id	sid	tid	token	lemma	pos
xg94hm54j30	1	1	optimal	optimal	ADJ
xg94hm54j30	1	2	decision	decision	NOUN
xg94hm54j30	1	3	-	-	PUNCT
xg94hm54j30	1	4	making	making	NOUN
xg94hm54j30	1	5	or	or	CCONJ
xg94hm54j30	1	6	policy	policy	NOUN
xg94hm54j30	1	7	satisfies	satisfie	NOUN
xg94hm54j30	1	8	both	both	CCONJ
xg94hm54j30	1	9	short	short	ADJ
xg94hm54j30	1	10	-	-	PUNCT
xg94hm54j30	1	11	term	term	NOUN
xg94hm54j30	1	12	and	and	CCONJ
xg94hm54j30	1	13	long	long	ADJ
xg94hm54j30	1	14	-	-	PUNCT
xg94hm54j30	1	15	term	term	NOUN
xg94hm54j30	1	16	objectives	objective	NOUN
xg94hm54j30	1	17	.	.	PUNCT
xg94hm54j30	2	1	policy	policy	NOUN
xg94hm54j30	2	2	optimization	optimization	NOUN
xg94hm54j30	2	3	in	in	ADP
xg94hm54j30	2	4	complex	complex	ADJ
xg94hm54j30	2	5	dynamic	dynamic	ADJ
xg94hm54j30	2	6	environments	environment	NOUN
xg94hm54j30	2	7	is	be	AUX
xg94hm54j30	2	8	notoriously	notoriously	ADV
xg94hm54j30	2	9	difficult	difficult	ADJ
xg94hm54j30	2	10	due	due	ADP
xg94hm54j30	2	11	to	to	ADP
xg94hm54j30	2	12	the	the	DET
xg94hm54j30	2	13	absence	absence	NOUN
xg94hm54j30	2	14	of	of	ADP
xg94hm54j30	2	15	succinct	succinct	ADJ
xg94hm54j30	2	16	mathematical	mathematical	ADJ
xg94hm54j30	2	17	models	model	NOUN
xg94hm54j30	2	18	.	.	PUNCT
xg94hm54j30	3	1	reinforcement	reinforcement	NOUN
xg94hm54j30	3	2	learning	learning	PROPN
xg94hm54j30	3	3	(	(	PUNCT
xg94hm54j30	3	4	rl	rl	PROPN
xg94hm54j30	3	5	)	)	PUNCT
xg94hm54j30	3	6	leverages	leverage	VERB
xg94hm54j30	3	7	data	datum	NOUN
xg94hm54j30	3	8	from	from	ADP
xg94hm54j30	3	9	experiments	experiment	NOUN
xg94hm54j30	3	10	or	or	CCONJ
xg94hm54j30	3	11	simulations	simulation	NOUN
xg94hm54j30	3	12	to	to	PART
xg94hm54j30	3	13	learn	learn	VERB
xg94hm54j30	3	14	optimal	optimal	ADJ
xg94hm54j30	3	15	policies	policy	NOUN
xg94hm54j30	3	16	that	that	PRON
xg94hm54j30	3	17	,	,	PUNCT
xg94hm54j30	3	18	in	in	ADP
xg94hm54j30	3	19	many	many	ADJ
xg94hm54j30	3	20	domains	domain	NOUN
xg94hm54j30	3	21	,	,	PUNCT
xg94hm54j30	3	22	have	have	AUX
xg94hm54j30	3	23	already	already	ADV
xg94hm54j30	3	24	outperformed	outperform	VERB
xg94hm54j30	3	25	policies	policy	NOUN
xg94hm54j30	3	26	designed	design	VERB
xg94hm54j30	3	27	by	by	ADP
xg94hm54j30	3	28	humans	human	NOUN
xg94hm54j30	3	29	.	.	PUNCT
xg94hm54j30	4	1	rl	rl	PROPN
xg94hm54j30	4	2	still	still	ADV
xg94hm54j30	4	3	faces	face	VERB
xg94hm54j30	4	4	numerous	numerous	ADJ
xg94hm54j30	4	5	challenges	challenge	NOUN
xg94hm54j30	4	6	in	in	ADP
