PUNCT
fcis-17186	2	37	xi’an	xi’an	PROPN
fcis-17186	2	38	shaanxi	shaanxi	PROPN
fcis-17186	2	39	,	,	PUNCT
fcis-17186	2	40	710600	710600	NUM
fcis-17186	2	41	,	,	PUNCT
fcis-17186	2	42	china	china	PROPN
fcis-17186	2	43	*	*	PUNCT
fcis-17186	2	44	corresponding	correspond	VERB
fcis-17186	2	45	author	author	NOUN
fcis-17186	2	46	:	:	PUNCT
fcis-17186	2	47	kaifang	kaifang	PROPN
fcis-17186	2	48	yang	yang	PROPN
fcis-17186	2	49	abstract	abstract	PROPN
fcis-17186	2	50	:	:	PUNCT
fcis-17186	2	51	convolutional	convolutional	ADJ
fcis-17186	2	52	neural	neural	ADJ
fcis-17186	2	53	networks	network	NOUN
fcis-17186	2	54	(	(	PUNCT
fcis-17186	2	55	cnns	cnns	PROPN
fcis-17186	2	56	)	)	PUNCT
fcis-17186	2	57	,	,	PUNCT
fcis-17186	2	58	as	as	ADP
fcis-17186	2	59	quintessential	quintessential	ADJ
fcis-17186	2	60	representatives	representative	NOUN
fcis-17186	2	61	of	of	ADP
fcis-17186	2	62	deep	deep	ADJ
fcis-17186	2	63	neural	neural	ADJ
fcis-17186	2	64	networks	network	NOUN
fcis-17186	2	65	,	,	PUNCT
fcis-17186	2	66	have	have	AUX
fcis-17186	2	67	found	find	VERB
fcis-17186	2	68	widespread	widespread	ADJ
fcis-17186	2	69	applications	application	NOUN
fcis-17186	2	70	in	in	ADP
fcis-17186	2	71	numerous	numerous	ADJ
fcis-17186	2	72	domains	domain	NOUN
fcis-17186	2	73	such	such	ADJ
fcis-17186	2	74	as	as	ADP
fcis-17186	2	75	image	image	NOUN
fcis-17186	2	76	recognition	recognition	NOUN
fcis-17186	2	77	,	,	PUNCT
fcis-17186	2	78	object	object	NOUN
fcis-17186	2	79	detection	detection	NOUN
fcis-17186	2	80	,	,	PUNCT
fcis-17186	2	81	and	and	CCONJ
fcis-17186	2	82	image	image	NOUN
fcis-17186	2	83	generation	generation	NOUN
fcis-17186	2	84	,	,	PUNCT
fcis-17186	2	85	owing	owe	VERB
fcis-17186	2	86	to	to	ADP
fcis-17186	2	87	their	their	PRON
fcis-17186	2	88	robust	robust	ADJ
fcis-17186	2	89	nonlinear	nonlinear	ADJ
fcis-17186	2	90	mapping	mapping	NOUN
fcis-17186	2	91	capabilities	capability	NOUN
fcis-17186	2	92	.	.	PUNCT
fcis-17186	3	1	however	however	ADV
fcis-17186	3	2	,	,	PUNCT
fcis-17186	3	3	the	the	DET
fcis-17186	3	4	practical	practical	ADJ
fcis-17186	3	5	deployment	deployment	NOUN
fcis-17186	3	6	of	of	ADP
fcis-17186	3	7	cnns	cnns	ADJ
fcis-17186	3	8	faces	face	NOUN
fcis-17186	3	9	challenges	challenge	NOUN
fcis-17186	3	10	such	such	ADJ
fcis-17186	3	11	as	as	ADP
fcis-17186	3	12	low	low	ADJ
fcis-17186	3	13	learning	learning	NOUN
fcis-17186	3	14	efficiency	efficiency	NOUN
fcis-17186	3	15	and	and	CCONJ
fcis-17186	3	16	subpar	subpar	ADJ
fcis-17186	3	17	accuracy	accuracy	NOUN
fcis-17186	3	18	due	due	ADP
fcis-17186	3	19	to	to	ADP
fcis-17186	3	20	the	the	DET
fcis-17186	3	21	intricacies	intricacy	NOUN
fcis-17186	3	22	in	in	ADP
fcis-17186	3	23	the	the	DET
fcis-17186	3	24	network	network	NOUN
fcis-17186	3	25	structure	structure	NOUN
fcis-17186	3	26	.	.	PUNCT
fcis-17186	4	1	this	this	DET
fcis-17186	4	2	paper	paper	NOUN
fcis-17186	4	3	introduces	introduce	VERB
fcis-17186	4	4	the	the	DET
fcis-17186	4	5	smoothing	smooth	VERB
fcis-17186	4	6	l1	l1	PROPN
fcis-17186	4	7	regularization	regularization	NOUN
fcis-17186	4	8	term	term	NOUN
fcis-17186	4	9	atop	atop	ADP
fcis-17186	4	10	the	the	DET
fcis-17186	4	11	conventional	conventional	ADJ
fcis-17186	4	12	entropy	entropy	NOUN
fcis-17186	4	13	loss	loss	NOUN
fcis-17186	4	14	function	function	NOUN
fcis-17186	4	15	,	,	PUNCT
fcis-17186	4	16	effectively	effectively	ADV
fcis-17186	4	17	enhancing	enhance	VERB
fcis-17186	4	18	both	both	CCONJ
fcis-17186	4	19	the	the	DET
fcis-17186	4	20	learning	learning	NOUN
fcis-17186	4	21	efficiency	efficiency	NOUN
fcis-17186	4	22	and	and	CCONJ
fcis-17186	4	23	algorithmic	algorithmic	ADJ
fcis-17186	4	24	precision	precision	NOUN
fcis-17186	4	25	of	of	ADP
fcis-17186	4	26	convolutional	convolutional	ADJ
fcis-17186	4	27	neural	neural	ADJ
fcis-17186	4	28	networks	network	NOUN
fcis-17186	4	29	.	.	PUNCT
fcis-17186	5	1	the	the	DET
fcis-17186	5	2	efficacy	efficacy	NOUN
fcis-17186	5	3	of	of	ADP
fcis-17186	5	4	the	the	DET
fcis-17186	5	5	improved	improved	ADJ
fcis-17186	5	6	algorithm	algorithm	NOUN
fcis-17186	5	7	is	be	AUX
fcis-17186	5	8	substantiated	substantiate	VERB
fcis-17186	5	9	through	through	ADP
fcis-17186	5	10	numerical	numerical	ADJ
fcis-17186	5	11	simulations	simulation	NOUN
fcis-17186	5	12	.	.	PUNCT
fcis-17186	6	1	keywords	keyword	NOUN
fcis-17186	6	2	:	:	PUNCT
fcis-17186	6	3	convolution	convolution	NOUN
fcis-17186	6	4	neural	neural	ADJ
fcis-17186	6	5	network	network	NOUN
fcis-17186	6	6	(	(	PUNCT
fcis-17186	6	7	cnn	cnn	PROPN
fcis-17186	6	8	)	)	PUNCT
fcis-17186	6	9	;	;	PUNCT
fcis-17186	6	10	smooth	smooth	ADJ
fcis-17186	6	11	l1	l1	PROPN
fcis-17186	6	12	regularization	regularization	NOUN
fcis-17186	6	13	;	;	PUNCT
fcis-17186	6	14	cross	cross	ADJ
fcis-17186	6	15	-	-	ADJ
fcis-17186	6	16	entropy	entropy	ADJ
fcis-17186	6	17	loss	loss	NOUN
fcis-17186	6	18	function	function	NOUN
fcis-17186	6	19	;	;	PUNCT
fcis-17186	6	20	classification	classification	NOUN
fcis-17186	6	21	problems	problem	NOUN
fcis-17186	6	22	.	.	PUNCT
fcis-17186	7	1	1	1	X
fcis-17186	7	2	.	.	X
fcis-17186	7	3	introduction	introduction	NOUN
fcis-17186	7	4	:	:	PUNCT
fcis-17186	7	5	the	the	DET
fcis-17186	7	6	introduction	introduction	NOUN
fcis-17186	7	7	of	of	ADP
fcis-17186	7	8	convolutional	convolutional	ADJ
fcis-17186	7	9	neural	neural	ADJ
fcis-17186	7	10	networks	network	NOUN
fcis-17186	7	11	(	(	PUNCT
fcis-17186	7	12	cnns	cnns	PROPN
fcis-17186	7	13	)	)	PUNCT
fcis-17186	7	14	signifies	signify	VERB
fcis-17186	7	15	a	a	DET
fcis-17186	7	16	significant	significant	ADJ
fcis-17186	7	17	advancement	advancement	NOUN
fcis-17186	7	18	in	in	ADP
fcis-17186	7	19	the	the	DET
fcis-17186	7	20	field	field	NOUN
fcis-17186	7	21	of	of	ADP
fcis-17186	7	22	deep	deep	ADJ
fcis-17186	7	23	learning	learning	NOUN
fcis-17186	7	24	.	.	PUNCT
fcis-17186	8	1	in	in	ADP
fcis-17186	8	2	1998	1998	NUM
fcis-17186	8	3	,	,	PUNCT
fcis-17186	8	4	lecun	lecun	ADJ
fcis-17186	8	5	and	and	CCONJ
fcis-17186	8	6	collaborators	collaborator	NOUN
fcis-17186	8	7	pioneered	pioneer	VERB
fcis-17186	8	8	cnns	cnn	NOUN
fcis-17186	8	9	,	,	PUNCT
fcis-17186	8	10	applying	apply	VERB
fcis-17186	8	11	them	they	PRON
fcis-17186	8	12	in	in	ADP
fcis-17186	8	13	the	the	DET
fcis-17186	8	14	form	form	NOUN
fcis-17186	8	15	of	of	ADP
fcis-17186	8	16	lenet-5	lenet-5	NUM
fcis-17186	8	17	to	to	PART
fcis-17186	8	18	handwritten	handwritten	ADJ
fcis-17186	8	19	digit	digit	NOUN
fcis-17186	8	20	recognition	recognition	NOUN
fcis-17186	8	21	tasks	task	NOUN
fcis-17186	8	22	,	,	PUNCT
fcis-17186	8	23	thus	thus	ADV
fcis-17186	8	24	propelling	propel	VERB
fcis-17186	8	25	the	the	DET
fcis-17186	8	26	development	development	NOUN
fcis-17186	8	27	of	of	ADP
fcis-17186	8	28	automatic	automatic	ADJ
fcis-17186	8	29	classification	classification	NOUN
fcis-17186	8	30	in	in	ADP
fcis-17186	8	31	digital	digital	ADJ
fcis-17186	8	32	image	image	NOUN
fcis-17186	8	33	research	research	NOUN
fcis-17186	8	34	.	.	PUNCT
fcis-17186	9	1	subsequently	subsequently	ADV
fcis-17186	9	2	,	,	PUNCT
fcis-17186	9	3	scholars	scholar	NOUN
fcis-17186	9	4	widely	widely	ADV
fcis-17186	9	5	applied	apply	VERB
fcis-17186	9	6	various	various	ADJ
fcis-17186	9	7	forms	form	NOUN
fcis-17186	9	8	of	of	ADP
fcis-17186	9	9	cnns	cnn	NOUN
fcis-17186	9	10	across	across	ADP
fcis-17186	9	11	diverse	diverse	ADJ
fcis-17186	9	12	fields	field	NOUN
fcis-17186	9	13	such	such	ADJ
fcis-17186	9	14	as	as	ADP
fcis-17186	9	15	computer	computer	NOUN
fcis-17186	9	16	vision	vision	NOUN
fcis-17186	9	17	,	,	PUNCT
fcis-17186	9	18	medical	medical	ADJ
fcis-17186	9	19	imaging	imaging	NOUN
fcis-17186	9	20	,	,	PUNCT
fcis-17186	9	21	natural	natural	ADJ
fcis-17186	9	22	language	language	NOUN
fcis-17186	9	23	processing	processing	NOUN
fcis-17186	9	24	,	,	PUNCT
fcis-17186	9	25	and	and	CCONJ
fcis-17186	9	26	video	video	NOUN
fcis-17186	9	27	analysis	analysis	NOUN
fcis-17186	9	28	,	,	PUNCT
fcis-17186	9	29	showcasing	showcase	VERB
fcis-17186	9	30	their	their	PRON
fcis-17186	9	31	extensive	extensive	ADJ
fcis-17186	9	32	application	application	NOUN
fcis-17186	9	33	prospects	prospect	NOUN
fcis-17186	9	34	[	[	X
fcis-17186	9	35	1	1	NUM
fcis-17186	9	36	-	-	SYM
fcis-17186	9	37	6	6	NUM
fcis-17186	9	38	]	]	PUNCT
fcis-17186	9	39	.	.	PUNCT
fcis-17186	10	1	as	as	SCONJ
fcis-17186	10	2	models	model	NOUN
fcis-17186	10	3	become	become	VERB
fcis-17186	10	4	more	more	ADV
fcis-17186	10	5	complex	complex	ADJ
fcis-17186	10	6	,	,	PUNCT
fcis-17186	10	7	they	they	PRON
fcis-17186	10	8	may	may	AUX
fcis-17186	10	9	merely	merely	ADV
fcis-17186	10	10	memorize	memorize	VERB
fcis-17186	10	11	noise	noise	NOUN
fcis-17186	10	12	and	and	CCONJ
fcis-17186	10	13	minor	minor	ADJ
fcis-17186	10	14	variations	variation	NOUN
fcis-17186	10	15	in	in	ADP
fcis-17186	10	16	training	training	NOUN
fcis-17186	10	17	data	datum	NOUN
fcis-17186	10	18	,	,	PUNCT
fcis-17186	10	19	resulting	result	VERB
fcis-17186	10	20	in	in	ADP
fcis-17186	10	21	redundant	redundant	ADJ
fcis-17186	10	22	information	information	NOUN
fcis-17186	10	23	and	and	CCONJ
fcis-17186	10	24	diminished	diminish	VERB
fcis-17186	10	25	generalization	generalization	NOUN
fcis-17186	10	26	capabilities	capability	NOUN
fcis-17186	10	27	within	within	ADP
fcis-17186	10	28	the	the	DET
fcis-17186	10	29	network	network	NOUN
fcis-17186	10	30	.	.	PUNCT
fcis-17186	11	1	regularization	regularization	NOUN
fcis-17186	11	2	techniques	technique	NOUN
fcis-17186	11	3	become	become	VERB
fcis-17186	11	4	crucial	crucial	ADJ
fcis-17186	11	5	to	to	PART
fcis-17186	11	6	address	address	VERB
fcis-17186	11	7	this	this	DET
fcis-17186	11	8	issue	issue	NOUN
fcis-17186	11	9	.	.	PUNCT
fcis-17186	12	1	despite	despite	SCONJ
fcis-17186	12	2	attempts	attempt	NOUN
fcis-17186	12	3	by	by	ADP
fcis-17186	12	4	researchers	researcher	NOUN
fcis-17186	12	5	to	to	PART
fcis-17186	12	6	optimize	optimize	VERB
fcis-17186	12	7	neural	neural	ADJ
