Deep Belief Nets in C++ and CUDA C: Volume 1: Restricted Boltzmann Machines and Supervised Feedforward Networks

Index

A

1. Activation file

2. Architecture of CUDA

B

1. Backpropagation of errors

   1. chain rule

   2. derivative of

   3. hidden-layers

   4. hidden neurons

   5. mean squared error

   6. MLFN

   7. neurons activation

   8. parameters

   9. SoftMax outputs

C

1. Clear all data option

2. Conjugate gradient optimization

   1. change

   2. concept of

   3. difficulties

   4. dimensions

   5. function minimization

   6. implementations

   7. multivariate

   8. negative gradient

   9. power of

   10. second-order methods

   11. simple quadratic function

   12. traditional backpropagation, momentum

   13. vectors

3. CUDA code, RBM training

   1. bias vectors and weight matrix

   2. calling parameter list

   3. components

   4. development

   5. device and clean up, parameters

   6. epoch loop

   7. fetching

   8. gradient length and dot product, reduction

   9. hidden-to-visible analysis

      1. issue efficiency and stall reasons

      2. load balancing

      3. memory access statistics

      4. occupancy

      5. pipe utilization

   10. initialization ...