Marco Martinolli's Projects
A TensorFlow implementation of the Differentiable Neural Computer.
Example of a typical repository with the basic structure to be included
Development of the memory dynamics in AuGMEnT network to enable learning on hierarchical tasks, like 12ax. We introduced a differentiable leaky control system that allows the network to forget non-relevant cues. Thus, we implemented a variant of the AuGMEnT network, named hybrid AuGMEnT, which employs an hybrid memory with both leaky and non-leaky cells in order to learn on multiple termporal scales.
Deep Learning library for Python. Convnets, recurrent neural networks, and more. Runs on TensorFlow or Theano.
A mesh simplification strategy for non-planar spatial regression with a penalized term. This is a joined project for NAPDE and APSC courses.
A C++ library implementing an edge collapse-based simplification routine for three-dimensional surface grids with distributed data
One Shot Learning using Memory-Augmented Neural Networks based on Neural Turing Machine architecture in Tensorflow
Examples and exercises for the PACS course
Computation using data flow graphs for scalable machine learning