xg94hm54j30	4	7	the	the	DET
xg94hm54j30	4	8	multi	multi	ADJ
xg94hm54j30	4	9	-	-	ADJ
xg94hm54j30	4	10	agent	agent	ADJ
xg94hm54j30	4	11	setting	setting	NOUN
xg94hm54j30	4	12	,	,	PUNCT
xg94hm54j30	4	13	where	where	SCONJ
xg94hm54j30	4	14	the	the	DET
xg94hm54j30	4	15	participating	participate	VERB
xg94hm54j30	4	16	agents	agent	NOUN
xg94hm54j30	4	17	interact	interact	VERB
xg94hm54j30	4	18	in	in	ADP
xg94hm54j30	4	19	shared	shared	ADJ
xg94hm54j30	4	20	environments	environment	NOUN
xg94hm54j30	4	21	.	.	PUNCT
xg94hm54j30	5	1	this	this	DET
xg94hm54j30	5	2	dissertation	dissertation	NOUN
xg94hm54j30	5	3	aims	aim	VERB
xg94hm54j30	5	4	to	to	PART
xg94hm54j30	5	5	address	address	VERB
xg94hm54j30	5	6	two	two	NUM
xg94hm54j30	5	7	challenges	challenge	NOUN
xg94hm54j30	5	8	in	in	ADP
xg94hm54j30	5	9	decentralized	decentralized	ADJ
xg94hm54j30	5	10	cooperative	cooperative	ADJ
xg94hm54j30	5	11	multi	multi	ADJ
xg94hm54j30	5	12	-	-	ADJ
xg94hm54j30	5	13	agent	agent	ADJ
xg94hm54j30	5	14	reinforcement	reinforcement	NOUN
xg94hm54j30	5	15	learning	learning	NOUN
xg94hm54j30	5	16	,	,	PUNCT
xg94hm54j30	5	17	a	a	DET
xg94hm54j30	5	18	new	new	ADJ
xg94hm54j30	5	19	training	training	NOUN
xg94hm54j30	5	20	paradigm	paradigm	NOUN
xg94hm54j30	5	21	that	that	PRON
xg94hm54j30	5	22	features	feature	VERB
xg94hm54j30	5	23	scalability	scalability	NOUN
xg94hm54j30	5	24	and	and	CCONJ
xg94hm54j30	5	25	privacy	privacy	NOUN
xg94hm54j30	5	26	guarantees	guarantee	NOUN
xg94hm54j30	5	27	for	for	ADP
xg94hm54j30	5	28	cooperative	cooperative	ADJ
xg94hm54j30	5	29	agents.in	agents.in	X
xg94hm54j30	5	30	the	the	DET
xg94hm54j30	5	31	first	first	ADJ
xg94hm54j30	5	32	part	part	NOUN
xg94hm54j30	5	33	,	,	PUNCT
xg94hm54j30	5	34	we	we	PRON
xg94hm54j30	5	35	study	study	VERB
xg94hm54j30	5	36	the	the	DET
xg94hm54j30	5	37	behavior	behavior	NOUN
xg94hm54j30	5	38	of	of	ADP
xg94hm54j30	5	39	the	the	DET
xg94hm54j30	5	40	cooperative	cooperative	ADJ
xg94hm54j30	5	41	agents	agent	NOUN
xg94hm54j30	5	42	in	in	ADP
xg94hm54j30	5	43	a	a	DET
xg94hm54j30	5	44	network	network	NOUN
xg94hm54j30	5	45	that	that	PRON
xg94hm54j30	5	46	includes	include	VERB
xg94hm54j30	5	47	adversarial	adversarial	ADJ
xg94hm54j30	5	48	agents	agent	NOUN
xg94hm54j30	5	49	.	.	PUNCT
xg94hm54j30	6	1	adversarial	adversarial	ADJ
xg94hm54j30	6	2	attacks	attack	NOUN
xg94hm54j30	6	3	in	in	ADP
xg94hm54j30	6	4	training	training	NOUN
xg94hm54j30	6	5	can	can	AUX