fcis-17186	12	8	networks	network	NOUN
fcis-17186	12	9	for	for	ADP
fcis-17186	12	10	sparse	sparse	ADJ
fcis-17186	12	11	solutions	solution	NOUN
fcis-17186	12	12	through	through	ADP
fcis-17186	12	13	l1	l1	PROPN
fcis-17186	12	14	regularization	regularization	NOUN
fcis-17186	12	15	,	,	PUNCT
fcis-17186	12	16	introducing	introduce	VERB
fcis-17186	12	17	l1	l1	PROPN
fcis-17186	12	18	regularization	regularization	NOUN
fcis-17186	12	19	terms	term	NOUN
fcis-17186	12	20	may	may	AUX
fcis-17186	12	21	lead	lead	VERB
fcis-17186	12	22	to	to	ADP
fcis-17186	12	23	discontinuities	discontinuity	NOUN
fcis-17186	12	24	in	in	ADP
fcis-17186	12	25	the	the	DET
fcis-17186	12	26	loss	loss	NOUN
fcis-17186	12	27	function	function	NOUN
fcis-17186	12	28	.	.	PUNCT
fcis-17186	13	1	therefore	therefore	ADV
fcis-17186	13	2	,	,	PUNCT
fcis-17186	13	3	we	we	PRON
fcis-17186	13	4	introduce	introduce	VERB
fcis-17186	13	5	the	the	DET
fcis-17186	13	6	technique	technique	NOUN
fcis-17186	13	7	of	of	ADP
fcis-17186	13	8	smoothing	smooth	VERB
fcis-17186	13	9	l1	l1	PROPN
fcis-17186	13	10	regularization	regularization	NOUN
fcis-17186	13	11	,	,	PUNCT
fcis-17186	13	12	considering	consider	VERB
fcis-17186	13	13	the	the	DET
fcis-17186	13	14	sum	sum	NOUN
fcis-17186	13	15	of	of	ADP
fcis-17186	13	16	absolute	absolute	ADJ
fcis-17186	13	17	values	value	NOUN
fcis-17186	13	18	in	in	ADP
fcis-17186	13	19	the	the	DET
fcis-17186	13	20	loss	loss	NOUN
fcis-17186	13	21	function	function	NOUN
fcis-17186	13	22	.	.	PUNCT
fcis-17186	14	1	by	by	ADP
fcis-17186	14	2	enhancing	enhance	VERB
fcis-17186	14	3	smoothness	smoothness	NOUN
fcis-17186	14	4	,	,	PUNCT
fcis-17186	14	5	this	this	PRON
fcis-17186	14	6	effectively	effectively	ADV
fcis-17186	14	7	reduces	reduce	VERB
fcis-17186	14	8	model	model	NOUN
fcis-17186	14	9	complexity	complexity	NOUN
fcis-17186	14	10	,	,	PUNCT
fcis-17186	14	11	mitigates	mitigate	VERB
fcis-17186	14	12	overfitting	overfitte	VERB
fcis-17186	14	13	,	,	PUNCT
fcis-17186	14	14	and	and	CCONJ
fcis-17186	14	15	improves	improve	VERB
fcis-17186	14	16	generalization	generalization	NOUN
fcis-17186	14	17	performance	performance	NOUN
fcis-17186	14	18	through	through	ADP
fcis-17186	14	19	the	the	DET
fcis-17186	14	20	pruning	pruning	NOUN
fcis-17186	14	21	of	of	ADP
fcis-17186	14	22	nodes	node	NOUN
fcis-17186	14	23	approaching	approach	VERB
fcis-17186	14	24	zero	zero	NUM
fcis-17186	14	25	.	.	PUNCT
fcis-17186	15	1	to	to	PART
fcis-17186	15	2	further	far	ADV
fcis-17186	15	3	explore	explore	VERB
fcis-17186	15	4	the	the	DET
fcis-17186	15	5	potential	potential	NOUN
fcis-17186	15	6	of	of	ADP
fcis-17186	15	7	convolutional	convolutional	ADJ
fcis-17186	15	8	neural	neural	ADJ
fcis-17186	15	9	networks	network	NOUN
fcis-17186	15	10	,	,	PUNCT
fcis-17186	15	11	this	this	DET
fcis-17186	15	12	paper	paper	NOUN
fcis-17186	15	13	employs	employ	VERB
fcis-17186	15	14	an	an	DET
fcis-17186	15	15	innovative	innovative	ADJ
fcis-17186	15	16	cnn	cnn	NOUN
fcis-17186	15	17	architecture	architecture	NOUN
fcis-17186	15	18	that	that	PRON
fcis-17186	15	19	integrates	integrate	VERB
fcis-17186	15	20	smoothing	smooth	VERB
fcis-17186	15	21	l1	l1	PROPN
fcis-17186	15	22	regularization	regularization	NOUN
fcis-17186	15	23	and	and	CCONJ
fcis-17186	15	24	cross	cross	ADJ
fcis-17186	15	25	-	-	ADJ
fcis-17186	15	26	entropy	entropy	ADJ
fcis-17186	15	27	loss	loss	NOUN
fcis-17186	15	28	function	function	NOUN
fcis-17186	15	29	[	[	X
fcis-17186	15	30	7	7	NUM
fcis-17186	15	31	-	-	SYM
fcis-17186	15	32	8	8	NUM
fcis-17186	15	33	]	]	PUNCT
fcis-17186	15	34	,	,	PUNCT
fcis-17186	15	35	fully	fully	ADV
fcis-17186	15	36	leveraging	leverage	VERB
fcis-17186	15	37	the	the	DET
fcis-17186	15	38	outstanding	outstanding	ADJ
fcis-17186	15	39	performance	performance	NOUN
fcis-17186	15	40	of	of	ADP
fcis-17186	15	41	cnns	cnn	NOUN
fcis-17186	15	42	in	in	ADP
fcis-17186	15	43	classification	classification	NOUN
fcis-17186	15	44	tasks	task	NOUN
fcis-17186	15	45	.	.	PUNCT
fcis-17186	16	1	in	in	ADP
fcis-17186	16	2	the	the	DET
fcis-17186	16	3	subsequent	subsequent	ADJ
fcis-17186	16	4	section	section	NOUN
fcis-17186	16	5	of	of	ADP
fcis-17186	16	6	this	this	DET
fcis-17186	16	7	paper	paper	NOUN
fcis-17186	16	8	,	,	PUNCT
fcis-17186	16	9	the	the	DET
fcis-17186	16	10	second	second	ADJ
fcis-17186	16	11	part	part	NOUN
fcis-17186	16	12	will	will	AUX
fcis-17186	16	13	introduce	introduce	VERB
fcis-17186	16	14	the	the	DET
fcis-17186	16	15	structure	structure	NOUN
fcis-17186	16	16	and	and	CCONJ
fcis-17186	16	17	working	work	VERB
fcis-17186	16	18	principles	principle	NOUN
fcis-17186	16	19	of	of	ADP
fcis-17186	16	20	convolutional	convolutional	ADJ
fcis-17186	16	21	neural	neural	ADJ
fcis-17186	16	22	networks	network	NOUN
fcis-17186	16	23	(	(	PUNCT
fcis-17186	16	24	cnns	cnns	PROPN
fcis-17186	16	25	)	)	PUNCT
fcis-17186	17	1	[	[	X
fcis-17186	17	2	9	9	NUM
fcis-17186	17	3	]	]	PUNCT
fcis-17186	17	4	.	.	PUNCT
fcis-17186	18	1	it	it	PRON
fcis-17186	18	2	will	will	AUX
fcis-17186	18	3	provide	provide	VERB
fcis-17186	18	4	a	a	DET
fcis-17186	18	5	detailed	detailed	ADJ
fcis-17186	18	6	exposition	exposition	NOUN
fcis-17186	18	7	of	of	ADP
fcis-17186	18	8	the	the	DET
fcis-17186	18	9	mathematical	mathematical	ADJ
fcis-17186	18	10	computation	computation	NOUN
fcis-17186	18	11	process	process	NOUN
fcis-17186	18	12	of	of	ADP
fcis-17186	18	13	the	the	DET
fcis-17186	18	14	sl1	sl1	PROPN
fcis-17186	18	15	regularization	regularization	NOUN
fcis-17186	18	16	term	term	NOUN
fcis-17186	18	17	in	in	ADP
fcis-17186	18	18	the	the	DET
fcis-17186	18	19	convolutional	convolutional	ADJ
fcis-17186	18	20	network	network	NOUN
fcis-17186	18	21	,	,	PUNCT
fcis-17186	18	22	showcasing	showcase	VERB
fcis-17186	18	23	the	the	DET
fcis-17186	18	24	model	model	NOUN
fcis-17186	18	25	construction	construction	NOUN
fcis-17186	18	26	process	process	NOUN
fcis-17186	18	27	.	.	PUNCT
fcis-17186	19	1	finally	finally	ADV
fcis-17186	19	2	,	,	PUNCT
fcis-17186	19	3	the	the	DET
fcis-17186	19	4	third	third	ADJ
fcis-17186	19	5	part	part	NOUN
fcis-17186	19	6	will	will	AUX
fcis-17186	19	7	describe	describe	VERB
fcis-17186	19	8	and	and	CCONJ
fcis-17186	19	9	analyze	analyze	VERB
fcis-17186	19	10	the	the	DET
fcis-17186	19	11	experimental	experimental	ADJ
fcis-17186	19	12	results	result	NOUN
fcis-17186	19	13	to	to	PART
fcis-17186	19	14	validate	validate	VERB
fcis-17186	19	15	the	the	DET
fcis-17186	19	16	effectiveness	effectiveness	NOUN
fcis-17186	19	17	of	of	ADP
fcis-17186	19	18	the	the	DET
fcis-17186	19	19	algorithm	algorithm	NOUN
fcis-17186	19	20	.	.	PUNCT
fcis-17186	20	1	2	2	X
fcis-17186	20	2	.	.	X
fcis-17186	20	3	related	relate	VERB
fcis-17186	20	4	work	work	NOUN
fcis-17186	20	5	:	:	PUNCT
fcis-17186	20	6	(	(	PUNCT
fcis-17186	20	7	1	1	X
fcis-17186	20	8	)	)	PUNCT
fcis-17186	20	9	convolutional	convolutional	ADJ
fcis-17186	20	10	neural	neural	ADJ
fcis-17186	20	11	network	network	NOUN
fcis-17186	20	12	(	(	PUNCT
fcis-17186	20	13	cnn	cnn	PROPN
fcis-17186	20	14	)	)	PUNCT
fcis-17186	20	15	structure	structure	NOUN
fcis-17186	20	16	and	and	CCONJ
fcis-17186	20	17	operating	operating	NOUN
fcis-17186	20	18	principles	principle	NOUN
fcis-17186	20	19	:	:	PUNCT
fcis-17186	20	20	the	the	DET
fcis-17186	20	21	core	core	ADJ
fcis-17186	20	22	architecture	architecture	NOUN
fcis-17186	20	23	of	of	ADP
fcis-17186	20	24	convolutional	convolutional	ADJ
fcis-17186	20	25	neural	neural	ADJ
fcis-17186	20	26	networks	network	NOUN
fcis-17186	20	27	(	(	PUNCT
fcis-17186	20	28	cnns	cnns	PROPN
fcis-17186	20	29	)	)	PUNCT
fcis-17186	20	30	encompasses	encompass	VERB
fcis-17186	20	31	convolutional	convolutional	ADJ
fcis-17186	20	32	layers	layer	NOUN
fcis-17186	20	33	,	,	PUNCT
fcis-17186	20	34	pooling	pool	VERB
fcis-17186	20	35	layers	layer	NOUN
fcis-17186	20	36	,	,	PUNCT
fcis-17186	20	37	and	and	CCONJ
fcis-17186	20	38	fully	fully	ADV
fcis-17186	20	39	connected	connected	ADJ
fcis-17186	20	40	layers	layer	NOUN
fcis-17186	20	41	.	.	PUNCT
fcis-17186	21	1	convolutional	convolutional	ADJ
fcis-17186	21	2	layers	layer	NOUN
fcis-17186	21	3	extract	extract	VERB
fcis-17186	21	4	pivotal	pivotal	ADJ
fcis-17186	21	5	features	feature	NOUN
fcis-17186	21	6	through	through	ADP
fcis-17186	21	7	convolutional	convolutional	ADJ
fcis-17186	21	8	operations	operation	NOUN
fcis-17186	21	9	and	and	CCONJ
fcis-17186	21	10	activation	activation	NOUN
fcis-17186	21	11	functions	function	NOUN
fcis-17186	21	12	,	,	PUNCT
fcis-17186	21	13	while	while	SCONJ
fcis-17186	21	14	pooling	pool	VERB
fcis-17186	21	15	layers	layer	NOUN
fcis-17186	21	16	reduce	reduce	VERB
fcis-17186	21	17	computational	computational	ADJ
fcis-17186	21	18	complexity	complexity	NOUN
fcis-17186	21	19	by	by	ADP
fcis-17186	21	20	downsizing	downsize	VERB
fcis-17186	21	21	feature	feature	NOUN
fcis-17186	21	22	map	map	NOUN
fcis-17186	21	23	dimensions	dimension	NOUN
fcis-17186	21	24	.	.	PUNCT
fcis-17186	22	1	by	by	ADP
fcis-17186	22	2	alternately	alternately	ADV
fcis-17186	22	3	stacking	stack	VERB
fcis-17186	22	4	these	these	DET
fcis-17186	22	5	layers	layer	NOUN
fcis-17186	22	6	,	,	PUNCT
fcis-17186	22	7	features	feature	NOUN
fcis-17186	22	8	are	be	AUX
fcis-17186	22	9	extracted	extract	VERB
fcis-17186	22	10	from	from	ADP
fcis-17186	22	11	input	input	NOUN
fcis-17186	22	12	data	datum	NOUN
fcis-17186	22	13	,	,	PUNCT
fcis-17186	22	14	mapped	map	VERB
fcis-17186	22	15	to	to	ADP
fcis-17186	22	16	outputs	output	NOUN
fcis-17186	22	17	through	through	ADP
fcis-17186	22	18	fully	fully	ADV
fcis-17186	22	19	connected	connected	ADJ
fcis-17186	22	20	layers	layer	NOUN
fcis-17186	22	21	,	,	PUNCT
fcis-17186	22	22	and	and	CCONJ
fcis-17186	22	23	a	a	DET
fcis-17186	22	24	probability	probability	NOUN
fcis-17186	22	25	distribution	distribution	NOUN
fcis-17186	22	26	is	be	AUX
fcis-17186	22	27	obtained	obtain	VERB
fcis-17186	22	28	using	use	VERB
fcis-17186	22	29	the	the	DET
