Machine Learning - Wyatt's Notes
sources:
- text: Standard textbook reference
What This Site Covers
Machine learning fundamentals through advanced topics: supervised learning (regression, classification), unsupervised learning (clustering, dimensionality reduction), neural networks, model evaluation, feature engineering, and deployment. Emphasis on practical implementation with Python/scikit-learn and TensorFlow.
Why This Matters
Machine learning is transforming industries from healthcare to finance to autonomous systems. Understanding the mathematical foundations (linear algebra, calculus, probability), algorithm design, and practical implementation details is essential for building models that generalise beyond training data. These notes bridge theory and practice, covering both the “why” and the “how” of modern ML systems.
Getting Started
Start with the mathematical foundations: linear algebra (vectors, matrices, eigenvalues), calculus (derivatives, gradients), and probability (distributions, Bayes’ theorem). Then learn the core algorithms: linear regression, logistic regression, decision trees, and k-nearest neighbours. Build up to neural networks and deep learning once the fundamentals are solid.
Practical Applications
- Computer vision: Image classification, object detection, and segmentation using CNNs and transformers. Applications include autonomous vehicles, medical imaging, and quality control.
- Natural language processing: Text classification, sentiment analysis, and language generation using transformers. Applications include chatbots, translation, and content moderation.
- Recommendation systems: Collaborative filtering and content-based recommendations for e-commerce, streaming, and social media platforms.
Key Topics
- Supervised Learning — Linear models, trees, ensembles, SVMs
- Unsupervised Learning — K-means, PCA, t-SNE
- Deep Learning — CNNs, RNNs, transformers, attention
- MLOps — Model serving, monitoring, A/B testing
Resources
- Scikit-learn documentation (scikit-learn.org) — Reference for classical ML algorithms
- TensorFlow tutorials (tensorflow.org/tutorials) — Hands-on deep learning guides
- fast.ai courses (course.fast.ai) — Practical deep learning for coders
Intuition
Machine learning is a subset of artificial intelligence focused on learning from data: Rather than following explicit rules, ML systems improve their performance on tasks through experience. The field combines statistics, computer science, and domain expertise.
Why it matters: ML is transforming industries from healthcare to finance to transportation, creating both opportunities and challenges for society.
The key insight: Data quality matters more than model complexity — the best algorithm with bad data performs worse than a simple algorithm with good data.
Common Mistakes
Confusing training accuracy with generalisation: A model can memorise training data (overfitting) while failing on new data. Always evaluate on a held-out test set. Training accuracy alone is misleading.
Ignoring data preprocessing: Raw data often contains missing values, outliers, and different scales. Feeding unscaled features into gradient-based optimisers causes slow convergence or divergence. Always normalise/standardise features before training.
Assuming more data always helps: More data helps only if the data is representative and relevant. Adding noisy or biased data can hurt performance. Data quality matters more than quantity.
Study Approach
Start with the mathematical foundations before jumping into algorithms. Linear algebra, calculus, and probability are prerequisites. Then learn classical algorithms before deep learning. Always evaluate models on held-out test data.