This repository has been archived by the owner on Sep 14, 2020. It is now read-only.

Commits

Commits on Apr 21, 2017

Fix README.md
peterrudenko
committedApr 21, 2017

Commits on Apr 10, 2017

Merge pull request #1 from zeakey/master
peterrudenko
committedApr 10, 2017
fix README markdown syntax error
Zhao Kai
committedApr 10, 2017

Commits on Mar 11, 2017

Merge branch 'hotfix/1'
peterrudenko
committedMar 11, 2017
Removed the deprecated failing tests.
peterrudenko
committedMar 11, 2017
Merge branch 'release/v0.5'
peterrudenko
committedMar 11, 2017
The project is no longer updated, so fixed the readme.
peterrudenko
committedMar 11, 2017

Commits on Jul 27, 2016

Added tanh activation and set it as the default one.
peterrudenko
committedJul 27, 2016

Commits on Jul 26, 2016

Implemented experimental dropout.
peterrudenko
committedJul 26, 2016
Added adjustable activation type and fixed LSTM architecture.
peterrudenko
committedJul 26, 2016

Commits on Jul 22, 2016

Removed TrainingContext that was a big mistake of mine.
peterrudenko
committedJul 22, 2016
Updated the copyright headers.
peterrudenko
committedJul 22, 2016
Added gradient clipping threshold.
peterrudenko
committedJul 22, 2016
Merge branch 'release/v0.4'
peterrudenko
committedJul 22, 2016
Updated readme.
peterrudenko
committedJul 22, 2016
Added one more serialization test.
peterrudenko
committedJul 22, 2016
Removed the remains of OpenCL dead code.
peterrudenko
committedJul 22, 2016
An include fix.
peterrudenko
committedJul 22, 2016
Namespace fix.
peterrudenko
committedJul 22, 2016
Returned installing build-essential to the Travis config.
peterrudenko
committedJul 22, 2016
CMake script argument fix for TravisCI.
peterrudenko
committedJul 22, 2016
Removed OpenCL-related stuff from CMake scripts.
peterrudenko
committedJul 22, 2016
Removed OpenCL-related stuff from Travis config.
peterrudenko
committedJul 22, 2016
Removed OpenCL support.
peterrudenko
committedJul 22, 2016
Added one loop instruction to the VM to minimize static network program size.
peterrudenko
committedJul 22, 2016

Commits on Jul 21, 2016

Revert "Implemented GPU-driven batch training."
peterrudenko
committedJul 21, 2016

Commits on Jul 7, 2016

Implemented GPU-driven batch training.
peterrudenko
committedJul 7, 2016

Commits on Jul 4, 2016

Merge branch 'release/v0.3'
peterrudenko
committedJul 4, 2016
Removed all classes of hardcoded networks.
peterrudenko
committedJul 4, 2016

Commits on Jul 1, 2016

Clipping the gradient in the backward phase to avoid ReLU activations going NaN.
peterrudenko
committedJul 1, 2016
Made initial random weights and biases way smaller.
peterrudenko
committedJul 1, 2016

Commits on Jun 30, 2016

Added some missing headers.
peterrudenko
committedJun 30, 2016

Commits on Jun 27, 2016

Replaced sigmoid with leaky ReLU.
peterrudenko
committedJun 27, 2016

Commits on May 29, 2016

Merge branch 'release/v0.2'
peterrudenko
committedMay 29, 2016

Commits on May 19, 2016

Fixed feed-only networks generation.
peterrudenko
committedMay 19, 2016