This repository gathers a series of self-contained mini-projects exploring advanced machine learning and generative modeling techniques — with a focus on time-series, stochastic processes, and financial modeling.
| File | Description |
|---|---|
SDE_Simulator_TimeSeries.ipynb |
Simulates stochastic differential equations (Brownian Motion, Ornstein–Uhlenbeck) using the Euler–Maruyama scheme. |
RealNVP_2D_vs_MLP.ipynb |
Implements a Normalizing Flow (RealNVP) on 2D datasets (spiral, Gaussian mixture) and compares it to a simple MLP regression baseline. |
FlowMatching_vs_Diffusion_1D2D.ipynb |
Compares Flow Matching and Diffusion Models on 1D/2D Gaussian mixtures — highlights sample quality, stability, and data efficiency. |
TS_RNN_Transformer_Sinus_Regimes.ipynb |
Forecasts a noisy sinusoidal time-series with volatility regimes using both LSTM and Transformer architectures. Evaluates MSE, NLL, and calibration. |
portfolio_optimization.ipynb |
Illustrates portfolio optimization (Markowitz) and basic strategy backtesting using Python & NumPy. |
Dashboard/ |
Streamlit-based dashboards for financial strategy simulation or risk analysis. |
Portfolio Strategy Simulator/ |
Modular app for portfolio backtesting (Buy & Hold, MA crossover, Vol Target, Markowitz optimization). |
VOLSURF_LAB/ |
Experimental module for volatility surface analysis and visualization. |
- Generative Modeling: Flow Matching, Diffusion, RealNVP
- Probabilistic Forecasting: Gaussian likelihoods, NLL, calibration
- Stochastic Processes: Brownian motion, Ornstein–Uhlenbeck
- Neural Architectures: MLPs, RNNs (LSTM), Transformers
- Quantitative Finance: Portfolio simulation, optimization, volatility analysis