xg94hm54j30	6	6	strongly	strongly	ADV
xg94hm54j30	6	7	influence	influence	VERB
xg94hm54j30	6	8	the	the	DET
xg94hm54j30	6	9	performance	performance	NOUN
xg94hm54j30	6	10	of	of	ADP
xg94hm54j30	6	11	multi	multi	ADJ
xg94hm54j30	6	12	-	-	NOUN
xg94hm54j30	6	13	agent	agent	ADJ
xg94hm54j30	6	14	rl	rl	NOUN
xg94hm54j30	6	15	algorithms	algorithm	NOUN
xg94hm54j30	6	16	.	.	PUNCT
xg94hm54j30	7	1	it	it	PRON
xg94hm54j30	7	2	is	be	AUX
xg94hm54j30	7	3	,	,	PUNCT
xg94hm54j30	7	4	thus	thus	ADV
xg94hm54j30	7	5	,	,	PUNCT
xg94hm54j30	7	6	highly	highly	ADV
xg94hm54j30	7	7	desirable	desirable	ADJ
xg94hm54j30	7	8	to	to	PART
xg94hm54j30	7	9	augment	augment	VERB
xg94hm54j30	7	10	existing	exist	VERB
xg94hm54j30	7	11	algorithms	algorithm	NOUN
xg94hm54j30	7	12	such	such	ADJ
xg94hm54j30	7	13	that	that	SCONJ
xg94hm54j30	7	14	the	the	DET
xg94hm54j30	7	15	impact	impact	NOUN
xg94hm54j30	7	16	of	of	ADP
xg94hm54j30	7	17	adversarial	adversarial	ADJ
xg94hm54j30	7	18	attacks	attack	NOUN
xg94hm54j30	7	19	on	on	ADP
xg94hm54j30	7	20	cooperative	cooperative	ADJ
xg94hm54j30	7	21	networks	network	NOUN
xg94hm54j30	7	22	is	be	AUX
xg94hm54j30	7	23	eliminated	eliminate	VERB
xg94hm54j30	7	24	,	,	PUNCT
xg94hm54j30	7	25	or	or	CCONJ
xg94hm54j30	7	26	at	at	ADP
xg94hm54j30	7	27	least	least	ADJ
xg94hm54j30	7	28	bounded	bound	VERB
xg94hm54j30	7	29	.	.	PUNCT
xg94hm54j30	8	1	we	we	PRON
xg94hm54j30	8	2	introduce	introduce	VERB
xg94hm54j30	8	3	a	a	DET
xg94hm54j30	8	4	resilient	resilient	ADJ
xg94hm54j30	8	5	projection	projection	NOUN
xg94hm54j30	8	6	-	-	PUNCT
xg94hm54j30	8	7	based	base	VERB
xg94hm54j30	8	8	consensus	consensus	NOUN
xg94hm54j30	8	9	multi	multi	NOUN
xg94hm54j30	8	10	-	-	ADJ
xg94hm54j30	8	11	agent	agent	ADJ
xg94hm54j30	8	12	actor	actor	NOUN
xg94hm54j30	8	13	-	-	PUNCT
xg94hm54j30	8	14	critic	critic	NOUN
xg94hm54j30	8	15	algorithm	algorithm	NOUN
xg94hm54j30	8	16	,	,	PUNCT
xg94hm54j30	8	17	whereby	whereby	SCONJ
xg94hm54j30	8	18	each	each	DET
xg94hm54j30	8	19	agent	agent	NOUN
xg94hm54j30	8	20	receives	receive	VERB
xg94hm54j30	8	21	a	a	DET
xg94hm54j30	8	22	private	private	ADJ
xg94hm54j30	8	23	reward	reward	NOUN
xg94hm54j30	8	24	and	and	CCONJ
xg94hm54j30	8	25	communicates	communicate	NOUN
xg94hm54j30	8	26	with	with	ADP
xg94hm54j30	8	27	its	its	PRON
xg94hm54j30	8	28	neighbors	neighbor	NOUN
xg94hm54j30	8	29	to	to	PART
xg94hm54j30	8	30	estimate	estimate	VERB
xg94hm54j30	8	31	the	the	DET
xg94hm54j30	8	32	team	team	NOUN
