ML.AI 2026
The full 6 week curriculum and resources from IITG.ai's summer course on Data Science and Machine Learning — from basic Python to complex neural networks.
Course Structure
6 weeks, basic Python to complex neural networks — ML.AI 2026 (concluded)
Week 1
Covers Python basics along with essential libraries like Pandas, NumPy, and Matplotlib, and introduces core concepts of Exploratory Data Analysis (EDA) and Feature Engineering.
Week 2
Familiarises you with Machine Learning basics, covering Linear Regression, Logistic Regression, Regularization, Naive Bayes, Evaluation Metrics and more, along with the Sklearn library.
Week 3
Covers important ML algorithms like SVM, KMeans, and KNN, and introduces unsupervised learning, Principal Component Analysis (PCA), dimensionality reduction, clustering, and anomaly detection.
Week 4
Covers Ensemble Learning and boosting algorithms like XGBoost, CatBoost, and LightGBM, along with semi-supervised approaches like self-training and label propagation.
Week 5
Starts your journey into neural networks and deep learning, covering the fundamentals of neural networks, activation functions, and backpropagation.
Hackathon
Participants put everything they learned to the test in the official Kaggle competition.
Getting Started with Python (Setup + Introduction to Python)
Setting up a Python development environment with Anaconda and learning the basics of Jupyter Notebook for interactive programming. Introduction to the Python language for machine learning.
Anaconda Installation + Jupyter Tutorial (Video)
Covers the basic data structures required to get started with machine learning.
Introduction to Numpy, the scientific computing library of Python. This will help you in doing heavy mathematical calculations
Learn to clean, summarize, and visualize datasets using Pandas and Matplotlib, with an intro to Exploratory Data Analysis (EDA) to uncover patterns, anomalies, and insights.
Explore and analyze datasets through EDA, handle missing data, detect outliers, and engineer features to improve model performance and extract deeper insights.
Machine Learning Introduction, Linear Regression
Introduction to Machine Learning and in particular, Supervised Learning. Covers Linear Regression model and the famous Gradient Descent algorithm.
Machine Learning, Supervised Learning (Regression) Introduction
(1 to 4)
Supervised Learning Introduction
(Blog)
Linear Regression Introduction
(9 to 14)
Linear Regression
(Blog)
Gradient Descent
(15 to 20)
Gradient Descent
(Blog)
Multivariate Linear Regression
(21 to 24)
Multivariate Regression
(Blog)
Going deeper into Linear Regression (25 to 30)
Covers the Sklearn Implementation of Linear Regression, and Gradient Descent in Python from Scratch. Also introduces you with a very popular Classification algorithm.
Gradient Descent Implementation of Linear Regression from Scratch
Completes the Logistic Regression model, and covers its Sklearn Implementation, and Gradient Descent from Scratch.
Gradient Descent Implementation of Logistic Regression from Scratch
Covers Regularization for addressing the problem of Overfitting, and its Implementation. Covers basic terminologies like Bias and Variance. Introduces you with the popular Naive Bayes Algorithm.
Overfitting and Regularization
(37 to 39)
Overfitting and Regularization
(Blog)
L1 and L2 Regularization
(Blog)
Multinomial Naive Bayes Algorithm
Gaussian Naive Bayes Algorithm
Naive Bayes
(Blog)
Introduces Evaluation Metrics to judge both Classification and Regression Models, and their Implementation. Covers ROC Curve and AUC.
Normal Equation Implementation (Blog)
SVM
SVM is a strong supervised learning technique used for classification and regression applications. It determines the optimum hyperplane for categorizing data points.
Comprehensive Explanation with Code for SVM
KNN is a simple yet powerful algorithm that can be used for both classification and regression tasks. Now some introduction to Unsupervised Learning, which includes algorithms that learn from unlabelled data.
What is Unsupervised Learning? (Blog)
Supervised vs Unsupervised Learning Video
KMeans clustering is a popular unsupervised learning algorithm for grouping data points into K clusters based on similarity.
Also includes the implementation of several clustering techniques.
Different Clustering Algorithms (Notebook)
PCA is a widely used unsupervised learning technique that reduces the dimensionality of high-dimensional data by identifying the most important features or principal components
Now, back to supervised learning. Decision Trees are a popular supervised learning algorithm for classification and regression tasks. They use a tree-like model to make decisions based on input features.
Decision Trees Explanation (Video)
Using Decision-Tree Classifier (Notebook)
Also covers an example of Unsupervised learning.
Decision Tree From Scratch (Notebook)
Unsupervised Learning Example: Anomaly detection using Isolation forest
Ensemble learning refers to a machine learning technique in which multiple models, known as base models or weak learners, are combined to form a stronger and more accurate predictive model.
Weak Learners vs Strong Learners
Today, we explore ensemble learning with a focus on Random Forest, a variant of Bagging, and compare it with standard bagging. We also delve into boosting and compare it to bagging, highlighting their differences.
Feature Importance in Random Forest
For the final day of Ensemble Learning, we delve into boosting algorithms like AdaBoost, GradBoost, XGBoost, LightGBM and Catboost. Boosting is a powerful technique that combines homogeneous weak learners to enhance predictive models.
XGBoost vs LightGBM vs CatBoost — Comparison Table & Deep Dive
Hyperparameter Tuning Intuition & Overfitting in Boosting
Semi Supervised Learning a technique that combines a small amount of labeled data with a large pool of unlabeled data to build better models. We discuss several methods like self-training, label propagation, etc. along with their practical applications including the maths behind the frameworks discussed in the blog.
Detailed dive into Semi-Supervised Learning
(Optional Exercise)
Pseudo Labeling — Semi-Supervised Technique
Self-Training Workflow & Real-World Applications
Reinforcement LearningA foundational concept in machine learning where an agent learns to make decisions by interacting with an environment. Unlike supervised learning, RL doesn't rely on labeled data—instead, it learns from feedback in the form of rewards or penalties.
Markov Decision Process (MDP) & Bellman Equation
Today we will be starting with basics of DL. First we will be seeing basics of DL and Neural Networks. Following this a reference notebook is also there.
Neural Network for binary Classification (Watch C1W2L01-C1W2L03)
Following from yesterday, we will be seeing more about NN, mathematics behind it and how it works.
Today we will be starting with how to do calculations in a vectorized manner and more technical details of implementation
Continuing from yesterday , we will look into what are Activations and basics of it. Further we will be looking into Gradient Descent in NN and it’s calculations
Vectorization more details (Watch C1W3L04-C1W3L05)