This project covers a wide range of tasks: mining and parsing data about recipes, which have a high
dimensional representation, using NLP, processing and modeling that data in order to apply
machine learning algorithms for various objectives (ranking, classification, generating new data…)

key words: NLP, GANs, recommendation, scoring, auto-encoders, data-generation

This project covers a wide range of tasks: mining and parsing data about recipes, which have a high dimensional representation, using NLP, processing and modeling that data in order to apply machine learning algorithms for various objectives (ranking, classification, generating new data…)

This project consists in implementing a gated recurrent neural network (RNN) model that has simple, predicable and non-chaotic dynamics, achieving performance comparable to well-known gated architectures, such as LSTMs and GRUs, on the image captioning task. This stands in stark contrast to more standard gated architectures, whose underlying dynamical systems exhibit chaotic behavior.

Original paper : https://arxiv.org/abs/1612.06212

Implementing a gated recurrent neural network (RNN) model that has simple, predicable and non-chaotic dynamics, achieving performance comparable to well-known gated architectures, such as LSTMs and GRUs, on the image captioning task. This stands in stark contrast to more standard gated architectures, whose underlying dynamics exhibit chaotic behavior.

This Project consists into handling a large dataset of patents (500 GB + in textual relational tables) and processing them in order to be able to classify each patent by its IPC (international patent classification) in a parallel fashion on a cloud environment.