Machine Learning Series

This series represents my exploration into different domains of machine learning, from computer vision and robotics to natural language processing and audio analysis. Each project tackles unique challenges and demonstrates different ML approaches and techniques.

The "Dexterous Tree" robot is a segmented pillar with a neural network that developed using evolutionary techniques to dodge objects in its environment when falling. This project explores how evolutionary algorithms can train neural networks for dynamic obstacle avoidance and adaptive movement in robotic systems.

This sound classification algorithm utilizes a convolutional neural network to distinguish the source and environment of sampled audio clips. The system demonstrates the application of deep learning to acoustic pattern recognition, enabling automatic identification and categorization of different audio signals based on their contextual characteristics.

A natural language processing project that uses bidirectional LSTMs to learn an artist's style from their most popular work and reproduce a new set of lyrics for the user. This system analyzes patterns in creative writing and generates coherent, stylistically consistent text that mimics the learned artistic style.

Each project represents a different branch of machine learning: computer vision for robotic manipulation, generative models for creative content, and signal processing for audio classification. Together, they demonstrate the versatility and wide-ranging applications of modern ML techniques.

This series provided hands-on experience with multiple ML frameworks, data preprocessing techniques, model architecture design, and performance optimization. The projects also highlighted the importance of domain-specific considerations when applying machine learning to real-world problems.