fcis-17186	22	30	softmax	softmax	ADJ
fcis-17186	22	31	activation	activation	NOUN
fcis-17186	22	32	function	function	NOUN
fcis-17186	22	33	.	.	PUNCT
fcis-17186	23	1	the	the	DET
fcis-17186	23	2	loss	loss	NOUN
fcis-17186	23	3	calculation	calculation	NOUN
fcis-17186	23	4	phase	phase	NOUN
fcis-17186	23	5	employs	employ	VERB
fcis-17186	23	6	a	a	DET
fcis-17186	23	7	cross	cross	ADJ
fcis-17186	23	8	-	-	ADJ
fcis-17186	23	9	entropy	entropy	ADJ
fcis-17186	23	10	loss	loss	NOUN
fcis-17186	23	11	function	function	NOUN
fcis-17186	23	12	to	to	PART
fcis-17186	23	13	compare	compare	VERB
fcis-17186	23	14	network	network	NOUN
fcis-17186	23	15	outputs	output	NOUN
fcis-17186	23	16	with	with	ADP
fcis-17186	23	17	actual	actual	ADJ
fcis-17186	23	18	labels	label	NOUN
fcis-17186	23	19	,	,	PUNCT
fcis-17186	23	20	yielding	yield	VERB
fcis-17186	23	21	the	the	DET
fcis-17186	23	22	loss	loss	NOUN
fcis-17186	23	23	value	value	NOUN
fcis-17186	23	24	.	.	PUNCT
fcis-17186	24	1	in	in	ADP
fcis-17186	24	2	the	the	DET
fcis-17186	24	3	backpropagation	backpropagation	NOUN
fcis-17186	24	4	phase	phase	NOUN
fcis-17186	24	5	,	,	PUNCT
fcis-17186	24	6	gradients	gradient	NOUN
fcis-17186	24	7	of	of	ADP
fcis-17186	24	8	layer	layer	NOUN
fcis-17186	24	9	weights	weight	NOUN
fcis-17186	24	10	are	be	AUX
fcis-17186	24	11	computed	compute	VERB
fcis-17186	24	12	using	use	VERB
fcis-17186	24	13	the	the	DET
fcis-17186	24	14	chain	chain	NOUN
fcis-17186	24	15	rule	rule	NOUN
fcis-17186	25	1	[	[	X
fcis-17186	25	2	10	10	NUM
fcis-17186	25	3	]	]	PUNCT
fcis-17186	25	4	and	and	CCONJ
fcis-17186	25	5	subsequently	subsequently	ADV
fcis-17186	25	6	updated	update	VERB
fcis-17186	25	7	.	.	PUNCT
fcis-17186	26	1	to	to	PART
fcis-17186	26	2	expedite	expedite	VERB
fcis-17186	26	3	backpropagation	backpropagation	NOUN
fcis-17186	26	4	,	,	PUNCT
fcis-17186	26	5	a	a	DET
fcis-17186	26	6	momentum	momentum	NOUN
fcis-17186	26	7	term	term	NOUN
fcis-17186	26	8	is	be	AUX
fcis-17186	26	9	introduced	introduce	VERB
fcis-17186	26	10	as	as	ADP
fcis-17186	26	11	an	an	DET
fcis-17186	26	12	optimization	optimization	NOUN
fcis-17186	26	13	method	method	NOUN
fcis-17186	26	14	,	,	PUNCT
fcis-17186	26	15	aiming	aim	VERB
fcis-17186	26	16	to	to	PART
fcis-17186	26	17	enhance	enhance	VERB
fcis-17186	26	18	the	the	DET
fcis-17186	26	19	convergence	convergence	NOUN
fcis-17186	26	20	speed	speed	NOUN
fcis-17186	26	21	of	of	ADP
fcis-17186	26	22	gradient	gradient	ADJ
fcis-17186	26	23	descent	descent	NOUN
fcis-17186	26	24	and	and	CCONJ
fcis-17186	26	25	improve	improve	VERB
fcis-17186	26	26	the	the	DET
fcis-17186	26	27	stability	stability	NOUN
fcis-17186	26	28	of	of	ADP
fcis-17186	26	29	model	model	NOUN
fcis-17186	26	30	training	training	NOUN
fcis-17186	26	31	.	.	PUNCT
fcis-17186	27	1	through	through	ADP
fcis-17186	27	2	iterative	iterative	NOUN
fcis-17186	27	3	processes	process	NOUN
fcis-17186	27	4	,	,	PUNCT
fcis-17186	27	5	network	network	NOUN
fcis-17186	27	6	parameters	parameter	NOUN
fcis-17186	27	7	are	be	AUX
fcis-17186	27	8	gradually	gradually	ADV
fcis-17186	27	9	adjusted	adjust	VERB
fcis-17186	27	10	to	to	PART
fcis-17186	27	11	enhance	enhance	VERB
fcis-17186	27	12	model	model	NOUN
fcis-17186	27	13	performance	performance	NOUN
fcis-17186	27	14	,	,	PUNCT
fcis-17186	27	15	the	the	DET
fcis-17186	27	16	convolutional	convolutional	ADJ
fcis-17186	27	17	network	network	NOUN
fcis-17186	27	18	structure	structure	NOUN
fcis-17186	27	19	we	we	PRON
fcis-17186	27	20	designed	design	VERB
fcis-17186	27	21	is	be	AUX
fcis-17186	27	22	illustrated	illustrate	VERB
fcis-17186	27	23	in	in	ADP
fcis-17186	27	24	the	the	DET
fcis-17186	27	25	accompanying	accompanying	ADJ
fcis-17186	27	26	diagram	diagram	NOUN
fcis-17186	27	27	.	.	PUNCT
fcis-17186	28	1	in	in	ADP
fcis-17186	28	2	cnn	cnn	PROPN
fcis-17186	28	3	,	,	PUNCT
fcis-17186	28	4	we	we	PRON
fcis-17186	28	5	process	process	VERB
fcis-17186	28	6	input	input	NOUN
fcis-17186	28	7	images	image	NOUN
fcis-17186	28	8	of	of	ADP
fcis-17186	28	9	size	size	NOUN
fcis-17186	28	10	28×28	28×28	NOUN
fcis-17186	28	11	.	.	PUNCT
fcis-17186	29	1	the	the	DET
fcis-17186	29	2	images	image	NOUN
fcis-17186	29	3	undergo	undergo	VERB
fcis-17186	29	4	convolutional	convolutional	ADJ
fcis-17186	29	5	operations	operation	NOUN
fcis-17186	29	6	with	with	ADP
fcis-17186	29	7	a	a	DET
fcis-17186	29	8	5×5	5×5	NUM
fcis-17186	29	9	convolutional	convolutional	ADJ
fcis-17186	29	10	kernel	kernel	NOUN
fcis-17186	29	11	,	,	PUNCT
fcis-17186	29	12	followed	follow	VERB
fcis-17186	29	13	by	by	ADP
fcis-17186	29	14	non	non	ADJ
fcis-17186	29	15	-	-	ADJ
fcis-17186	29	16	linear	linear	ADJ
fcis-17186	29	17	processing	processing	NOUN
fcis-17186	29	18	through	through	ADP
fcis-17186	29	19	the	the	DET
fcis-17186	29	20	sigmoid	sigmoid	NOUN
fcis-17186	29	21	activation	activation	NOUN
fcis-17186	29	22	function	function	NOUN
fcis-17186	29	23	.	.	PUNCT
fcis-17186	30	1	the	the	DET
fcis-17186	30	2	resulting	result	VERB
fcis-17186	30	3	feature	feature	NOUN
fcis-17186	30	4	maps	map	NOUN
fcis-17186	30	5	are	be	AUX
fcis-17186	30	6	then	then	ADV
fcis-17186	30	7	subjected	subject	VERB
fcis-17186	30	8	to	to	ADP
fcis-17186	30	9	average	average	ADJ
fcis-17186	30	10	pooling	pooling	NOUN
fcis-17186	30	11	,	,	PUNCT
fcis-17186	30	12	reducing	reduce	VERB
fcis-17186	30	13	the	the	DET
fcis-17186	30	14	image	image	NOUN
fcis-17186	30	15	dimensions	dimension	NOUN
fcis-17186	30	16	to	to	ADP
fcis-17186	30	17	12×12	12×12	NUM
fcis-17186	30	18	.	.	PUNCT
fcis-17186	31	1	the	the	DET
fcis-17186	31	2	images	image	NOUN
fcis-17186	31	3	then	then	ADV
fcis-17186	31	4	pass	pass	VERB
fcis-17186	31	5	through	through	ADP
fcis-17186	31	6	a	a	DET
fcis-17186	31	7	second	second	ADJ
fcis-17186	31	8	convolutional	convolutional	ADJ
fcis-17186	31	9	layer	layer	NOUN
fcis-17186	31	10	with	with	ADP
fcis-17186	31	11	the	the	DET
fcis-17186	31	12	same	same	ADJ
fcis-17186	31	13	pooling	pooling	NOUN
fcis-17186	31	14	operation	operation	NOUN
fcis-17186	31	15	as	as	ADP
fcis-17186	31	16	the	the	DET
fcis-17186	31	17	first	first	ADJ
fcis-17186	31	18	layer	layer	NOUN
fcis-17186	31	19	.	.	PUNCT
fcis-17186	32	1	this	this	DET
fcis-17186	32	2	step	step	NOUN
fcis-17186	32	3	involves	involve	VERB
fcis-17186	32	4	convolution	convolution	NOUN
fcis-17186	32	5	and	and	CCONJ
fcis-17186	32	6	pooling	pool	VERB
fcis-17186	32	7	processes	process	NOUN
fcis-17186	32	8	again	again	ADV
fcis-17186	32	9	to	to	PART
fcis-17186	32	10	extract	extract	VERB
fcis-17186	32	11	features	feature	NOUN
fcis-17186	32	12	.	.	PUNCT
fcis-17186	33	1	the	the	DET
fcis-17186	33	2	processed	process	VERB
fcis-17186	33	3	results	result	NOUN
fcis-17186	33	4	are	be	AUX
fcis-17186	33	5	flattened	flatten	VERB
fcis-17186	33	6	into	into	ADP
fcis-17186	33	7	a	a	DET
fcis-17186	33	8	16×1	16×1	PROPN
fcis-17186	33	9	vector	vector	NOUN
fcis-17186	33	10	through	through	ADP
fcis-17186	33	11	a	a	DET
fcis-17186	33	12	fully	fully	ADV
fcis-17186	33	13	connected	connect	VERB
fcis-17186	33	14	layer	layer	NOUN
fcis-17186	33	15	.	.	PUNCT
fcis-17186	34	1	the	the	DET
fcis-17186	34	2	vectors	vector	NOUN
fcis-17186	34	3	are	be	AUX
fcis-17186	34	4	concatenated	concatenate	VERB
fcis-17186	34	5	to	to	PART
fcis-17186	34	6	form	form	VERB
fcis-17186	34	7	a	a	DET
fcis-17186	34	8	192×1	192×1	NUM
fcis-17186	34	9	vector	vector	NOUN
fcis-17186	34	10	,	,	PUNCT
fcis-17186	34	11	aiding	aid	VERB
fcis-17186	34	12	in	in	ADP
fcis-17186	34	13	the	the	DET
fcis-17186	34	14	fusion	fusion	NOUN
fcis-17186	34	15	of	of	ADP
fcis-17186	34	16	features	feature	NOUN
fcis-17186	34	17	extracted	extract	VERB
fcis-17186	34	18	from	from	ADP
fcis-17186	34	19	different	different	ADJ
fcis-17186	34	20	convolutional	convolutional	ADJ
fcis-17186	34	21	and	and	CCONJ
fcis-17186	34	22	pooling	pool	VERB
fcis-17186	34	23	layers	layer	NOUN
fcis-17186	34	24	.	.	PUNCT
fcis-17186	35	1	finally	finally	ADV
fcis-17186	35	2	,	,	PUNCT
fcis-17186	35	3	by	by	ADP
fcis-17186	35	4	applying	apply	VERB
fcis-17186	35	5	the	the	DET
fcis-17186	35	6	softmax	softmax	NOUN
fcis-17186	35	7	activation	activation	NOUN
fcis-17186	35	8	function	function	NOUN
fcis-17186	35	9	,	,	PUNCT
fcis-17186	35	10	this	this	DET
fcis-17186	35	11	vector	vector	NOUN
fcis-17186	35	12	is	be	AUX
fcis-17186	35	13	mapped	map	VERB
fcis-17186	35	14	to	to	ADP
fcis-17186	35	15	a	a	DET
fcis-17186	35	16	70	70	NUM
fcis-17186	35	17	category	category	NOUN
fcis-17186	35	18	probability	probability	NOUN
fcis-17186	35	19	distribution	distribution	NOUN
fcis-17186	35	20	for	for	ADP
fcis-17186	35	21	label	label	NOUN
fcis-17186	35	22	-	-	PUNCT
fcis-17186	35	23	based	base	VERB
fcis-17186	35	24	classification	classification	NOUN
fcis-17186	35	25	.	.	PUNCT
fcis-17186	36	1	figure	figure	NOUN
fcis-17186	36	2	1	1	NUM
fcis-17186	36	3	.	.	PUNCT
fcis-17186	36	4	convolutional	convolutional	ADJ
fcis-17186	36	5	neural	neural	ADJ
fcis-17186	36	6	network	network	NOUN
fcis-17186	36	7	(	(	PUNCT
fcis-17186	36	8	2	2	NUM
fcis-17186	36	9	)	)	PUNCT
fcis-17186	36	10	mathematical	mathematical	ADJ
fcis-17186	36	11	computations	computation	NOUN
fcis-17186	36	12	convolution	convolution	NOUN
fcis-17186	36	13	operation	operation	NOUN
fcis-17186	36	14	is	be	AUX
fcis-17186	36	15	a	a	DET
fcis-17186	36	16	crucial	crucial	ADJ
fcis-17186	36	17	step	step	NOUN
fcis-17186	36	18	in	in	ADP
fcis-17186	36	19	feature	feature	NOUN
fcis-17186	36	20	extraction	extraction	NOUN
fcis-17186	36	21	in	in	ADP
fcis-17186	36	22	deep	deep	ADJ
fcis-17186	36	23	learning	learning	NOUN
fcis-17186	36	24	,	,	PUNCT
fcis-17186	36	25	especially	especially	ADV
fcis-17186	36	26	in	in	ADP
fcis-17186	36	27	the	the	DET
fcis-17186	36	28	context	context	NOUN
fcis-17186	36	29	of	of	ADP
fcis-17186	36	30	image	image	NOUN
fcis-17186	36	31	processing	processing	NOUN
fcis-17186	36	32	.	.	PUNCT
fcis-17186	37	1	for	for	ADP
fcis-17186	37	2	the	the	DET
