Matrix factorization model for item prediction optimized using BPR-Opt;
Declaring type: BPRMF
Namespace: MyMediaLite.ItemRecommender
Assembly: MyMediaLite
Collapse/Expand Protected Fields
  Name Description
Protected Field fast_sampling Fast, but memory-intensive sampling
Protected Field fast_sampling_memory_limit Fast sampling memory limit, in MiB
Protected Field item_bias Use the first item latent factor as a bias term if set to true
Protected Field iteration_length One iteration is iteration_length * number of entries in the training matrix
Protected Field learn_rate Learning rate alpha
Protected Field random Random number generator
Protected Field reg_i Regularization parameter for positive item factors
Protected Field reg_j Regularization parameter for negative item factors
Protected Field reg_u Regularization parameter for user factors
Protected Field user_neg_items support data structure for fast sampling
Protected Field user_pos_items support data structure for fast sampling
Collapse/Expand Public Constructors
  Name Description
Public Method BPRMF

There is no summary.

Collapse/Expand Public Methods (see also: Protected Methods)
  Name Description
Public Method Virtual AddFeedback

There is no summary.

Public Method Virtual AddItem

There is no summary.

Public Method Virtual AddUser

There is no summary.

Public Method Virtual ComputeFit Compute approximate fit (AUC on training data)
Public Method Virtual Iterate Perform one iteration of stochastic gradient ascent over the training data. One iteration is iteration_length * number of entries in the training matrix
Public Method Virtual LoadModel

There is no summary.

Public Method Virtual RemoveFeedback

There is no summary.

Public Method Virtual RemoveItem

There is no summary.

Public Method Virtual RemoveUser

There is no summary.

Public Method Virtual ToString

There is no summary.

Public Method Virtual Train

There is no summary.

Collapse/Expand Protected Methods
  Name Description
Protected Method CheckSampling

There is no summary.

Protected Method Virtual RetrainItem Retrain the latent factors of a given item
Protected Method Virtual RetrainUser Retrain the latent factors of a given user
Protected Method Virtual SampleItemPair Sample a pair of items, given a user
Protected Method Virtual SampleOtherItem Sample another item, given the first one and the user
Protected Method SampleTriple Sample a triple for BPR learning
Protected Method Virtual SampleUser Sample a user that has viewed at least one and not all items
Protected Method Virtual UpdateFactors Update latent factors according to the stochastic gradient descent update rule
Collapse/Expand Public Properties
  Name Description
Public Property FastSamplingMemoryLimit Fast sampling memory limit, in MiB
Public Property ItemBias Use the first item latent factor as a bias term if set to true
Public Property IterationLength One iteration is iteration_length * number of entries in the training matrix
Public Property LearnRate Learning rate alpha
Public Property RegI Regularization parameter for positive item factors
Public Property RegJ Regularization parameter for negative item factors
Public Property RegU Regularization parameter for user factors