Research Article

Deep Interest-Shifting Network with Meta-Embeddings for Fresh Item Recommendation

Table 1

Notations.

NotationDefinition or descriptions

A pair of user and item
The id embedding of item
The embedded features (except id) of item
The embedded features (include id) of user
, Common contexts and scenario-specific contexts
, Item/user representation networks
Decision-making network
Context network
, Latent user/item vectors in the interest space
, Output vector/matrix of context network
, Cold-start and warmed item id embeddings in the meta-training procedure
, Support and query sets of a cold-start item in
, Predicted labels on and
, , Cold-start, warmed, and meta-loss in the meta-training procedure
, , Auxiliary, warm-start, and cold-start datasets
, , Ground-truth label and predicted labels of context-aware models and DisNet
Interest-shifting operator
The id embeddings of top relevant items
Attention scores
, Weight matrices and bias vectors of the NN operator
, Weight matrices and bias vector of RM-IdEG
, , Parameters of attentional embedding aggregator
, Item sets of and