fcis-17186	37	3	convolution	convolution	NOUN
fcis-17186	37	4	operation	operation	NOUN
fcis-17186	37	5	,	,	PUNCT
fcis-17186	37	6	the	the	DET
fcis-17186	37	7	calculation	calculation	NOUN
fcis-17186	37	8	formula	formula	NOUN
fcis-17186	37	9	between	between	ADP
fcis-17186	37	10	the	the	DET
fcis-17186	37	11	input	input	NOUN
fcis-17186	37	12	matrix	matrix	NOUN
fcis-17186	37	13	x	x	PUNCT
fcis-17186	37	14	and	and	CCONJ
fcis-17186	37	15	the	the	DET
fcis-17186	37	16	convolutional	convolutional	ADJ
fcis-17186	37	17	kernel	kernel	NOUN
fcis-17186	37	18	k	k	PROPN
fcis-17186	37	19	is	be	AUX
fcis-17186	37	20	:	:	PUNCT
fcis-17186	37	21	(	(	PUNCT
fcis-17186	37	22	,	,	PUNCT
fcis-17186	37	23	)	)	PUNCT
fcis-17186	37	24	(	(	PUNCT
fcis-17186	37	25	,	,	PUNCT
fcis-17186	37	26	)	)	PUNCT
fcis-17186	37	27	(	(	PUNCT
fcis-17186	37	28	,	,	PUNCT
fcis-17186	37	29	)	)	PUNCT
fcis-17186	37	30	,	,	PUNCT
fcis-17186	37	31	m	m	VERB
fcis-17186	37	32	n	n	ADP
fcis-17186	37	33	a	a	PRON
fcis-17186	38	1	i	i	INTJ
fcis-17186	38	2	j	j	NOUN
fcis-17186	39	1	x	x	VERB
fcis-17186	39	2	i	i	PRON
fcis-17186	39	3	m	m	VERB
fcis-17186	39	4	j	j	PROPN
fcis-17186	40	1	n	n	CCONJ
fcis-17186	40	2	k	k	PROPN
fcis-17186	40	3	m	m	VERB
fcis-17186	40	4	n	n	PROPN
fcis-17186	40	5			ADV
fcis-17186	40	6			PUNCT
fcis-17186	40	7			X
fcis-17186	40	8	(	(	PUNCT
fcis-17186	40	9	1	1	X
fcis-17186	40	10	)	)	PUNCT
fcis-17186	40	11	given	give	VERB
fcis-17186	40	12	the	the	DET
fcis-17186	40	13	convolved	convolve	VERB
fcis-17186	40	14	output	output	NOUN
fcis-17186	40	15	matrix	matrix	NOUN
fcis-17186	40	16	a	a	PRON
fcis-17186	40	17	and	and	CCONJ
fcis-17186	40	18	the	the	DET
fcis-17186	40	19	size	size	NOUN
fcis-17186	40	20	of	of	ADP
fcis-17186	40	21	the	the	DET
fcis-17186	40	22	pooling	pooling	NOUN
fcis-17186	40	23	window	window	NOUN
fcis-17186	40	24	k×k	k×k	PROPN
fcis-17186	40	25	,	,	PUNCT
fcis-17186	40	26	the	the	DET
fcis-17186	40	27	computation	computation	NOUN
fcis-17186	40	28	process	process	NOUN
fcis-17186	40	29	of	of	ADP
fcis-17186	40	30	the	the	DET
fcis-17186	40	31	output	output	NOUN
fcis-17186	40	32	matrix	matrix	NOUN
fcis-17186	40	33	is	be	AUX
fcis-17186	40	34	as	as	SCONJ
fcis-17186	40	35	follows	follow	VERB
fcis-17186	40	36	:	:	PUNCT
fcis-17186	40	37	1	1	NUM
fcis-17186	40	38	1	1	NUM
fcis-17186	40	39	2	2	NUM
fcis-17186	40	40	0	0	NUM
fcis-17186	40	41	0	0	NUM
fcis-17186	40	42	1	1	NUM
fcis-17186	40	43	(	(	PUNCT
fcis-17186	40	44	,	,	PUNCT
fcis-17186	40	45	)	)	PUNCT
fcis-17186	40	46	(	(	PUNCT
fcis-17186	40	47	,	,	PUNCT
fcis-17186	40	48	)	)	PUNCT
fcis-17186	40	49	,	,	PUNCT
fcis-17186	41	1	k	k	PROPN
fcis-17186	41	2	k	k	PROPN
fcis-17186	41	3	m	m	VERB
fcis-17186	41	4	n	n	ADV
fcis-17186	41	5	o	o	X
fcis-17186	42	1	i	i	PRON
fcis-17186	42	2	j	j	PROPN
fcis-17186	43	1	a	a	PRON
fcis-17186	43	2	i	i	PRON
fcis-17186	44	1	k	k	PROPN
fcis-17186	44	2	m	m	VERB
fcis-17186	45	1	j	j	PROPN
fcis-17186	45	2	k	k	PROPN
fcis-17186	45	3	n	n	CCONJ
fcis-17186	45	4	k	k	PROPN
fcis-17186	45	5			PROPN
fcis-17186	45	6			PROPN
fcis-17186	45	7			NUM
fcis-17186	45	8			PRON
fcis-17186	46	1			PROPN
fcis-17186	46	2			PROPN
fcis-17186	46	3			ADJ
fcis-17186	46	4			NUM
fcis-17186	46	5			NOUN
fcis-17186	46	6	(	(	PUNCT
fcis-17186	46	7	2	2	NUM
fcis-17186	46	8	)	)	PUNCT
fcis-17186	46	9	in	in	ADP
fcis-17186	46	10	(	(	PUNCT
fcis-17186	46	11	1	1	NUM
fcis-17186	46	12	)	)	PUNCT
fcis-17186	46	13	and	and	CCONJ
fcis-17186	46	14	(	(	PUNCT
fcis-17186	46	15	2	2	NUM
fcis-17186	46	16	)	)	PUNCT
fcis-17186	46	17	,	,	PUNCT
fcis-17186	46	18	denotes	denote	VERB
fcis-17186	46	19	the	the	DET
fcis-17186	46	20	element	element	NOUN
fcis-17186	46	21	in	in	ADP
fcis-17186	46	22	the	the	DET
fcis-17186	46	23	i	i	PROPN
fcis-17186	46	24	-	-	PUNCT
fcis-17186	46	25	th	th	X
fcis-17186	46	26	row	row	NOUN
fcis-17186	46	27	and	and	CCONJ
fcis-17186	46	28	j	j	PROPN
fcis-17186	46	29	-	-	PUNCT
fcis-17186	46	30	th	th	VERB
fcis-17186	46	31	column	column	NOUN
fcis-17186	46	32	of	of	ADP
fcis-17186	46	33	the	the	DET
fcis-17186	46	34	input	input	NOUN
fcis-17186	46	35	matrix	matrix	NOUN
fcis-17186	46	36	,	,	PUNCT
fcis-17186	46	37	representing	represent	VERB
fcis-17186	46	38	the	the	DET
fcis-17186	46	39	i	i	PROPN
fcis-17186	46	40	-	-	PUNCT
fcis-17186	46	41	th	th	X
fcis-17186	46	42	row	row	NOUN
fcis-17186	46	43	and	and	CCONJ
fcis-17186	46	44	j	j	PROPN
fcis-17186	46	45	-	-	PUNCT
fcis-17186	46	46	th	th	VERB
fcis-17186	46	47	column	column	NOUN
fcis-17186	46	48	element	element	NOUN
fcis-17186	46	49	in	in	ADP
fcis-17186	46	50	the	the	DET
fcis-17186	46	51	convolutional	convolutional	ADJ
fcis-17186	46	52	kernel	kernel	NOUN
fcis-17186	46	53	k.	k.	PROPN
fcis-17186	47	1	the	the	DET
fcis-17186	47	2	calculation	calculation	NOUN
fcis-17186	47	3	formula	formula	NOUN
fcis-17186	47	4	for	for	ADP
fcis-17186	47	5	the	the	DET
fcis-17186	47	6	fully	fully	ADV
fcis-17186	47	7	connected	connect	VERB
fcis-17186	47	8	layer	layer	NOUN
fcis-17186	47	9	is	be	AUX
fcis-17186	47	10	the	the	DET
fcis-17186	47	11	same	same	ADJ
fcis-17186	47	12	as	as	ADP
fcis-17186	47	13	that	that	PRON
fcis-17186	47	14	of	of	ADP
fcis-17186	47	15	a	a	DET
fcis-17186	47	16	conventional	conventional	ADJ
fcis-17186	47	17	fully	fully	ADV
fcis-17186	47	18	connected	connect	VERB
fcis-17186	47	19	layer	layer	NOUN
fcis-17186	47	20	.	.	PUNCT
fcis-17186	48	1	for	for	ADP
fcis-17186	48	2	a	a	DET
fcis-17186	48	3	fully	fully	ADV
fcis-17186	48	4	connected	connect	VERB
fcis-17186	48	5	layer	layer	NOUN
fcis-17186	48	6	,	,	PUNCT
fcis-17186	48	7	given	give	VERB
fcis-17186	48	8	the	the	DET
fcis-17186	48	9	output	output	NOUN
fcis-17186	48	10	of	of	ADP
fcis-17186	48	11	the	the	DET
fcis-17186	48	12	previous	previous	ADJ
fcis-17186	48	13	layer	layer	NOUN
fcis-17186	48	14	,	,	PUNCT
fcis-17186	48	15	weight	weight	NOUN
fcis-17186	48	16	matrix	matrix	NOUN
fcis-17186	48	17	,	,	PUNCT
fcis-17186	48	18	and	and	CCONJ
fcis-17186	48	19	bias	bias	NOUN
fcis-17186	48	20	vector	vector	NOUN
fcis-17186	48	21	,	,	PUNCT
fcis-17186	48	22	the	the	DET
fcis-17186	48	23	calculation	calculation	NOUN
fcis-17186	48	24	process	process	NOUN
fcis-17186	48	25	for	for	ADP
fcis-17186	48	26	the	the	DET
fcis-17186	48	27	output	output	NOUN
fcis-17186	48	28	is	be	AUX
fcis-17186	48	29	as	as	SCONJ
fcis-17186	48	30	follows	follow	VERB
fcis-17186	48	31	:	:	PUNCT
fcis-17186	48	32	re	re	X
fcis-17186	48	33	(	(	PUNCT
fcis-17186	48	34	)	)	PUNCT
fcis-17186	48	35	,	,	PUNCT
fcis-17186	48	36	s	s	VERB
fcis-17186	48	37	lu	lu	NOUN
fcis-17186	48	38	o	o	X
fcis-17186	48	39	u	u	NOUN
fcis-17186	48	40	b	b	PROPN
fcis-17186	48	41			PROPN
fcis-17186	48	42			PROPN
fcis-17186	48	43	(	(	PUNCT
fcis-17186	48	44	3	3	X
fcis-17186	48	45	)	)	PUNCT
fcis-17186	48	46	where	where	SCONJ
fcis-17186	48	47	represents	represent	VERB
fcis-17186	48	48	matrix	matrix	NOUN
fcis-17186	48	49	multiplication	multiplication	NOUN
fcis-17186	48	50	,	,	PUNCT
fcis-17186	48	51	o	o	NOUN
fcis-17186	48	52	is	be	AUX
fcis-17186	48	53	the	the	DET
fcis-17186	48	54	output	output	NOUN
fcis-17186	48	55	vector	vector	NOUN
fcis-17186	48	56	from	from	ADP
fcis-17186	48	57	the	the	DET
fcis-17186	48	58	previous	previous	ADJ
fcis-17186	48	59	layer	layer	NOUN
fcis-17186	48	60	,	,	PUNCT
fcis-17186	48	61	u	u	NOUN
fcis-17186	48	62	is	be	AUX
fcis-17186	48	63	the	the	DET
fcis-17186	48	64	weight	weight	NOUN
fcis-17186	48	65	matrix	matrix	NOUN
fcis-17186	48	66	of	of	ADP
fcis-17186	48	67	the	the	DET
fcis-17186	48	68	fully	fully	ADV
fcis-17186	48	69	connected	connected	ADJ
fcis-17186	48	70	layer	layer	NOUN
fcis-17186	48	71	,	,	PUNCT
fcis-17186	48	72	and	and	CCONJ
fcis-17186	48	73	b	b	NOUN
fcis-17186	48	74	is	be	AUX
fcis-17186	48	75	the	the	DET
fcis-17186	48	76	bias	bias	NOUN
fcis-17186	48	77	vector	vector	NOUN
fcis-17186	48	78	.	.	PUNCT
fcis-17186	49	1	relu	relu	NOUN
fcis-17186	49	2	is	be	AUX
fcis-17186	49	3	a	a	DET
fcis-17186	49	4	commonly	commonly	ADV
fcis-17186	49	5	used	use	VERB
fcis-17186	49	6	activation	activation	NOUN
fcis-17186	49	7	function	function	NOUN
fcis-17186	49	8	defined	define	VERB
fcis-17186	49	9	as	as	ADP
fcis-17186	49	10	r	r	NOUN
fcis-17186	49	11	e	e	X
fcis-17186	49	12	(	(	PUNCT
fcis-17186	49	13	)	)	PUNCT
fcis-17186	49	14	m	m	VERB
fcis-17186	49	15	ax(0	ax(0	PROPN
fcis-17186	49	16	,	,	PUNCT
fcis-17186	49	17	)	)	PUNCT
fcis-17186	49	18	lu	lu	PROPN
fcis-17186	50	1	x	x	X
fcis-17186	50	2	x	x	PROPN
fcis-17186	51	1	.the	.the	PRON
fcis-17186	51	2	introduction	introduction	NOUN
fcis-17186	51	3	of	of	ADP
fcis-17186	51	4	this	this	DET
fcis-17186	51	5	non	non	ADJ
fcis-17186	51	6	-	-	ADJ
fcis-17186	51	7	linear	linear	ADJ
fcis-17186	51	8	activation	activation	NOUN
fcis-17186	51	9	function	function	NOUN
fcis-17186	51	10	enables	enable	VERB
fcis-17186	51	11	the	the	DET
fcis-17186	51	12	network	network	NOUN
fcis-17186	51	13	to	to	PART
fcis-17186	51	14	better	well	ADV
fcis-17186	51	15	learn	learn	VERB
fcis-17186	51	16	complex	complex	ADJ
fcis-17186	51	17	feature	feature	NOUN
fcis-17186	51	18	mappings	mapping	NOUN
fcis-17186	51	19	.	.	PUNCT
fcis-17186	52	1	the	the	DET
fcis-17186	52	2	output	output	NOUN
fcis-17186	52	3	s	s	VERB
fcis-17186	52	4	of	of	ADP
fcis-17186	52	5	the	the	DET
fcis-17186	52	6	fully	fully	ADV
fcis-17186	52	7	connected	connect	VERB
fcis-17186	52	8	layer	layer	NOUN
fcis-17186	52	9	can	can	AUX
fcis-17186	52	10	be	be	AUX
fcis-17186	52	11	utilized	utilize	VERB
fcis-17186	52	12	for	for	ADP
fcis-17186	52	13	classification	classification	NOUN
fcis-17186	52	14	tasks	task	NOUN
fcis-17186	52	15	,	,	PUNCT
fcis-17186	52	16	and	and	CCONJ
fcis-17186	52	17	based	base	VERB
fcis-17186	52	18	on	on	ADP
fcis-17186	52	19	the	the	DET
fcis-17186	52	20	entropy	entropy	NOUN
fcis-17186	52	21	error	error	NOUN
fcis-17186	52	22	function	function	NOUN
fcis-17186	52	23	,	,	PUNCT
fcis-17186	52	24	the	the	DET
fcis-17186	52	25	error	error	NOUN
fcis-17186	52	26	function	function	NOUN
