This repository contains a collection of small projects demonstrating my work with machine learning, data analysis and neural networks.
The goal of these projects is to explore different stages of the ML workflow: data preprocessing, model building, evaluation and experimentation with different algorithms.
This project focuses on data preprocessing and classical machine learning models using the Titanic dataset.
Key steps:
- data cleaning and handling missing values
- feature engineering
- exploratory data analysis
- training and evaluating classification models to predict passenger survival
The goal was to understand how data preprocessing affects model performance.
This project explores neural network architectures for image classification using the MNIST dataset.
The experiments start with a simple baseline Multilayer Perceptron (MLP), followed by a baseline Convolutional Neural Network (CNN). After establishing the CNN baseline, hyperparameter tuning was performed to investigate how architectural changes affect model performance.
The tuning experiments include modifications such as increasing network width and depth, as well as adding techniques like Batch Normalization and Dropout.
- Python
- PyTorch
- NumPy
- Pandas
- Matplotlib
- Scikit-learn
- TensorFlow (for dataset loading)