Deep convex network with joint use of nonlinear random projection, restricted boltzmann machine and batch-based parallelizable optimization

Deep convex network with joint use of nonlinear random projection, restricted boltzmann machine and batch-based parallelizable optimization

  • CN 102,737,278 A
  • Filed: 03/30/2012
  • Published: 10/17/2012
  • Est. Priority Date: 03/31/2011
  • Status: Active Application
First Claim
Patent Images

1. method comprises that following computing machine can carry out action:

  • Reception is used to train the training data of the dark protruding network of the module that comprises a plurality of interconnection, and each module in the module of wherein said a plurality of interconnection comprises linear layer and non-linear layer;

    AndMake processor at least partly come wherein to train said dark protruding network to comprise the corresponding weight matrix of output of the non-linear layer of study and at least one module to train the part of said dark protruding network based on mode in batches based on said training data.

View all claims
    ×
    ×

    Thank you for your feedback

    ×
    ×