fcis-17186	52	27	of	of	ADP
fcis-17186	52	28	the	the	DET
fcis-17186	52	29	convolutional	convolutional	ADJ
fcis-17186	52	30	neural	neural	ADJ
fcis-17186	52	31	network	network	NOUN
fcis-17186	52	32	is	be	AUX
fcis-17186	52	33	defined	define	VERB
fcis-17186	52	34	as	as	ADP
fcis-17186	52	35	:	:	SYM
fcis-17186	52	36	1	1	NUM
fcis-17186	52	37	1	1	NUM
fcis-17186	52	38	[	[	PUNCT
fcis-17186	52	39	ln	ln	X
fcis-17186	52	40	(	(	PUNCT
fcis-17186	52	41	1	1	NUM
fcis-17186	52	42	)	)	PUNCT
fcis-17186	52	43	ln	ln	NOUN
fcis-17186	52	44	]	]	X
fcis-17186	52	45	.	.	PUNCT
fcis-17186	53	1	1	1	NUM
fcis-17186	53	2	j	j	PROPN
fcis-17186	53	3	jj	jj	PROPN
fcis-17186	53	4	j	j	PROPN
fcis-17186	53	5	j	j	PROPN
fcis-17186	53	6	j	j	PROPN
fcis-17186	53	7	j	j	PROPN
fcis-17186	53	8	y	y	PROPN
fcis-17186	53	9	y	y	PROPN
fcis-17186	53	10	e	e	PROPN
fcis-17186	53	11			NOUN
fcis-17186	53	12			VERB
fcis-17186	53	13			VERB
fcis-17186	53	14			PUNCT
fcis-17186	53	15			NOUN
fcis-17186	53	16			PROPN
fcis-17186	53	17			PROPN
fcis-17186	53	18			VERB
fcis-17186	53	19			PROPN
fcis-17186	53	20			PROPN
fcis-17186	53	21	(	(	PUNCT
fcis-17186	53	22	4	4	NUM
fcis-17186	53	23	)	)	PUNCT
fcis-17186	53	24	(	(	PUNCT
fcis-17186	53	25	3	3	X
fcis-17186	53	26	)	)	PUNCT
fcis-17186	53	27	smoothed	smooth	VERB
fcis-17186	53	28	l1	l1	PROPN
fcis-17186	53	29	regularization	regularization	NOUN
fcis-17186	53	30	term	term	NOUN
fcis-17186	53	31	smooth	smooth	ADJ
fcis-17186	53	32	approximation	approximation	NOUN
fcis-17186	53	33	is	be	AUX
fcis-17186	53	34	a	a	DET
fcis-17186	53	35	strategy	strategy	NOUN
fcis-17186	53	36	that	that	PRON
fcis-17186	53	37	substitutes	substitute	VERB
fcis-17186	53	38	the	the	DET
fcis-17186	53	39	absolute	absolute	ADJ
fcis-17186	53	40	value	value	NOUN
fcis-17186	53	41	function	function	NOUN
fcis-17186	53	42	with	with	ADP
fcis-17186	53	43	a	a	DET
fcis-17186	53	44	continuously	continuously	ADV
fcis-17186	53	45	differentiable	differentiable	ADJ
fcis-17186	53	46	function	function	NOUN
fcis-17186	53	47	,	,	PUNCT
fcis-17186	53	48	aiding	aid	VERB
fcis-17186	53	49	in	in	ADP
fcis-17186	53	50	better	well	ADJ
fcis-17186	53	51	support	support	NOUN
fcis-17186	53	52	for	for	ADP
fcis-17186	53	53	the	the	DET
fcis-17186	53	54	optimization	optimization	NOUN
fcis-17186	53	55	process	process	NOUN
fcis-17186	53	56	and	and	CCONJ
fcis-17186	53	57	enhancing	enhance	VERB
fcis-17186	53	58	the	the	DET
fcis-17186	53	59	convergence	convergence	NOUN
fcis-17186	53	60	and	and	CCONJ
fcis-17186	53	61	generalization	generalization	NOUN
fcis-17186	53	62	of	of	ADP
fcis-17186	53	63	the	the	DET
fcis-17186	53	64	model	model	NOUN
fcis-17186	53	65	during	during	ADP
fcis-17186	53	66	training	training	NOUN
fcis-17186	53	67	.	.	PUNCT
fcis-17186	54	1	given	give	VERB
fcis-17186	54	2	a	a	DET
fcis-17186	54	3	training	training	NOUN
fcis-17186	54	4	sample	sample	NOUN
fcis-17186	54	5	1	1	NUM
fcis-17186	54	6	{	{	PUNCT
fcis-17186	54	7	,	,	PUNCT
fcis-17186	54	8	}	}	PUNCT
fcis-17186	54	9	j	j	PROPN
fcis-17186	54	10	j	j	PROPN
fcis-17186	55	1	j	j	PROPN
fcis-17186	55	2	jx	jx	PROPN
fcis-17186	56	1	o	o	PROPN
fcis-17186	57	1			NOUN
fcis-17186	57	2	,	,	PUNCT
fcis-17186	57	3	where	where	SCONJ
fcis-17186	57	4	oj	oj	PART
fcis-17186	57	5	represents	represent	VERB
fcis-17186	57	6	the	the	DET
fcis-17186	57	7	ideal	ideal	ADJ
fcis-17186	57	8	output	output	NOUN
fcis-17186	57	9	,	,	PUNCT
fcis-17186	57	10	we	we	PRON
fcis-17186	57	11	define	define	VERB
fcis-17186	57	12	the	the	DET
fcis-17186	57	13	following	follow	VERB
fcis-17186	57	14	smooth	smooth	ADJ
fcis-17186	57	15	function	function	NOUN
fcis-17186	57	16	:	:	PUNCT
fcis-17186	57	17	2	2	NUM
fcis-17186	57	18	2	2	NUM
fcis-17186	57	19	2	2	NUM
fcis-17186	57	20	,	,	PUNCT
fcis-17186	57	21	(	(	PUNCT
fcis-17186	57	22	,	,	PUNCT
fcis-17186	57	23	)	)	PUNCT
fcis-17186	57	24	,	,	PUNCT
fcis-17186	57	25	,	,	PUNCT
fcis-17186	57	26	k	k	PROPN
fcis-17186	57	27	kk	kk	PROPN
fcis-17186	57	28	k	k	PROPN
fcis-17186	57	29	k	k	PROPN
fcis-17186	57	30	u	u	PROPN
fcis-17186	57	31	uu	uu	PRON
fcis-17186	57	32	h	h	NOUN
fcis-17186	57	33	u	u	NOUN
fcis-17186	57	34	u	u	NOUN
fcis-17186	57	35			NOUN
fcis-17186	57	36			X
fcis-17186	57	37			X
fcis-17186	57	38			X
fcis-17186	57	39			VERB
fcis-17186	57	40			ADV
fcis-17186	57	41			NUM
fcis-17186	57	42			PROPN
fcis-17186	57	43			NUM
fcis-17186	57	44			NUM
fcis-17186	57	45			PROPN
fcis-17186	57	46	(	(	PUNCT
fcis-17186	57	47	5	5	NUM
fcis-17186	57	48	)	)	PUNCT
fcis-17186	57	49	where	where	SCONJ
fcis-17186	57	50	0	0	PROPN
fcis-17186	57	51			VERB
fcis-17186	57	52	,	,	PUNCT
fcis-17186	57	53	the	the	DET
fcis-17186	57	54	weight	weight	NOUN
fcis-17186	57	55	vector	vector	NOUN
fcis-17186	57	56	uk	uk	PROPN
fcis-17186	57	57	represents	represent	VERB
fcis-17186	57	58	the	the	DET
fcis-17186	57	59	vector	vector	NOUN
fcis-17186	57	60	connecting	connect	VERB
fcis-17186	57	61	to	to	ADP
fcis-17186	57	62	the	the	DET
fcis-17186	57	63	k	k	NOUN
fcis-17186	57	64	-	-	PUNCT
fcis-17186	57	65	th	th	X
fcis-17186	57	66	node	node	NOUN
fcis-17186	57	67	of	of	ADP
fcis-17186	57	68	the	the	DET
fcis-17186	57	69	fully	fully	ADV
fcis-17186	57	70	connected	connect	VERB
fcis-17186	57	71	layer	layer	NOUN
fcis-17186	57	72	,	,	PUNCT
fcis-17186	57	73	and	and	CCONJ
fcis-17186	57	74	the	the	DET
fcis-17186	57	75	k	k	NOUN
fcis-17186	57	76	-	-	PUNCT
fcis-17186	57	77	th	th	VERB
fcis-17186	57	78	node	node	NOUN
fcis-17186	57	79	connecting	connect	VERB
fcis-17186	57	80	to	to	ADP
fcis-17186	57	81	all	all	DET
fcis-17186	57	82	output	output	NOUN
fcis-17186	57	83	nodes	node	NOUN
fcis-17186	57	84	in	in	ADP
fcis-17186	57	85	the	the	DET
fcis-17186	57	86	fully	fully	ADV
fcis-17186	57	87	connected	connect	VERB
fcis-17186	57	88	layer	layer	NOUN
fcis-17186	57	89	.	.	PUNCT
fcis-17186	58	1	subsequently	subsequently	ADV
fcis-17186	58	2	,	,	PUNCT
fcis-17186	58	3	the	the	DET
fcis-17186	58	4	h	h	NOUN
fcis-17186	58	5	(	(	PUNCT
fcis-17186	58	6	uk	uk	PROPN
fcis-17186	58	7	,	,	PUNCT
fcis-17186	58	8			X
fcis-17186	58	9	)	)	PUNCT
fcis-17186	58	10	gradient	gradient	NOUN
fcis-17186	58	11	of	of	ADP
fcis-17186	58	12	and	and	CCONJ
fcis-17186	58	13	the	the	DET
fcis-17186	58	14	gradient	gradient	NOUN
fcis-17186	58	15	of	of	ADP
fcis-17186	58	16	the	the	DET
fcis-17186	58	17	vector	vector	NOUN
fcis-17186	58	18	uk	uk	PROPN
fcis-17186	58	19	are	be	AUX
fcis-17186	58	20	as	as	SCONJ
fcis-17186	58	21	follows	follow	VERB
fcis-17186	58	22	:	:	PUNCT
fcis-17186	58	23	,	,	PUNCT
fcis-17186	58	24	(	(	PUNCT
fcis-17186	58	25	,	,	PUNCT
fcis-17186	58	26	)	)	PUNCT
fcis-17186	58	27	,	,	PUNCT
fcis-17186	58	28	,	,	PUNCT
fcis-17186	59	1	k	k	PROPN
fcis-17186	59	2	k	k	PROPN
fcis-17186	60	1	k	k	PROPN
fcis-17186	60	2	k	k	PROPN
fcis-17186	61	1	k	k	PROPN
fcis-17186	61	2	u	u	X
fcis-17186	61	3	k	k	PROPN
fcis-17186	61	4	k	k	X
fcis-17186	61	5	u	u	X
fcis-17186	61	6	u	u	VERB
fcis-17186	61	7	u	u	NOUN
fcis-17186	61	8	u	u	NOUN
fcis-17186	61	9	h	h	NOUN
fcis-17186	61	10	u	u	NOUN
fcis-17186	61	11	u	u	NOUN
fcis-17186	61	12			NOUN
fcis-17186	61	13			X
fcis-17186	61	14			X
fcis-17186	61	15			X
fcis-17186	61	16			ADP
fcis-17186	61	17			ADJ
fcis-17186	61	18			PROPN
fcis-17186	61	19			NUM
fcis-17186	61	20			NUM
fcis-17186	61	21			X
fcis-17186	61	22	(	(	PUNCT
fcis-17186	61	23	6	6	NUM
fcis-17186	61	24	)	)	PUNCT
fcis-17186	61	25	the	the	DET
fcis-17186	61	26	error	error	NOUN
fcis-17186	61	27	function	function	NOUN
fcis-17186	61	28	of	of	ADP
fcis-17186	61	29	the	the	DET
fcis-17186	61	30	convolutional	convolutional	ADJ
fcis-17186	61	31	neural	neural	ADJ
fcis-17186	61	32	network	network	NOUN
fcis-17186	61	33	based	base	VERB
fcis-17186	61	34	on	on	ADP
fcis-17186	61	35	l1	l1	PROPN
fcis-17186	61	36	smooth	smooth	ADJ
fcis-17186	61	37	regularization	regularization	NOUN
fcis-17186	61	38	of	of	ADP
fcis-17186	61	39	the	the	DET
fcis-17186	61	40	entropy	entropy	NOUN
fcis-17186	61	41	error	error	NOUN
fcis-17186	61	42	function	function	NOUN
fcis-17186	61	43	,	,	PUNCT
fcis-17186	61	44	as	as	SCONJ
fcis-17186	61	45	indicated	indicate	VERB
fcis-17186	61	46	by	by	ADP
fcis-17186	61	47	the	the	DET
fcis-17186	61	48	above	above	ADJ
fcis-17186	61	49	formula	formula	NOUN
fcis-17186	61	50	,	,	PUNCT
fcis-17186	61	51	is	be	AUX
fcis-17186	61	52	defined	define	VERB
fcis-17186	61	53	as	as	SCONJ
fcis-17186	61	54	follows	follow	VERB
fcis-17186	61	55	:	:	PUNCT
fcis-17186	61	56	10	10	NUM
fcis-17186	61	57	1	1	NUM
fcis-17186	61	58	1	1	NUM
fcis-17186	61	59	1	1	NUM
fcis-17186	61	60	1	1	NUM
fcis-17186	61	61	(	(	PUNCT
fcis-17186	61	62	)	)	PUNCT
fcis-17186	61	63	1	1	NUM
fcis-17186	61	64	(	(	PUNCT
fcis-17186	61	65	)	)	PUNCT
fcis-17186	61	66	(	(	PUNCT
fcis-17186	61	67	)	)	PUNCT
fcis-17186	62	1	[	[	PUNCT
fcis-17186	62	2	ln	ln	X
fcis-17186	62	3	(	(	PUNCT
fcis-17186	62	4	1	1	NUM
fcis-17186	62	5	)	)	PUNCT
fcis-17186	62	6	ln	ln	NOUN
fcis-17186	62	7	]	]	PUNCT
fcis-17186	62	8	1	1	NUM
fcis-17186	62	9	(	(	PUNCT
fcis-17186	62	10	)	)	PUNCT
fcis-17186	62	11	.	.	PUNCT
fcis-17186	63	1	j	j	PROPN
fcis-17186	63	2	jj	jj	PROPN
fcis-17186	63	3	j	j	PROPN
fcis-17186	64	1	ji	ji	INTJ
fcis-17186	65	1	i	i	PRON
fcis-17186	65	2	j	j	PROPN
fcis-17186	66	1	j	j	PROPN
fcis-17186	66	2	j	j	PROPN
fcis-17186	67	1	i	i	PRON
fcis-17186	67	2	qj	qj	PROPN
fcis-17186	67	3	j	j	PROPN
fcis-17186	68	1	k	k	PROPN
fcis-17186	68	2	j	j	PROPN
fcis-17186	69	1	k	k	PROPN
fcis-17186	69	2	g	g	PROPN
fcis-17186	69	3	u	u	PROPN
fcis-17186	69	4	s	s	PROPN
fcis-17186	69	5	g	g	NOUN
fcis-17186	69	6	u	u	NOUN
fcis-17186	69	7	s	s	PROPN
fcis-17186	69	8	e	e	NOUN
fcis-17186	69	9	w	w	NOUN
fcis-17186	69	10	h	h	NOUN
fcis-17186	69	11	u	u	NOUN
fcis-17186	69	12	x	x	NOUN
fcis-17186	69	13			NOUN
fcis-17186	69	14			NOUN
fcis-17186	69	15			NOUN
fcis-17186	69	16			NOUN
fcis-17186	69	17			ADJ
fcis-17186	69	18			PROPN
fcis-17186	69	19			NUM
fcis-17186	69	20			NUM
fcis-17186	69	21			PROPN
fcis-17186	69	22			ADJ
