Pierre-Simon Laplace
- 3rd year Econ @ Bocconi
- Currently exploring Extreme Value Theory
With Braico and Vitto we explored ARC–AGI with Reinforcement Learning:
- ARC (v1) : Custom DSL and action space with Wolpertinger Actor-Critic for large combinatorial domains.
- ARC (v2) : Added learnable action embeddings and ViT state encoding with a pretrained world model.
- ARC (v3) : World modeling.
Under the CS department @ Bocconi we are working on:
- 'RL World Models': Benchmarking JEPA vs. ViT encoder–decoder world models in RL environments.
- Statistics-30457 : Vanilla risk parity vs. alternative model portfolios.
- Econometrics-30462 : Demand modeling using cross-sectional and panel techniques.
- Weather Derivatives : Climate modeling with Ornstein-Uhlenbeck processes using Wavelet Networks.