Mocha.jl <https://github.com/pluskid/Mocha.jl> v0.0.6: Mocha is a deep learning framework for Julia. Check out our new tutorial on unsupervised pre-training with stacked denoising auto-encoders <http://mochajl.readthedocs.org/en/latest/tutorial/mnist-sDA.html>. v0.0.6 2014.12.31 - Infrastructure - Numerical gradient checking in unit-tests (@pcmoritz <https://github.com/pcmoritz>) - Simple ref-counting for shared parameters - Network - RandomMaskLayer, TiedInnerProductLayer, IdentityLayer - Freezing / unfreezing some layers of a network to allow layer-wise pre-training - Documentation - A new tutorial on MNIST that compares unsupervised pre-training via stacked denoising auto-encoders and random initialization
Happy New Year! - pluskid