fcis-17186	69	23			NOUN
fcis-17186	69	24			NUM
fcis-17186	69	25			NUM
fcis-17186	69	26			PROPN
fcis-17186	69	27			PUNCT
fcis-17186	69	28			PROPN
fcis-17186	69	29			PROPN
fcis-17186	69	30			CCONJ
fcis-17186	69	31			PROPN
fcis-17186	69	32			X
fcis-17186	69	33			X
fcis-17186	69	34	(	(	PUNCT
fcis-17186	69	35	7	7	NUM
fcis-17186	69	36	)	)	SYM
fcis-17186	69	37	3	3	NUM
fcis-17186	69	38	.	.	PUNCT
fcis-17186	69	39	numerical	numerical	ADJ
fcis-17186	69	40	experiments	experiment	NOUN
fcis-17186	69	41	(	(	PUNCT
fcis-17186	69	42	1	1	X
fcis-17186	69	43	)	)	PUNCT
fcis-17186	69	44	dataset	dataset	NOUN
fcis-17186	69	45	introduction	introduction	NOUN
fcis-17186	69	46	the	the	DET
fcis-17186	69	47	dataset	dataset	NOUN
fcis-17186	69	48	is	be	AUX
fcis-17186	69	49	a	a	DET
fcis-17186	69	50	classic	classic	ADJ
fcis-17186	69	51	collection	collection	NOUN
fcis-17186	69	52	of	of	ADP
fcis-17186	69	53	handwritten	handwritten	ADJ
fcis-17186	69	54	digit	digit	NOUN
fcis-17186	69	55	images	image	NOUN
fcis-17186	69	56	,	,	PUNCT
fcis-17186	69	57	used	use	VERB
fcis-17186	69	58	for	for	ADP
fcis-17186	69	59	machine	machine	NOUN
fcis-17186	69	60	learning	learn	VERB
fcis-17186	69	61	training	training	NOUN
fcis-17186	69	62	and	and	CCONJ
fcis-17186	69	63	testing	testing	NOUN
fcis-17186	69	64	.	.	PUNCT
fcis-17186	70	1	it	it	PRON
fcis-17186	70	2	comprises	comprise	VERB
fcis-17186	70	3	60,000	60,000	NUM
fcis-17186	70	4	training	training	NOUN
fcis-17186	70	5	images	image	NOUN
fcis-17186	70	6	and	and	CCONJ
fcis-17186	70	7	10,000	10,000	NUM
fcis-17186	70	8	28×28	28×28	NUM
fcis-17186	70	9	-	-	PUNCT
fcis-17186	70	10	pixel	pixel	ADJ
fcis-17186	70	11	test	test	NOUN
fcis-17186	70	12	images	image	NOUN
fcis-17186	70	13	.	.	PUNCT
fcis-17186	71	1	through	through	ADP
fcis-17186	71	2	training	train	VERB
fcis-17186	71	3	our	our	PRON
fcis-17186	71	4	model	model	NOUN
fcis-17186	71	5	,	,	PUNCT
fcis-17186	71	6	we	we	PRON
fcis-17186	71	7	can	can	AUX
fcis-17186	71	8	achieve	achieve	VERB
fcis-17186	71	9	accurate	accurate	ADJ
fcis-17186	71	10	recognition	recognition	NOUN
fcis-17186	71	11	of	of	ADP
fcis-17186	71	12	handwritten	handwritten	ADJ
fcis-17186	71	13	digits	digit	NOUN
fcis-17186	71	14	.	.	PUNCT
fcis-17186	72	1	(	(	PUNCT
fcis-17186	72	2	2	2	X
fcis-17186	72	3	)	)	PUNCT
fcis-17186	72	4	innparameter	innparameter	NOUN
fcis-17186	72	5	configuration	configuration	NOUN
fcis-17186	72	6	in	in	ADP
fcis-17186	72	7	our	our	PRON
fcis-17186	72	8	convolutional	convolutional	ADJ
fcis-17186	72	9	neural	neural	ADJ
fcis-17186	72	10	network	network	NOUN
fcis-17186	72	11	,	,	PUNCT
fcis-17186	72	12	the	the	DET
fcis-17186	72	13	convolutional	convolutional	ADJ
fcis-17186	72	14	layer	layer	NOUN
fcis-17186	72	15	extracts	extract	NOUN
fcis-17186	72	16	feature	feature	NOUN
fcis-17186	72	17	information	information	NOUN
fcis-17186	72	18	from	from	ADP
fcis-17186	72	19	the	the	DET
fcis-17186	72	20	input	input	NOUN
fcis-17186	72	21	image	image	NOUN
fcis-17186	72	22	by	by	ADP
fcis-17186	72	23	performing	perform	VERB
fcis-17186	72	24	element	element	ADJ
fcis-17186	72	25	-	-	ADJ
fcis-17186	72	26	wise	wise	ADJ
fcis-17186	72	27	multiplication	multiplication	NOUN
fcis-17186	72	28	and	and	CCONJ
fcis-17186	72	29	summation	summation	NOUN
fcis-17186	72	30	operations	operation	NOUN
fcis-17186	72	31	using	use	VERB
fcis-17186	72	32	a	a	DET
fcis-17186	72	33	5×5	5×5	NUM
fcis-17186	72	34	convolutional	convolutional	ADJ
fcis-17186	72	35	kernel	kernel	NOUN
fcis-17186	72	36	.	.	PUNCT
fcis-17186	73	1	to	to	PART
fcis-17186	73	2	maintain	maintain	VERB
fcis-17186	73	3	the	the	DET
fcis-17186	73	4	spatial	spatial	ADJ
fcis-17186	73	5	dimensions	dimension	NOUN
fcis-17186	73	6	of	of	ADP
fcis-17186	73	7	the	the	DET
fcis-17186	73	8	output	output	NOUN
fcis-17186	73	9	,	,	PUNCT
fcis-17186	73	10	we	we	PRON
fcis-17186	73	11	have	have	AUX
fcis-17186	73	12	chosen	choose	VERB
fcis-17186	73	13	a	a	DET
fcis-17186	73	14	stride	stride	NOUN
fcis-17186	73	15	of	of	ADP
fcis-17186	73	16	1	1	NUM
fcis-17186	73	17	71	71	NUM
fcis-17186	73	18	and	and	CCONJ
fcis-17186	73	19	zero	zero	NUM
fcis-17186	73	20	-	-	PUNCT
fcis-17186	73	21	padding	padding	NOUN
fcis-17186	73	22	.	.	PUNCT
fcis-17186	74	1	the	the	DET
fcis-17186	74	2	pooling	pool	VERB
fcis-17186	74	3	layer	layer	NOUN
fcis-17186	74	4	employs	employ	VERB
fcis-17186	74	5	2×2	2×2	NUM
fcis-17186	74	6	average	average	ADJ
fcis-17186	74	7	pooling	pooling	NOUN
fcis-17186	74	8	to	to	PART
fcis-17186	74	9	reduce	reduce	VERB
fcis-17186	74	10	data	data	NOUN
fcis-17186	74	11	dimensions	dimension	NOUN
fcis-17186	74	12	and	and	CCONJ
fcis-17186	74	13	enhance	enhance	VERB
fcis-17186	74	14	the	the	DET
fcis-17186	74	15	model	model	NOUN
fcis-17186	74	16	's	's	PART
fcis-17186	74	17	robustness	robustness	NOUN
fcis-17186	74	18	.	.	PUNCT
fcis-17186	75	1	the	the	DET
fcis-17186	75	2	softmax	softmax	NOUN
fcis-17186	75	3	activation	activation	NOUN
fcis-17186	75	4	function	function	NOUN
fcis-17186	75	5	is	be	AUX
fcis-17186	75	6	used	use	VERB
fcis-17186	75	7	to	to	PART
fcis-17186	75	8	transform	transform	VERB
fcis-17186	75	9	the	the	DET
fcis-17186	75	10	neural	neural	ADJ
fcis-17186	75	11	network	network	NOUN
fcis-17186	75	12	output	output	NOUN
fcis-17186	75	13	into	into	ADP
fcis-17186	75	14	a	a	DET
fcis-17186	75	15	probability	probability	NOUN
fcis-17186	75	16	distribution	distribution	NOUN
fcis-17186	75	17	.	.	PUNCT
fcis-17186	76	1	during	during	ADP
fcis-17186	76	2	the	the	DET
fcis-17186	76	3	parameter	parameter	NOUN
fcis-17186	76	4	update	update	NOUN
fcis-17186	76	5	phase	phase	NOUN
fcis-17186	76	6	,	,	PUNCT
fcis-17186	76	7	we	we	PRON
fcis-17186	76	8	introduce	introduce	VERB
fcis-17186	76	9	a	a	DET
fcis-17186	76	10	momentum	momentum	NOUN
fcis-17186	76	11	parameter	parameter	NOUN
fcis-17186	76	12	set	set	VERB
fcis-17186	76	13	to	to	ADP
fcis-17186	76	14	0.2	0.2	NUM
fcis-17186	76	15	,	,	PUNCT
fcis-17186	76	16	which	which	PRON
fcis-17186	76	17	aids	aid	VERB
fcis-17186	76	18	in	in	ADP
fcis-17186	76	19	accelerating	accelerate	VERB
fcis-17186	76	20	convergence	convergence	NOUN
fcis-17186	76	21	,	,	PUNCT
fcis-17186	76	22	especially	especially	ADV
fcis-17186	76	23	in	in	ADP
fcis-17186	76	24	the	the	DET
fcis-17186	76	25	early	early	ADJ
fcis-17186	76	26	stages	stage	NOUN
fcis-17186	76	27	of	of	ADP
fcis-17186	76	28	training	training	NOUN
fcis-17186	76	29	.	.	PUNCT
fcis-17186	77	1	interestingly	interestingly	ADV
fcis-17186	77	2	,	,	PUNCT
fcis-17186	77	3	in	in	ADP
fcis-17186	77	4	the	the	DET
fcis-17186	77	5	initial	initial	ADJ
fcis-17186	77	6	20	20	NUM
fcis-17186	77	7	iterations	iteration	NOUN
fcis-17186	77	8	,	,	PUNCT
fcis-17186	77	9	we	we	PRON
fcis-17186	77	10	gradually	gradually	ADV
fcis-17186	77	11	increase	increase	VERB
fcis-17186	77	12	the	the	DET
fcis-17186	77	13	momentum	momentum	NOUN
fcis-17186	77	14	value	value	NOUN
fcis-17186	77	15	to	to	PART
fcis-17186	77	16	better	well	ADV
fcis-17186	77	17	explore	explore	VERB
fcis-17186	77	18	local	local	ADJ
fcis-17186	77	19	minima	minima	NOUN
fcis-17186	77	20	of	of	ADP
fcis-17186	77	21	the	the	DET
fcis-17186	77	22	loss	loss	NOUN
fcis-17186	77	23	function	function	NOUN
fcis-17186	77	24	and	and	CCONJ
fcis-17186	77	25	then	then	ADV
fcis-17186	77	26	rapidly	rapidly	ADV
fcis-17186	77	27	converge	converge	VERB
fcis-17186	77	28	towards	towards	ADP
fcis-17186	77	29	the	the	DET
fcis-17186	77	30	global	global	ADJ
fcis-17186	77	31	minimum	minimum	NOUN
fcis-17186	77	32	.	.	PUNCT
fcis-17186	78	1	the	the	DET
fcis-17186	78	2	experimental	experimental	ADJ
fcis-17186	78	3	parameter	parameter	NOUN
fcis-17186	78	4	settings	setting	NOUN
fcis-17186	78	5	are	be	AUX
fcis-17186	78	6	presented	present	VERB
fcis-17186	78	7	in	in	ADP
fcis-17186	78	8	the	the	DET
fcis-17186	78	9	table	table	NOUN
fcis-17186	78	10	below	below	ADV
fcis-17186	78	11	.	.	PUNCT
fcis-17186	79	1	table	table	NOUN
fcis-17186	79	2	1	1	NUM
fcis-17186	79	3	.	.	PUNCT
fcis-17186	79	4	model	model	NOUN
fcis-17186	79	5	parameter	parameter	PROPN
fcis-17186	79	6	settings	setting	NOUN
fcis-17186	79	7	size	size	VERB
fcis-17186	79	8	operations	operation	NOUN
fcis-17186	79	9	28×28×6	28×28×6	NUM
fcis-17186	79	10	convolutional	convolutional	ADJ
fcis-17186	79	11	stride	stride	NOUN
fcis-17186	79	12	=	=	SYM
fcis-17186	79	13	1	1	NUM
fcis-17186	79	14	,	,	PUNCT
fcis-17186	79	15	padding	padding	NOUN
fcis-17186	79	16	=	=	SYM
fcis-17186	79	17	0	0	NUM
fcis-17186	79	18	24×24×6	24×24×6	NUM
fcis-17186	79	19	2×2	2×2	NUM
fcis-17186	79	20	average	average	ADJ
fcis-17186	79	21	pooling	pooling	NOUN
fcis-17186	79	22	with	with	ADP
fcis-17186	79	23	relu	relu	NOUN
fcis-17186	79	24	activation	activation	NOUN
fcis-17186	79	25	function	function	NOUN
fcis-17186	79	26	12×12×12	12×12×12	PUNCT
fcis-17186	79	27	convolution	convolution	NOUN
fcis-17186	79	28	stride	stride	NOUN
fcis-17186	79	29	=	=	SYM
fcis-17186	79	30	1	1	NUM
fcis-17186	79	31	,	,	PUNCT
fcis-17186	79	32	padding	padding	NOUN
fcis-17186	79	33	=	=	SYM
fcis-17186	79	34	0	0	NUM
fcis-17186	79	35	8×8×12	8×8×12	NUM
fcis-17186	79	36	2×2	2×2	NUM
fcis-17186	79	37	average	average	ADJ
fcis-17186	79	38	pooling	pooling	NOUN
fcis-17186	79	39	with	with	ADP
fcis-17186	79	40	relu	relu	NOUN
fcis-17186	79	41	activation	activation	NOUN
fcis-17186	79	42	function	function	NOUN
fcis-17186	79	43	16×1	16×1	PROPN
fcis-17186	79	44	fully	fully	ADV
fcis-17186	79	45	connected	connected	ADJ
fcis-17186	79	46	192×1	192×1	NUM
fcis-17186	79	47	fully	fully	ADV
fcis-17186	79	48	connected	connect	VERB
fcis-17186	79	49	10×192	10×192	NUM
fcis-17186	79	50	fully	fully	ADV
fcis-17186	79	51	connected	connect	VERB
fcis-17186	79	52	(	(	PUNCT
fcis-17186	79	53	3	3	X
fcis-17186	79	54	)	)	PUNCT
fcis-17186	79	55	results	result	VERB
fcis-17186	79	56	analysis	analysis	NOUN
fcis-17186	79	57	figure	figure	NOUN
fcis-17186	79	58	2	2	NUM
fcis-17186	79	59	.	.	PUNCT
