Mattia Scardecchia's repositories
Entropic-Optimization-Algorithms
Studied a class of entropic optimization algorithms for deep learning, in the unsupervised learning setting. Found evidence that they improve generalization and denoising capabilities of autoencoders.
hierarchy-learning
Learning a hierarchical model of data with neural networks.
OrchestraChess
UCI-compliant chess engine written in Rust. Reached 2300+ Elo rating on Lichess. Features a Pygame GUI.
Robustness-of-Minima-and-Finetuning
In deep learning, investigate the impact on finetuning performance of the robustness of a pretrained checkpoint to random and adversarial perturbations of inputs and weight.
Potts-model-and-phase-transitions
Numerical study of the Potts model using Monte Carlo methods. Found evidence of a phase transition whose nature depends on the number of states, confirming analytical results.
Single-Cell-Sequencing-Analysis
Analyze single-cell sequencing data and predict hypoxia/normoxia condition of cancerous cells. Presented at the Department of Oncology of Oxford University in July 2022.