xg94hm54j30	8	33	-	-	PUNCT
xg94hm54j30	8	34	average	average	ADJ
xg94hm54j30	8	35	reward	reward	NOUN
xg94hm54j30	8	36	and	and	CCONJ
xg94hm54j30	8	37	value	value	NOUN
xg94hm54j30	8	38	function	function	NOUN
xg94hm54j30	8	39	.	.	PUNCT
xg94hm54j30	9	1	we	we	PRON
xg94hm54j30	9	2	show	show	VERB
xg94hm54j30	9	3	that	that	SCONJ
xg94hm54j30	9	4	in	in	ADP
xg94hm54j30	9	5	the	the	DET
xg94hm54j30	9	6	presence	presence	NOUN
xg94hm54j30	9	7	of	of	ADP
xg94hm54j30	9	8	byzantine	byzantine	ADJ
xg94hm54j30	9	9	agents	agent	NOUN
xg94hm54j30	9	10	,	,	PUNCT
xg94hm54j30	9	11	whose	whose	DET
xg94hm54j30	9	12	estimation	estimation	NOUN
xg94hm54j30	9	13	and	and	CCONJ
xg94hm54j30	9	14	communication	communication	NOUN
xg94hm54j30	9	15	strategies	strategy	NOUN
xg94hm54j30	9	16	are	be	AUX
xg94hm54j30	9	17	arbitrary	arbitrary	ADJ
xg94hm54j30	9	18	,	,	PUNCT
xg94hm54j30	9	19	the	the	DET
xg94hm54j30	9	20	estimates	estimate	NOUN
xg94hm54j30	9	21	of	of	ADP
xg94hm54j30	9	22	the	the	DET
xg94hm54j30	9	23	cooperative	cooperative	ADJ
xg94hm54j30	9	24	agents	agent	NOUN
xg94hm54j30	9	25	converge	converge	VERB
xg94hm54j30	9	26	to	to	ADP
xg94hm54j30	9	27	a	a	DET
xg94hm54j30	9	28	bounded	bound	VERB
xg94hm54j30	9	29	consensus	consensus	NOUN
xg94hm54j30	9	30	value	value	NOUN
xg94hm54j30	9	31	,	,	PUNCT
xg94hm54j30	9	32	provided	provide	VERB
xg94hm54j30	9	33	that	that	SCONJ
xg94hm54j30	9	34	there	there	PRON
xg94hm54j30	9	35	are	be	VERB
xg94hm54j30	9	36	at	at	ADV
xg94hm54j30	9	37	most	most	ADJ
xg94hm54j30	9	38	h	h	NOUN
xg94hm54j30	9	39	byzantine	byzantine	ADJ
xg94hm54j30	9	40	agents	agent	NOUN
xg94hm54j30	9	41	in	in	ADP
xg94hm54j30	9	42	the	the	DET
xg94hm54j30	9	43	network	network	NOUN
xg94hm54j30	9	44	that	that	PRON
xg94hm54j30	9	45	is	be	AUX
xg94hm54j30	9	46	(	(	PUNCT
xg94hm54j30	9	47	2h+1)-robust	2h+1)-robust	PROPN
xg94hm54j30	9	48	.	.	PUNCT
xg94hm54j30	10	1	furthermore	furthermore	ADV
xg94hm54j30	10	2	,	,	PUNCT
xg94hm54j30	10	3	we	we	PRON
xg94hm54j30	10	4	prove	prove	VERB
xg94hm54j30	10	5	that	that	SCONJ
xg94hm54j30	10	6	the	the	DET
xg94hm54j30	10	7	joint	joint	ADJ
xg94hm54j30	10	8	cooperative	cooperative	ADJ
xg94hm54j30	10	9	policy	policy	NOUN
xg94hm54j30	10	10	converges	converge	NOUN
xg94hm54j30	10	11	to	to	ADP
xg94hm54j30	10	12	a	a	DET
xg94hm54j30	10	13	bounded	bound	VERB
xg94hm54j30	10	14	neighborhood	neighborhood	NOUN
xg94hm54j30	10	15	around	around	ADP
xg94hm54j30	10	16	a	a	DET