fcis-17186	79	60	network	network	NOUN
fcis-17186	79	61	visualization	visualization	NOUN
fcis-17186	79	62	figure	figure	NOUN
fcis-17186	79	63	3	3	NUM
fcis-17186	79	64	.	.	PUNCT
fcis-17186	79	65	loss	loss	NOUN
fcis-17186	79	66	function	function	NOUN
fcis-17186	79	67	plot	plot	NOUN
fcis-17186	79	68	in	in	ADP
fcis-17186	79	69	figure	figure	NOUN
fcis-17186	79	70	2	2	NUM
fcis-17186	79	71	,	,	PUNCT
fcis-17186	79	72	we	we	PRON
fcis-17186	79	73	present	present	VERB
fcis-17186	79	74	a	a	DET
fcis-17186	79	75	detailed	detailed	ADJ
fcis-17186	79	76	visualization	visualization	NOUN
fcis-17186	79	77	of	of	ADP
fcis-17186	79	78	handwritten	handwritten	ADJ
fcis-17186	79	79	digit	digit	NOUN
fcis-17186	79	80	recognition	recognition	NOUN
fcis-17186	79	81	.	.	PUNCT
fcis-17186	80	1	figure	figure	VERB
fcis-17186	80	2	3	3	NUM
fcis-17186	80	3	further	far	ADV
fcis-17186	80	4	compares	compare	VERB
fcis-17186	80	5	the	the	DET
fcis-17186	80	6	error	error	NOUN
fcis-17186	80	7	functions	function	NOUN
fcis-17186	80	8	of	of	ADP
fcis-17186	80	9	different	different	ADJ
fcis-17186	80	10	algorithms	algorithm	NOUN
fcis-17186	80	11	,	,	PUNCT
fcis-17186	80	12	with	with	ADP
fcis-17186	80	13	a	a	DET
fcis-17186	80	14	particular	particular	ADJ
fcis-17186	80	15	focus	focus	NOUN
fcis-17186	80	16	on	on	ADP
fcis-17186	80	17	the	the	DET
fcis-17186	80	18	outstanding	outstanding	ADJ
fcis-17186	80	19	performance	performance	NOUN
fcis-17186	80	20	of	of	ADP
fcis-17186	80	21	the	the	DET
fcis-17186	80	22	sl1	sl1	PROPN
fcis-17186	80	23	algorithm	algorithm	NOUN
fcis-17186	80	24	in	in	ADP
fcis-17186	80	25	enhancing	enhance	VERB
fcis-17186	80	26	network	network	NOUN
fcis-17186	80	27	performance	performance	NOUN
fcis-17186	80	28	.	.	PUNCT
fcis-17186	81	1	the	the	DET
fcis-17186	81	2	figure	figure	NOUN
fcis-17186	81	3	illustrates	illustrate	VERB
fcis-17186	81	4	the	the	DET
fcis-17186	81	5	rapid	rapid	ADJ
fcis-17186	81	6	reduction	reduction	NOUN
fcis-17186	81	7	of	of	ADP
fcis-17186	81	8	network	network	NOUN
fcis-17186	81	9	error	error	NOUN
fcis-17186	81	10	during	during	ADP
fcis-17186	81	11	the	the	DET
fcis-17186	81	12	early	early	ADJ
fcis-17186	81	13	stages	stage	NOUN
fcis-17186	81	14	of	of	ADP
fcis-17186	81	15	training	training	NOUN
fcis-17186	81	16	using	use	VERB
fcis-17186	81	17	the	the	DET
fcis-17186	81	18	sl1	sl1	PROPN
fcis-17186	81	19	algorithm	algorithm	NOUN
fcis-17186	81	20	,	,	PUNCT
fcis-17186	81	21	ultimately	ultimately	ADV
fcis-17186	81	22	stabilizing	stabilize	VERB
fcis-17186	81	23	and	and	CCONJ
fcis-17186	81	24	exhibiting	exhibit	VERB
fcis-17186	81	25	a	a	DET
fcis-17186	81	26	gradual	gradual	ADJ
fcis-17186	81	27	convergence	convergence	NOUN
fcis-17186	81	28	trend	trend	NOUN
fcis-17186	81	29	,	,	PUNCT
fcis-17186	81	30	while	while	SCONJ
fcis-17186	81	31	other	other	ADJ
fcis-17186	81	32	algorithms	algorithm	NOUN
fcis-17186	81	33	show	show	VERB
fcis-17186	81	34	relatively	relatively	ADV
fcis-17186	81	35	weaker	weak	ADJ
fcis-17186	81	36	performance	performance	NOUN
fcis-17186	81	37	.	.	PUNCT
fcis-17186	82	1	this	this	PRON
fcis-17186	82	2	clearly	clearly	ADV
fcis-17186	82	3	indicates	indicate	VERB
fcis-17186	82	4	a	a	DET
fcis-17186	82	5	significant	significant	ADJ
fcis-17186	82	6	advantage	advantage	NOUN
fcis-17186	82	7	of	of	ADP
fcis-17186	82	8	the	the	DET
fcis-17186	82	9	sl1	sl1	PROPN
fcis-17186	82	10	algorithm	algorithm	NOUN
fcis-17186	82	11	over	over	ADP
fcis-17186	82	12	others	other	NOUN
fcis-17186	82	13	in	in	ADP
fcis-17186	82	14	terms	term	NOUN
fcis-17186	82	15	of	of	ADP
fcis-17186	82	16	error	error	NOUN
fcis-17186	82	17	reduction	reduction	NOUN
fcis-17186	82	18	speed	speed	NOUN
fcis-17186	82	19	and	and	CCONJ
fcis-17186	82	20	convergence	convergence	NOUN
fcis-17186	82	21	.	.	PUNCT
fcis-17186	83	1	through	through	ADP
fcis-17186	83	2	the	the	DET
fcis-17186	83	3	incorporation	incorporation	NOUN
fcis-17186	83	4	of	of	ADP
fcis-17186	83	5	smooth	smooth	ADJ
fcis-17186	83	6	l1	l1	PROPN
fcis-17186	83	7	regularization	regularization	NOUN
fcis-17186	83	8	and	and	CCONJ
fcis-17186	83	9	entropy	entropy	NOUN
fcis-17186	83	10	error	error	NOUN
fcis-17186	83	11	functions	function	NOUN
fcis-17186	83	12	,	,	PUNCT
fcis-17186	83	13	the	the	DET
fcis-17186	83	14	sl1	sl1	PROPN
fcis-17186	83	15	algorithm	algorithm	NOUN
fcis-17186	83	16	achieves	achieve	VERB
fcis-17186	83	17	higher	high	ADJ
fcis-17186	83	18	precision	precision	NOUN
fcis-17186	83	19	during	during	ADP
fcis-17186	83	20	the	the	DET
fcis-17186	83	21	model	model	NOUN
fcis-17186	83	22	inference	inference	NOUN
fcis-17186	83	23	phase	phase	NOUN
fcis-17186	83	24	,	,	PUNCT
fcis-17186	83	25	successfully	successfully	ADV
fcis-17186	83	26	enhancing	enhance	VERB
fcis-17186	83	27	the	the	DET
fcis-17186	83	28	accuracy	accuracy	NOUN
fcis-17186	83	29	of	of	ADP
fcis-17186	83	30	handwritten	handwritten	ADJ
fcis-17186	83	31	digit	digit	NOUN
fcis-17186	83	32	recognition	recognition	NOUN
fcis-17186	83	33	.	.	PUNCT
fcis-17186	84	1	4	4	X
fcis-17186	84	2	.	.	X
fcis-17186	84	3	conclusion	conclusion	NOUN
fcis-17186	84	4	this	this	DET
fcis-17186	84	5	study	study	NOUN
fcis-17186	84	6	has	have	AUX
fcis-17186	84	7	achieved	achieve	VERB
fcis-17186	84	8	success	success	NOUN
fcis-17186	84	9	in	in	ADP
fcis-17186	84	10	optimizing	optimize	VERB
fcis-17186	84	11	convolutional	convolutional	ADJ
fcis-17186	84	12	neural	neural	ADJ
fcis-17186	84	13	networks	network	NOUN
fcis-17186	84	14	(	(	PUNCT
fcis-17186	84	15	cnns	cnns	PROPN
fcis-17186	84	16	)	)	PUNCT
fcis-17186	84	17	by	by	ADP
fcis-17186	84	18	introducing	introduce	VERB
fcis-17186	84	19	a	a	DET
fcis-17186	84	20	smoothed	smooth	VERB
fcis-17186	84	21	l1	l1	PROPN
fcis-17186	84	22	regularization	regularization	NOUN
fcis-17186	84	23	term	term	NOUN
fcis-17186	84	24	and	and	CCONJ
fcis-17186	84	25	an	an	DET
fcis-17186	84	26	entropy	entropy	NOUN
fcis-17186	84	27	error	error	NOUN
fcis-17186	84	28	function	function	NOUN
fcis-17186	84	29	.	.	PUNCT
fcis-17186	85	1	the	the	DET
fcis-17186	85	2	smoothed	smooth	VERB
fcis-17186	85	3	l1	l1	PROPN
fcis-17186	85	4	regularization	regularization	NOUN
fcis-17186	85	5	effectively	effectively	ADV
fcis-17186	85	6	prunes	prune	VERB
fcis-17186	85	7	relevant	relevant	ADJ
fcis-17186	85	8	nodes	node	NOUN
fcis-17186	85	9	,	,	PUNCT
fcis-17186	85	10	significantly	significantly	ADV
fcis-17186	85	11	enhancing	enhance	VERB
fcis-17186	85	12	the	the	DET
fcis-17186	85	13	network	network	NOUN
fcis-17186	85	14	's	's	PART
fcis-17186	85	15	computational	computational	ADJ
fcis-17186	85	16	efficiency	efficiency	NOUN
fcis-17186	85	17	and	and	CCONJ
fcis-17186	85	18	generalization	generalization	NOUN
fcis-17186	85	19	performance	performance	NOUN
fcis-17186	85	20	.	.	PUNCT
fcis-17186	86	1	in	in	ADP
fcis-17186	86	2	comparison	comparison	NOUN
fcis-17186	86	3	to	to	ADP
fcis-17186	86	4	traditional	traditional	ADJ
fcis-17186	86	5	regularization	regularization	NOUN
fcis-17186	86	6	methods	method	NOUN
fcis-17186	86	7	,	,	PUNCT
fcis-17186	86	8	it	it	PRON
fcis-17186	86	9	provides	provide	VERB
fcis-17186	86	10	a	a	DET
fcis-17186	86	11	more	more	ADV
fcis-17186	86	12	robust	robust	ADJ
fcis-17186	86	13	mathematical	mathematical	ADJ
fcis-17186	86	14	foundation	foundation	NOUN
fcis-17186	86	15	for	for	ADP
fcis-17186	86	16	neural	neural	ADJ
fcis-17186	86	17	network	network	NOUN
fcis-17186	86	18	training	training	NOUN
fcis-17186	86	19	.	.	PUNCT
fcis-17186	87	1	simultaneously	simultaneously	ADV
fcis-17186	87	2	,	,	PUNCT
fcis-17186	87	3	the	the	DET
fcis-17186	87	4	application	application	NOUN
fcis-17186	87	5	of	of	ADP
fcis-17186	87	6	the	the	DET
fcis-17186	87	7	entropy	entropy	NOUN
fcis-17186	87	8	error	error	NOUN
fcis-17186	87	9	function	function	NOUN
fcis-17186	87	10	enhances	enhance	VERB
fcis-17186	87	11	the	the	DET
fcis-17186	87	12	assessment	assessment	NOUN
fcis-17186	87	13	of	of	ADP
fcis-17186	87	14	differences	difference	NOUN
fcis-17186	87	15	between	between	ADP
fcis-17186	87	16	model	model	NOUN
fcis-17186	87	17	outputs	output	NOUN
fcis-17186	87	18	and	and	CCONJ
fcis-17186	87	19	label	label	NOUN
fcis-17186	87	20	distributions	distribution	NOUN
fcis-17186	87	21	.	.	PUNCT
fcis-17186	88	1	a	a	DET
fcis-17186	88	2	series	series	NOUN
fcis-17186	88	3	of	of	ADP
fcis-17186	88	4	experiments	experiment	NOUN
fcis-17186	88	5	demonstrate	demonstrate	VERB
fcis-17186	88	6	that	that	SCONJ
fcis-17186	88	7	our	our	PRON
fcis-17186	88	8	proposed	propose	VERB
fcis-17186	88	9	algorithm	algorithm	NOUN
fcis-17186	88	10	effectively	effectively	ADV
fcis-17186	88	11	balances	balance	VERB
fcis-17186	88	12	algorithm	algorithm	NOUN
fcis-17186	88	13	development	development	NOUN
fcis-17186	88	14	and	and	CCONJ
fcis-17186	88	15	exploration	exploration	NOUN
fcis-17186	88	16	.	.	PUNCT
fcis-17186	89	1	results	result	NOUN
fcis-17186	89	2	from	from	ADP
fcis-17186	89	3	classification	classification	NOUN
fcis-17186	89	4	tests	test	NOUN
fcis-17186	89	5	on	on	ADP
fcis-17186	89	6	the	the	DET
fcis-17186	89	7	mnist	mnist	NOUN
fcis-17186	89	8	dataset	dataset	PROPN
fcis-17186	89	9	indicate	indicate	VERB
fcis-17186	89	10	that	that	SCONJ
fcis-17186	89	11	the	the	DET
fcis-17186	89	12	algorithm	algorithm	NOUN
fcis-17186	89	13	exhibits	exhibit	VERB
fcis-17186	89	14	good	good	ADJ
fcis-17186	89	15	generalization	generalization	NOUN
fcis-17186	89	16	performance	performance	NOUN
fcis-17186	89	17	and	and	CCONJ
fcis-17186	89	18	stability	stability	NOUN
fcis-17186	89	19	.	.	PUNCT
fcis-17186	90	1	this	this	PRON
fcis-17186	90	2	suggests	suggest	VERB
fcis-17186	90	3	that	that	SCONJ
fcis-17186	90	4	the	the	DET
fcis-17186	90	5	optimization	optimization	NOUN
fcis-17186	90	6	approach	approach	NOUN
fcis-17186	90	7	,	,	PUNCT
fcis-17186	90	8	considering	consider	VERB
fcis-17186	90	9	both	both	CCONJ
fcis-17186	90	10	the	the	DET
fcis-17186	90	11	smoothed	smooth	VERB
