how i learn

how i learn machine learning /artificial intelligence as a Google DeepMind scholar BEFORE YOU BEGIN you need to spend some time understanding python basics (loops, functions, arrays) how data is stored (tables, vectors, matrices) plotting data and reading graphs why? data handling is fundamental to ML if you skip this, you won't understand why 'randomly' changes you won't know if bad results are from the model or the data A machine doesn't understand meaning. It only learns from numbers. If the data is messy, the learning will be too. this part can be boring, but necessary, especially if you want to be great the mathematical foundation ML is essentially applied stats and linear algebra Linear Algebra: Understanding tensors, matrices, and vectors. Most ML operations are just large scale matrix multiplications. Calculus: Specifically 'Partial Derivatives' and the 'chain rule' These are engines behind Gradient Descent (how models learn) Probability & Statistics: Concepts like Bayes' Theorem, distributions, and hypothesis testing. This helps you understand uncertainty and data patterns. I grinded this book for my interviews. MATHEMATICS FOR MACHINE LEARNING Marc Peter Deisenroth, A. Aldo Faisal, Cheng Soon Ong Machine Learning Theory Most real world problems are solved using these three techniques. Supervised Learning: -Regression: Predicting Continuous Values -Classification: Predicting Categories (Spam vs not spam) -Key Algorithms: Linear Regression, Logistic Regression, Decision Trees, SVM Unsupervised Learning: -Clustering: Grouping similar data points together -Dimensionality Reduction: Simplifying complex data points, but keeping important parts Key Algorithms: K-Means Clustering, PCA Reinforcement Learning: -Learning through trial and error to achieve a goal (mostly used in robotics and gaming) DEEP LEARNING a subset of machine learning Understand the Basic Math of a single Neuron: The Perceptron: Learn how a single "unit" takes inputs, applies Weights and Biases, and passes them through an Activation Function (like ReLU or Sigmoid). Backpropagation: This is the most important concept in all of AI. It's the mathematical "undo" button that allows the model to calculate its error and fix itself. The Workflow: Build a "Shallow" Neural Network from scratch in Python (using just NumPy) to see the math in action. DEEP LEARNING a subset of machine learning Understand how Computer Vision works: Once you understand how flat layers work, you learn how AI "sees." This is the easiest way to visualize how deep learning layers extract features. Convolutional Neural Networks (CNNs): Learn how "filters" scan an image to find edges, then shapes, then objects. Key Concepts: Pooling (simplifying data), Padding, and Dropout (preventing the model from just memorizing the data). project idea you can look into to (learn on the hob) A model that identifies handwritten digits (MNIST) or classifies images of dogs vs. cats. DEEP LEARNING a subset of machine learning Understanding Sequences: The Old Way (RNNs & LSTMs): Understand Recurrent Neural Networks. They process data one step at a time (like reading a sentence word-by-word). Learn why they were "slow" and "forgetful." The Attention Mechanism: It allows a model to look at an entire sentence at once and decide which words are most important to each other. The Transformer: This is the architecture that replaced RNNs. Learn about the Encoder (understanding input) and the Decoder (generating output). DEEP LEARNING a subset of machine learning Large Language Models and Generative AI (Think of ChatGPT or Gemini): Large Language Models (LLMs): Learn about Pre-training (teaching a model on the whole internet) vs. Fine-tuning (specializing it for a specific task). Embeddings: How to turn words or images into "coordinates" in a giant multi-dimensional map. RAG (Retrieval-Augmented Generation): How to connect an AI to a private database so it can answer questions about specific files. RESOURCES MATHEMATICS FOR MACHINE LEARNING two fantastic books, Mathematics for ML is arguably the best ML math book. Simply AI breaks all the concepts of AI down, teaches anyone you can learn and understand it. 100% recomend. (just download the online version for free, not hard to find) AI SIMPLY FACTS MADE FAST AI SUMMER Google AI Essentials Certificate of Completion Google AI Essentials Course: 9+ hour course with a certification, after go for the Google Cloud Gen AI specialisation course. AI Summer is the best website to learn anything AI. Been using it for years.