Collaborative Caching in Edge Computing via Federated Learning and Deep Reinforcement Learning

Table 2

Key notations and descriptions.

Notation	Description
	Set of BSs
	Set of UE
	Set of content
	Set of UE requesting content from in period
	Request vector for each UE in slot
	Local content popularity for all content in in time slot
	Content delivery variables via local, collaborative and cloud
	Caching decisions for content
	Size of content
	Energy consumption of each bit of data cached by MEC server
	The transmission power from to , the channel gain between and
	The variance of additive Gaussian white noise
,	Transmission rate from to , transfer rate from CM to cloud server
	Costs delivered via on-premises, collaboration and cloud
	Total cost, cache cost, transmission cost
	The feature vector of the content
	The category label of the content
	The probability that user in time slot requests content , the predicted probability that user in time slot requests content
	The cumulative number of requested samples for user
	The user preference model parameter vector of user , the user preference model parameter vector learned by user at the -th iteration. The parameter vector of the regional integrated model
	The logistic loss of user
	The gradient vector of the -th sample with respect to ’s logistic loss, the sum of the gradient vectors of the logistic loss of the first r samples
;	Positive regularization parameter; tuning parameters
	The nonincreasing learning rate, the learning rate of the -th feature, and the parameter related to the learning rate
	Feature dimension