Skip to main content

Advances, Systems and Applications

Table 1 List of mathematical notations

From: A split-federated learning and edge-cloud based efficient and privacy-preserving large-scale item recommendation model

Notations

Description

\({u}_{i}\), \({v}_{j}\)

User \(i\) and item \(j\)

\({M}_{u}\left(u,{\uptheta }_{u}\right)\), \({M}_{v}\left(v,{\uptheta }_{v}\right)\)

User and item model

\({\widetilde{u}}_{i}\), \({\widetilde{v}}_{j}\)

Output of user and item model

\(t\)

Training round

\({\theta }_{u}^{0}, {\theta }_{v}^{0}\)

Initial user and item model’s parameters

\({\uptheta }_{u}^{t}, {\uptheta }_{v}^{t}\)

User and item model’s parameters at training round \(t\)

\({C}^{t}\)

Random selected subset of clients at training round \(t\)

\(K\)

Number of selected clients in each training round

\({B}_{k}^{t}\)

Batch of training set for client \(k\) at training round \(t\)

\({V}_{k}^{-}\), \({V}_{k}^{+}\)

Subset of non-clicked and clicked items for client \(k\)

\(\nabla {\theta }_{u}^{t}, \nabla {\theta }_{v}^{t}\)

The gradients of User and item model at training round \(t\)

\(\mathrm{\rm P}\)

Negative sampling rate

\({\upeta }_{u}\), \({\upeta }_{v}\)

Learning rate for user and item model update

\({\upbeta }_{1}\), \({\upbeta }_{2}\),\(\uptau\)

Hyperparameters for FedAdam optimizer

\(\upsigma\), \(\updelta\)

Activation functions

\(r\)

Reduction ratio