JAX compilation of RDDL description files, and a differentiable planner in JAX.

Implementation of fast.ai deep learning courses: "Practical Deep Learning for Coders", "From Deep Learning Foundations to Stable Diffusion"

Optimizer Comparison Study - Empirical analysis of SGD vs Adam performance on MNIST with various initialization and scheduler configurations

A pure NumPy implementation of Ridge Regression (L2 Regularization) from scratch. Features vectorized Minibatch Stochastic Gradient Descent (SGD) and manual hyperparameter grid search without using scikit-learn.

Classification of positive and negative IMDB movie reviews with neural network, training with SGD and Genetic Algorithms comparison

My repository for python implementation of a few prominent machine learning algorithms

A practical comparison of classical optimization algorithms (GD, SGD, Momentum, Adam, RMSProp, Adagrad, Newton) analyzing convergence speed, stability, and loss minimization for machine learning.

Comparative study of User-Based Collaborative Filtering vs Matrix Factorization for recommendation systems. Performance analysis on MovieLens 100k dataset evaluating RMSE, training time, and prediction latency.

A simple, from-scratch C++ Feedforward Neural Network with zero dependencies, built to demystify the fundamentals of deep learning.

A set of PyTorch/Jupyter notebooks exploring optimization algorithms and training strategies (e.g., SGD/Adam variants) with convergence and loss/metric visualizations.

Implementing PyTorch Optimizers from Scratch

SGDOX: Sistema de gestión documental desarrollado en Ecuador con React + Symfony. Diseñado como una alternativa moderna y soberana a Quipux, con trazabilidad legal completa, firma electrónica integrada (FirmaEC) y cumplimiento de los estándares EGSI y LOPDP.

EnsLoss: Stochastic Calibrated Loss Ensembles for Preventing Overfitting in Classification

Reinforcement Learning Implementation Inspired by Bilibili Professor Zhao Shiyu's Lecture at Westlake University

Implemented Perceptron, Logistic Regression, Linear SVM, and SGD with proper scaling and regularization. Compared decision boundaries, training dynamics, and coefficient-based interpretability, and outlined strategies for imbalance and hyperparameter tuning in fast, production-friendly models.

My undergraduate project on matrix completion.

SGD Machine learning implementation investigation in Rust

Linear Regression experiments on the California Housing dataset across five phases, using NumPy and scikit-learn only (no pandas). Includes EDA, polynomial features, SGD with scaling, residuals, 5-fold CV, and an LNCS-style report with figures.