fcis-17186	90	12	l1	l1	PROPN
fcis-17186	90	13	regularization	regularization	NOUN
fcis-17186	90	14	term	term	NOUN
fcis-17186	90	15	and	and	CCONJ
fcis-17186	90	16	entropy	entropy	NOUN
fcis-17186	90	17	error	error	NOUN
fcis-17186	90	18	function	function	NOUN
fcis-17186	90	19	,	,	PUNCT
fcis-17186	90	20	is	be	AUX
fcis-17186	90	21	successful	successful	ADJ
fcis-17186	90	22	and	and	CCONJ
fcis-17186	90	23	viable	viable	ADJ
fcis-17186	90	24	for	for	ADP
fcis-17186	90	25	improving	improve	VERB
fcis-17186	90	26	cnn	cnn	PROPN
fcis-17186	90	27	performance	performance	NOUN
fcis-17186	90	28	.	.	PUNCT
fcis-17186	91	1	references	reference	NOUN
fcis-17186	91	2	[	[	X
fcis-17186	91	3	1	1	NUM
fcis-17186	91	4	]	]	PUNCT
fcis-17186	91	5	leon	leon	PROPN
fcis-17186	91	6	.	.	PUNCT
fcis-17186	92	1	cnn	cnn	PROPN
fcis-17186	92	2	:	:	PUNCT
fcis-17186	92	3	a	a	DET
fcis-17186	92	4	vision	vision	NOUN
fcis-17186	92	5	of	of	ADP
fcis-17186	92	6	complexity	complexity	NOUN
fcis-17186	92	7	,	,	PUNCT
fcis-17186	92	8	international	international	ADJ
fcis-17186	92	9	journal	journal	NOUN
fcis-17186	92	10	of	of	ADP
fcis-17186	92	11	bifurcation	bifurcation	NOUN
fcis-17186	92	12	and	and	CCONJ
fcis-17186	92	13	chaos	chaos	NOUN
fcis-17186	92	14	,	,	PUNCT
fcis-17186	92	15	7	7	NUM
fcis-17186	92	16	(	(	PUNCT
fcis-17186	92	17	1997	1997	NUM
fcis-17186	92	18	)	)	PUNCT
fcis-17186	92	19	:	:	PUNCT
fcis-17186	92	20	2219	2219	NUM
fcis-17186	92	21	-	-	SYM
fcis-17186	92	22	2425	2425	NUM
fcis-17186	92	23	.	.	PUNCT
fcis-17186	93	1	[	[	X
fcis-17186	93	2	2	2	NUM
fcis-17186	93	3	]	]	PUNCT
fcis-17186	93	4	b.	b.	PROPN
fcis-17186	93	5	t.	t.	PROPN
fcis-17186	93	6	li	li	PROPN
fcis-17186	93	7	.	.	PUNCT
fcis-17186	94	1	learning	learn	VERB
fcis-17186	94	2	deep	deep	ADJ
fcis-17186	94	3	neural	neural	ADJ
fcis-17186	94	4	networks	network	NOUN
fcis-17186	94	5	for	for	ADP
fcis-17186	94	6	node	node	ADJ
fcis-17186	94	7	classification	classification	NOUN
fcis-17186	94	8	,	,	PUNCT
fcis-17186	94	9	expert	expert	NOUN
fcis-17186	94	10	systems	system	NOUN
fcis-17186	94	11	with	with	ADP
fcis-17186	94	12	applications	application	NOUN
fcis-17186	94	13	,	,	PUNCT
fcis-17186	94	14	137(2019	137(2019	NUM
fcis-17186	94	15	):	):	PUNCT
fcis-17186	94	16	324	324	NUM
fcis-17186	94	17	-	-	SYM
fcis-17186	94	18	334	334	NUM
fcis-17186	94	19	.	.	PUNCT
fcis-17186	95	1	[	[	X
fcis-17186	95	2	3	3	X
fcis-17186	95	3	]	]	X
fcis-17186	95	4	f.	f.	PROPN
fcis-17186	95	5	ghasemi	ghasemi	PROPN
fcis-17186	95	6	.	.	PUNCT
fcis-17186	96	1	deep	deep	ADJ
fcis-17186	96	2	neural	neural	ADJ
fcis-17186	96	3	network	network	NOUN
fcis-17186	96	4	in	in	ADP
fcis-17186	96	5	qsar	qsar	NOUN
fcis-17186	96	6	studies	study	NOUN
fcis-17186	96	7	using	use	VERB
fcis-17186	96	8	deep	deep	ADJ
fcis-17186	96	9	belief	belief	NOUN
fcis-17186	96	10	network	network	NOUN
fcis-17186	96	11	,	,	PUNCT
fcis-17186	96	12	applied	apply	VERB
fcis-17186	96	13	softcomputing	softcomputing	NOUN
fcis-17186	96	14	,	,	PUNCT
fcis-17186	96	15	62	62	NUM
fcis-17186	96	16	(	(	PUNCT
fcis-17186	96	17	2018	2018	NUM
fcis-17186	96	18	):	):	PUNCT
fcis-17186	96	19	251	251	NUM
fcis-17186	96	20	-	-	SYM
fcis-17186	96	21	258	258	NUM
fcis-17186	96	22	.	.	PUNCT
fcis-17186	97	1	[	[	X
fcis-17186	97	2	4	4	X
fcis-17186	97	3	]	]	X
fcis-17186	97	4	y.	y.	PROPN
fcis-17186	97	5	s.	s.	PROPN
fcis-17186	97	6	xu	xu	PROPN
fcis-17186	97	7	.	.	PUNCT
fcis-17186	98	1	convergence	convergence	NOUN
fcis-17186	98	2	of	of	ADP
fcis-17186	98	3	deep	deep	ADJ
fcis-17186	98	4	convolutional	convolutional	ADJ
fcis-17186	98	5	neural	neural	ADJ
fcis-17186	98	6	networks	network	NOUN
fcis-17186	98	7	,	,	PUNCT
fcis-17186	98	8	neural	neural	ADJ
fcis-17186	98	9	networks	network	NOUN
fcis-17186	98	10	,	,	PUNCT
fcis-17186	98	11	153	153	NUM
fcis-17186	98	12	(	(	PUNCT
fcis-17186	98	13	2022	2022	NUM
fcis-17186	98	14	):	):	PUNCT
fcis-17186	98	15	553	553	NUM
fcis-17186	98	16	-	-	SYM
fcis-17186	98	17	563	563	NUM
fcis-17186	98	18	.	.	PUNCT
fcis-17186	99	1	[	[	X
fcis-17186	99	2	5	5	X
fcis-17186	99	3	]	]	PUNCT
fcis-17186	99	4	s.	s.	PROPN
fcis-17186	99	5	f.	f.	PROPN
fcis-17186	99	6	wen	wen	PROPN
fcis-17186	99	7	,	,	PUNCT
fcis-17186	99	8	translation	translation	NOUN
fcis-17186	99	9	analysis	analysis	NOUN
fcis-17186	99	10	of	of	ADP
fcis-17186	99	11	english	english	ADJ
fcis-17186	99	12	address	address	NOUN
fcis-17186	99	13	image	image	NOUN
fcis-17186	99	14	recognition	recognition	NOUN
fcis-17186	99	15	based	base	VERB
fcis-17186	99	16	on	on	ADP
fcis-17186	99	17	image	image	NOUN
fcis-17186	99	18	recognition	recognition	NOUN
fcis-17186	99	19	,	,	PUNCT
fcis-17186	99	20	image	image	NOUN
fcis-17186	99	21	and	and	CCONJ
fcis-17186	99	22	video	video	NOUN
fcis-17186	99	23	processing	processing	NOUN
fcis-17186	99	24	,	,	PUNCT
fcis-17186	99	25	11	11	NUM
fcis-17186	99	26	(	(	PUNCT
fcis-17186	99	27	2019	2019	NUM
fcis-17186	99	28	):	):	PUNCT
fcis-17186	99	29	1	1	NUM
fcis-17186	99	30	-	-	SYM
fcis-17186	99	31	9	9	NUM
fcis-17186	99	32	.	.	PUNCT
fcis-17186	100	1	[	[	X
fcis-17186	100	2	6	6	NUM
fcis-17186	100	3	]	]	SYM
fcis-17186	100	4	m.y.park	m.y.park	NOUN
fcis-17186	100	5	,	,	PUNCT
fcis-17186	100	6	t.hastie	t.hastie	NOUN
fcis-17186	100	7	,	,	PUNCT
fcis-17186	100	8	l1	l1	PROPN
fcis-17186	100	9	-	-	PUNCT
fcis-17186	100	10	regularization	regularization	NOUN
fcis-17186	100	11	path	path	NOUN
fcis-17186	100	12	algorithm	algorithm	NOUN
fcis-17186	100	13	for	for	ADP
fcis-17186	100	14	generalized	generalized	ADJ
fcis-17186	100	15	linear	linear	PROPN
fcis-17186	100	16	model	model	NOUN
fcis-17186	100	17	,	,	PUNCT
fcis-17186	100	18	journal	journal	NOUN
fcis-17186	100	19	of	of	ADP
fcis-17186	100	20	the	the	DET
fcis-17186	100	21	royal	royal	ADJ
fcis-17186	100	22	statistical	statistical	ADJ
fcis-17186	100	23	society	society	NOUN
fcis-17186	100	24	series	series	PROPN
fcis-17186	100	25	b	b	PROPN
fcis-17186	100	26	:	:	PUNCT
fcis-17186	100	27	statistical	statistical	ADJ
fcis-17186	100	28	methodology	methodology	NOUN
fcis-17186	100	29	,	,	PUNCT
fcis-17186	100	30	69	69	NUM
fcis-17186	100	31	(	(	PUNCT
fcis-17186	100	32	2007	2007	NUM
fcis-17186	100	33	):	):	PUNCT
fcis-17186	100	34	659	659	NUM
fcis-17186	100	35	–	–	PUNCT
fcis-17186	100	36	677	677	NUM
fcis-17186	100	37	.	.	PUNCT
fcis-17186	101	1	[	[	X
fcis-17186	101	2	7	7	X
fcis-17186	101	3	]	]	PUNCT
fcis-17186	101	4	c.yang	c.yang	NOUN
fcis-17186	101	5	.	.	PUNCT
fcis-17186	101	6	structured	structure	VERB
fcis-17186	101	7	pruning	pruning	NOUN
fcis-17186	101	8	of	of	ADP
fcis-17186	101	9	convolutional	convolutional	ADJ
fcis-17186	101	10	neural	neural	ADJ
fcis-17186	101	11	networks	network	NOUN
fcis-17186	101	12	via	via	ADP
fcis-17186	101	13	l1	l1	PROPN
fcis-17186	101	14	regularization	regularization	NOUN
fcis-17186	101	15	,	,	PUNCT
fcis-17186	101	16	ieee	ieee	NOUN
fcis-17186	101	17	access	access	NOUN
fcis-17186	101	18	,	,	PUNCT
fcis-17186	101	19	7	7	NUM
fcis-17186	101	20	(	(	PUNCT
fcis-17186	101	21	2019	2019	NUM
fcis-17186	101	22	):	):	PUNCT
fcis-17186	101	23	106385	106385	NUM
fcis-17186	101	24	-	-	SYM
fcis-17186	101	25	106394	106394	NUM
fcis-17186	101	26	.	.	PUNCT
fcis-17186	102	1	[	[	X
fcis-17186	102	2	8	8	NUM
fcis-17186	102	3	]	]	X
fcis-17186	102	4	l.	l.	PROPN
fcis-17186	102	5	li	li	PROPN
fcis-17186	102	6	.	.	PUNCT
fcis-17186	102	7	approximating	approximate	VERB
fcis-17186	102	8	the	the	DET
fcis-17186	102	9	gradient	gradient	NOUN
fcis-17186	102	10	of	of	ADP
fcis-17186	102	11	cross	cross	ADJ
fcis-17186	102	12	-	-	ADJ
fcis-17186	102	13	entropy	entropy	ADJ
fcis-17186	102	14	loss	loss	NOUN
fcis-17186	102	15	function	function	NOUN
fcis-17186	102	16	,	,	PUNCT
fcis-17186	102	17	ieee	ieee	NOUN
fcis-17186	102	18	access	access	NOUN
fcis-17186	102	19	,	,	PUNCT
fcis-17186	102	20	2020	2020	NUM
fcis-17186	102	21	(	(	PUNCT
fcis-17186	102	22	8)	8)	NUM
fcis-17186	102	23	:	:	PUNCT
fcis-17186	102	24	111626	111626	NUM
fcis-17186	102	25	-	-	SYM
fcis-17186	102	26	111635	111635	NUM
fcis-17186	102	27	.	.	PUNCT
fcis-17186	103	1	[	[	X
fcis-17186	103	2	9	9	NUM
fcis-17186	103	3	]	]	PUNCT
fcis-17186	103	4	t.wiatowski.a	t.wiatowski.a	PROPN
fcis-17186	103	5	mathematical	mathematical	ADJ
fcis-17186	103	6	theory	theory	NOUN
fcis-17186	103	7	of	of	ADP
fcis-17186	103	8	deep	deep	ADJ
fcis-17186	103	9	convolutional	convolutional	ADJ
fcis-17186	103	10	neural	neural	ADJ
fcis-17186	103	11	networks	network	NOUN
fcis-17186	103	12	for	for	ADP
fcis-17186	103	13	feature	feature	NOUN
fcis-17186	103	14	extraction	extraction	NOUN
fcis-17186	103	15	,	,	PUNCT
fcis-17186	103	16	ieee	ieee	NOUN
fcis-17186	103	17	transactions	transaction	NOUN
fcis-17186	103	18	on	on	ADP
fcis-17186	103	19	information	information	NOUN
fcis-17186	103	20	theory	theory	NOUN
fcis-17186	103	21	,	,	PUNCT
fcis-17186	103	22	64(2017	64(2017	NUM
fcis-17186	103	23	):	):	PUNCT
fcis-17186	103	24	1845	1845	NUM
fcis-17186	103	25	-	-	SYM
fcis-17186	103	26	1866	1866	NUM
fcis-17186	103	27	.	.	PUNCT
fcis-17186	104	1	[	[	X
fcis-17186	104	2	10	10	NUM
fcis-17186	104	3	]	]	X
fcis-17186	104	4	p.y.wang	p.y.wang	PROPN
fcis-17186	104	5	.	.	PUNCT
fcis-17186	105	1	differentially	differentially	ADV
fcis-17186	105	2	private	private	ADJ
fcis-17186	105	3	sgd	sgd	NOUN
fcis-17186	105	4	with	with	ADP
fcis-17186	105	5	non	non	ADJ
fcis-17186	105	6	-	-	ADJ
fcis-17186	105	7	smooth	smooth	ADJ
fcis-17186	105	8	losses	loss	NOUN
fcis-17186	105	9	,	,	PUNCT
fcis-17186	105	10	applied	apply	VERB
fcis-17186	105	11	and	and	CCONJ
fcis-17186	105	12	computational	computational	ADJ
fcis-17186	105	13	harmonic	harmonic	ADJ
fcis-17186	105	14	analysis	analysis	NOUN
fcis-17186	105	15	,	,	PUNCT
fcis-17186	105	16	56	56	NUM
fcis-17186	105	17	(	(	PUNCT
fcis-17186	105	18	2022	2022	NUM
fcis-17186	105	19	):	):	PUNCT
fcis-17186	105	20	306	306	NUM
fcis-17186	105	21	-	-	SYM
fcis-17186	105	22	336	336	NUM
fcis-17186	105	23	.	.	PUNCT