Semi-Supervised-Labelling

This project elaborates on solving the problem of semi-supervised labelling / clustering, by leveraging gradient-based methods (Gradient Method, Block Coordinate Gradient Descent with randomized and cyclic approach). Gradient-based methods are implemented, and compared based on publicly available data and accuracy VS CPU analysis.

Kaggle.CrowdCounting

This project is devoted to the following kaggle competition: https://www.kaggle.com/datasets/fmena14/crowd-counting

Microcontrollers

Stm32F072 Microcontrollers programming

Robust-Feature-Selection

This project is devoted to the course on Cognitive, Behavioral and Social Data. The problem in psychology, is that the selected features are model-dependent. Here we experiment on feature selection techniques, which are model-independent and robust to the dataset choice.

RestrictedBolzmannMachine

Project with deep belief network, with Omniglot dataset. Analyze feature extraction, performed by DBN, compare performance with dense neural network, supervised learning approach.