xg94hm54j30	10	17	locally	locally	ADV
xg94hm54j30	10	18	optimal	optimal	ADJ
xg94hm54j30	10	19	cooperative	cooperative	NOUN
xg94hm54j30	10	20	policy.in	policy.in	X
xg94hm54j30	10	21	the	the	DET
xg94hm54j30	10	22	second	second	ADJ
xg94hm54j30	10	23	part	part	NOUN
xg94hm54j30	10	24	,	,	PUNCT
xg94hm54j30	10	25	we	we	PRON
xg94hm54j30	10	26	consider	consider	VERB
xg94hm54j30	10	27	a	a	DET
xg94hm54j30	10	28	fully	fully	ADV
xg94hm54j30	10	29	cooperative	cooperative	ADJ
xg94hm54j30	10	30	network	network	NOUN
xg94hm54j30	10	31	subject	subject	ADJ
xg94hm54j30	10	32	to	to	ADP
xg94hm54j30	10	33	communication	communication	NOUN
xg94hm54j30	10	34	delays	delay	NOUN
xg94hm54j30	10	35	and	and	CCONJ
xg94hm54j30	10	36	packet	packet	NOUN
xg94hm54j30	10	37	dropouts	dropout	NOUN
xg94hm54j30	10	38	.	.	PUNCT
xg94hm54j30	11	1	the	the	DET
xg94hm54j30	11	2	assumption	assumption	NOUN
xg94hm54j30	11	3	about	about	ADP
xg94hm54j30	11	4	disrupted	disrupt	VERB
xg94hm54j30	11	5	communication	communication	NOUN
xg94hm54j30	11	6	is	be	AUX
xg94hm54j30	11	7	reasonable	reasonable	ADJ
xg94hm54j30	11	8	in	in	ADP
xg94hm54j30	11	9	online	online	ADJ
xg94hm54j30	11	10	decentralized	decentralized	ADJ
xg94hm54j30	11	11	training	training	NOUN
xg94hm54j30	11	12	where	where	SCONJ
xg94hm54j30	11	13	agents	agent	NOUN
xg94hm54j30	11	14	continuously	continuously	ADV
xg94hm54j30	11	15	accumulate	accumulate	VERB
xg94hm54j30	11	16	new	new	ADJ
xg94hm54j30	11	17	experiences	experience	NOUN
xg94hm54j30	11	18	from	from	ADP
xg94hm54j30	11	19	the	the	DET
xg94hm54j30	11	20	environment	environment	NOUN
xg94hm54j30	11	21	and	and	CCONJ
xg94hm54j30	11	22	communicate	communicate	VERB
xg94hm54j30	11	23	periodically	periodically	ADV
xg94hm54j30	11	24	.	.	PUNCT
xg94hm54j30	12	1	we	we	PRON
xg94hm54j30	12	2	present	present	VERB
xg94hm54j30	12	3	a	a	DET
xg94hm54j30	12	4	multi	multi	ADJ
xg94hm54j30	12	5	-	-	ADJ
xg94hm54j30	12	6	agent	agent	ADJ
xg94hm54j30	12	7	actor	actor	NOUN
xg94hm54j30	12	8	-	-	PUNCT
xg94hm54j30	12	9	critic	critic	NOUN
xg94hm54j30	12	10	algorithm	algorithm	NOUN
xg94hm54j30	12	11	with	with	ADP
xg94hm54j30	12	12	td	td	PROPN
xg94hm54j30	12	13	error	error	NOUN
xg94hm54j30	12	14	aggregation	aggregation	NOUN
xg94hm54j30	12	15	,	,	PUNCT
xg94hm54j30	12	16	where	where	SCONJ
xg94hm54j30	12	17	the	the	DET
xg94hm54j30	12	18	aggregation	aggregation	NOUN
xg94hm54j30	12	19	of	of	ADP
xg94hm54j30	12	20	td	td	NOUN
