Elod Pal Csirmaz
I’m principal engineer at NewsNow with a keen interest in artificial intelligence, filesystems and computer science in general, as well as in creative writing and literary criticism.
Computing

Interpreting Neural Networks by Reducing Nonlinearities during Training A short paper and sample code demonstrating a simple solution that makes it possible to extract rules from a neural network that employs Parametric Rectified Linear Units (PReLUs). We introduce a force, applied in parallel to backpropagation, that aims to reduce PReLUs into the identity function, which then causes the neural network to collapse into a smaller system of linear functions and inequalities suitable for review or use by human decision makers.
Read the paper »
More on GitHub » 
Superloop: Complex RNN Models A "superloop" is an extra loop that connects the output of a recursive neural network (RNN) back to its input in the next timestep via one or more external systems. This project contains: a framework to build RNNs with a superloop in TensorFlow/Keras, an implementation of a simple gated recurrent unit, an implementation of a registerbased memory unit, and an implementation of an attention system that allows the network to move freely along the input data. More »