Master Practical Machine Learning: A Data Scientist's Guide
What you will learn:
- Develop a robust understanding of AI, Machine Learning, and Deep Learning for data science applications.
- Master the fundamentals and inner workings of supervised learning models: Linear Regression, Logistic Regression, Support Vector Machines, Deep Neural Networks, Decision Trees, and Random Forests.
- Gain proficiency in unsupervised learning techniques for dimensionality reduction and clustering.
- Build practical machine learning models and pipelines using Python, scikit-learn, pandas, Keras, and TensorFlow.
- Solve real-world problems such as image classification, text classification, and price prediction.
Description
Become a proficient machine learning practitioner with this comprehensive course designed for data scientists and aspiring AI engineers. We'll demystify artificial intelligence, machine learning, and deep learning, clarifying their interconnectedness within the data science landscape. Learn to communicate effectively within an AI team, understanding the potential and limitations of AI projects. This course covers supervised learning techniques, including linear models (Linear Regression, Logistic Regression, Support Vector Machines), and non-linear models (Polynomial Regression, Kernel SVM, Deep Neural Networks). We'll provide a systematic approach to tackling any ML problem—from data preparation and exploratory data analysis (EDA) to model selection, evaluation, design, fine-tuning, and regularization. Each step is illustrated with practical Google Colab notebooks. Master machine learning meta-algorithms and ensemble methods such as voting, bagging, boosting, decision trees, and random forests. Delve into unsupervised learning, exploring dimensionality reduction algorithms (like Locally Linear Embedding and Principal Component Analysis) and clustering techniques (like K-means). Throughout the course, you'll utilize Python, scikit-learn, pandas, and Keras to build robust and effective models. This hands-on learning experience will empower you to confidently tackle real-world machine learning challenges.
Curriculum
Introduction
This introductory section sets the stage, providing a foundational overview of the course content and what you can expect to learn. The 'Introduction' lecture provides a concise summary of the course's scope and learning objectives (5:13).
Module 1: Introduction to AI
This module explores the fundamental concepts of Artificial Intelligence. You'll delve into 'What is AI?' (24:40), understanding its principles and applications. You'll then learn how to effectively 'Work in an AI team' (20:23), mastering communication and collaboration skills. 'AI and Society' (15:41) explores the societal impact of AI and ethical considerations. Finally, 'How AI works' (1:28:36) offers a comprehensive explanation of the underlying mechanisms.
Module 2: Supervised Learning
This module is dedicated to supervised learning. Beginning with a 'Module roadmap' (2:45), you'll dissect the 'AI models anatomy' (16:46), followed by an exploration of the essential 'Supervised Learning ingredients' (18:35). A 'Example Keras program' (19:11) provides a practical demonstration. Learn about 'Learning problems types and design patterns' (11:17), 'Losses and output layers types' (36:38), and get an optional overview of the 'Scikit-learn library' (5:06). Master 'Introduction to Optimizers' (9:10), 'Linear Regression and Normal equation' (15:57), and address 'Issues with Normal equation and closed form solution' (4:11). We cover 'Iterative solution and Gradient based optimization' (16:48), exploring 'Gradient Descent hyperparameters' (17:29) and 'Batch size hyperparameter' (9:56). Further topics include 'Logistic Regression' (4:33), 'Support Vector Machines (SVM)' (18:05), 'Non-linear models' (11:45), 'Deep Neural Networks (DNN)' (17:19), 'Overfitting and underfitting' (14:54), and 'Regularization' (9:09). The module concludes with a comprehensive 'Summary' (1:47).
Module 3: Universal Supervised Machine Learning Process
This module provides a systematic approach to supervised machine learning. It starts with a 'Module roadmap' (2:00) and covers 'Terminologies and ML process overview' (8:15), 'Defining the problem and data assembly' (12:57), and 'Losses and Metrics' (9:31). Learn about 'Decision boundaries and thresholds' (10:03), 'Precision, Recall and F1' (8:41), the 'Imbalanced data problem' (18:49), and 'Imbalanced data solutions' (15:21). You'll master the 'Model selection process' (22:04) and 'Model selection techniques' (6:50), including 'Resampling methods' (18:00). The practical application is covered in 'Data preparation' (0:50), 'Exploratory Data Analysis (EDA)' (9:23), 'Data preprocessing' (18:13), 'Baseline model' (4:10), 'Scaling up' (5:10), and 'Regularization' (7:54).
Module 4: Machine Learning Meta Algorithms
This module delves into meta-algorithms. Starting with a 'Module roadmap' (3:30), you'll revisit 'Model selection revisited and bootstrapped resampling' (9:18), explore 'Model Ensembles types and Voting method' (13:43), 'BAGGing' (5:48), 'Decision Trees' (25:50), and 'Random Forests' (4:16). Learn about 'Boosting' (12:05) through a comprehensive 'End-to-end HousePrices prediction' process encompassing EDA (22:37), data preparation (14:54), and model selection and tuning (12:27). The module concludes with a 'Summary' (1:55).
Module 5: Unsupervised Learning
This module introduces unsupervised learning. After a 'Module roadmap' (1:40) and 'Unsupervised Learning overview' (7:14), you'll learn about the 'Curse of dimensionality and Dimensionality reduction' (13:26). Understand the 'Types of dimensionality reduction - Manifold Learning vs. Projection methods' (9:53), including 'Principal Component Analysis (PCA)' (21:35). Explore 'Clustering algorithms' (9:09) and the 'K-means clustering algorithm' (25:04), concluding with an introduction to 'Semi-supervised Learning' (11:19).
Material
This section provides access to supplementary course materials, including slides (0:01).