xg94hm54j30	12	21	errors	error	NOUN
xg94hm54j30	12	22	ensures	ensure	VERB
xg94hm54j30	12	23	cooperation	cooperation	NOUN
xg94hm54j30	12	24	between	between	ADP
xg94hm54j30	12	25	the	the	DET
xg94hm54j30	12	26	agents	agent	NOUN
xg94hm54j30	12	27	.	.	PUNCT
xg94hm54j30	13	1	the	the	DET
xg94hm54j30	13	2	assumptions	assumption	NOUN
xg94hm54j30	13	3	about	about	ADP
xg94hm54j30	13	4	the	the	DET
xg94hm54j30	13	5	communication	communication	NOUN
xg94hm54j30	13	6	lead	lead	NOUN
xg94hm54j30	13	7	to	to	ADP
xg94hm54j30	13	8	an	an	DET
xg94hm54j30	13	9	increased	increase	VERB
xg94hm54j30	13	10	communication	communication	NOUN
xg94hm54j30	13	11	burden	burden	NOUN
xg94hm54j30	13	12	for	for	ADP
xg94hm54j30	13	13	every	every	DET
xg94hm54j30	13	14	agent	agent	NOUN
xg94hm54j30	13	15	as	as	SCONJ
xg94hm54j30	13	16	measured	measure	VERB
xg94hm54j30	13	17	by	by	ADP
xg94hm54j30	13	18	the	the	DET
xg94hm54j30	13	19	dimension	dimension	NOUN
xg94hm54j30	13	20	of	of	ADP
xg94hm54j30	13	21	the	the	DET
xg94hm54j30	13	22	transmitted	transmit	VERB
xg94hm54j30	13	23	data	datum	NOUN
xg94hm54j30	13	24	;	;	PUNCT
xg94hm54j30	13	25	nonetheless	nonetheless	ADV
xg94hm54j30	13	26	,	,	PUNCT
xg94hm54j30	13	27	the	the	DET
xg94hm54j30	13	28	communication	communication	NOUN
xg94hm54j30	13	29	burden	burden	NOUN
xg94hm54j30	13	30	is	be	AUX
xg94hm54j30	13	31	only	only	ADV
xg94hm54j30	13	32	quadratic	quadratic	ADJ
xg94hm54j30	13	33	in	in	ADP
xg94hm54j30	13	34	the	the	DET
xg94hm54j30	13	35	graph	graph	NOUN
xg94hm54j30	13	36	size	size	NOUN
xg94hm54j30	13	37	,	,	PUNCT
xg94hm54j30	13	38	and	and	CCONJ
xg94hm54j30	13	39	thus	thus	ADV
xg94hm54j30	13	40	the	the	DET
xg94hm54j30	13	41	algorithm	algorithm	PROPN
xg94hm54j30	13	42	is	be	AUX
xg94hm54j30	13	43	applicable	applicable	ADJ
xg94hm54j30	13	44	in	in	ADP
xg94hm54j30	13	45	large	large	ADJ
xg94hm54j30	13	46	networks	network	NOUN
xg94hm54j30	13	47	.	.	PUNCT
xg94hm54j30	14	1	we	we	PRON
xg94hm54j30	14	2	prove	prove	VERB
xg94hm54j30	14	3	analytically	analytically	ADV
xg94hm54j30	14	4	that	that	SCONJ
xg94hm54j30	14	5	the	the	DET
xg94hm54j30	14	6	agents	agent	NOUN
xg94hm54j30	14	7	approximately	approximately	ADV
xg94hm54j30	14	8	maximize	maximize	VERB
xg94hm54j30	14	9	the	the	DET
xg94hm54j30	14	10	team	team	NOUN
xg94hm54j30	14	11	-	-	PUNCT
xg94hm54j30	14	12	average	average	ADJ
xg94hm54j30	14	13	objective	objective	ADJ
xg94hm54j30	14	14	function	function	NOUN
xg94hm54j30	14	15	.	.	PUNCT