Programme Content
Motivations and Introduction to different ML Paradigm
Linear Algebra for ML
Vectors and Matrices
Vector Space and Subspace
System of Linear Equations
The Concept of Rank and Independent Vectors
Inner Product Space
Norms, Positive Definite Matrix
Matrix factorisation (EVD, SVD, QR, LR, etc.)
Projection and Orthogonality
Probability and Statistics for Data science
Random Variables
Distribution and Density Functions
Conditional Probability, Bayes Theorem
Joint Distribution
Concept of Independence, Covariance, and Correlation
Introductory Statistical Inference (Likelihood, MAP, etc.)
Concept of Entropy
Mutual Information, and KL Divergence
Optimisation
Function and Derivatives
Gradient Descent
Stochastic Gradient Descents
Convex Optimisation
Formulation and Optimality Conditions
ADAM Optimiser
Hands-on Demo 1: Linear Algebra using NumPy
Concepts of Linear Algebra and Probability Basics
Optimisation with Practical ML Applications
Learning Outcome
Learners will develop a comprehensive understanding and application of linear algebra concepts, probability and statistics, and optimisation in real-world machine learning tasks.
2. Regression MethodsSimple and Multiple Linear Regression
Hands-on Demo 2: SLR/MLR
Least Squares Approach
Moving Beyond Linearity: Non-linear Regression
Hands-on Demo 3: NLR
Model Selection, Regularisation and Bias-Variance Trade-off
M2 Project: Regression application
Learning Outcome
Learners will master simple and multiple linear regression, non-linear regression, and the least squares approach, gaining practical experience through hands-on demos. They will also learn model selection, regularisation, and the bias-variance trade-off, culminating in a regression application project discussion.
3. Classical Machine LearningMotivation and Introduction to Classification Problems
Logistic Regression
Logistic Regression
Hands-on Demo 4: Logistic Regression
Decision Tree
Introduction to Decision Trees
Random Forests, Bagging, and Boosting
Hands-on Demo 5: Random Forests
Interpretability of Machine Learning Models
Hyperplanes
Concept of Hyperplane Classifier
SVM
Support Vector Machines, Kernel SVM
Hands-on Demo 6: SVM
Multi-class Classifiers
Clustering
Clustering Methods
Hands-on Demo 7: Clustering
Project
Classification Application
Learning Outcome
Learners will develop expertise in logistic regression, decision trees, random forests, and support vector machines, gaining practical experience through hands-on demos. The will also learn clustering methods and the interpretability of machine learning models, culminating in a classification application project discussion.
4. Deep LearningNeural Networks
Fundamentals of Neural Network and Feedforward Network
Concept of Training and Backpropagation
Hands-on Demo 8: ANN
Convolutional Neural Networks
Fundamentals of Convolution
Convolutional Neural Network Architecture
Hands-on Demo 9: CNN
Recurrent Neural Networks/LSTM
Introduction to Time Series and Sequential Data
Introduction to Language Modelling and NLP
Recurrent Neural Network and LSTM/GRU
Hands-on Demo 10
Graph Neural Networks
Introduction to Graph Data
Graph Neural Network Architecture
Hands-on Demo 11
Learning Outcome
Master the fundamentals of neural networks, including feedforward networks, training, and backpropagation, with practical experience through hands-on demos. Additionally, learn advanced topics such as convolutional neural networks, recurrent neural networks, graph neural networks, transformers, and generative AI, applying these concepts to real-world applications.
5. Generative AITransformers
Core mechanics — self-attention, positional encodings, causal mask
Efficiency & fine-tuning — Flash/linear attention, LoRA-FT/adapters
Multimodal extensions — vision-language models
Generative AI
Autoencoder, Variational Autoencoders, Generative Adversarial Networks (GANs)
Diffusion for images and text modalities
LLM Alignment
Alignment pipeline — SFT → reward model → RLHF/DPO/PPO
Alternative approaches — Constitutional AI, RLAIF
Details
1. Linear Regression Lab
Is there a connection between sales and different types of ad expenditure? In this lab, we try to forecast the sales of a product assuming ad sales are available.
2. Logistic Regression Lab
Sentiment Analysis of consumers. Can we directly infer the quality of any product based on its reviews?
3. Decision Tree, Random Forest, XGBoost
In-depth analysis of algorithms on benchmark datasets.
4. Support Vector Machines
Image classification on fashion MNIST dataset, intuition of soft margin, hard margin, solving SVM using CVXPY
5. Neural networks
Basic understanding and implementation of each layer of NN. Writing and understanding gradient descent/backpropagation algorithm in Python
Comparison of Neural Networks and SVM on image classification datasets
6. Convolutional Neural Networks (CNN)
Ever wondered how computers identify faces? We will see how CNN has revolutionized the field of Computer Vision
Understanding layers, visualization of the learning process, Occlusion, GRADCAM
7. Sequential Model (Recurrent Neural Network/Long Short-Term Memory)
Implementation of RNN/LSTM. Hands-on implementation for Caption/Summary generation from images/videos.
8. Understanding and implementation of Variational AutoEncoder on MNIST dataset. We will see how to encode images in a latent space of lower dimensions.
9. Is it possible to generate new images which never existed? Understanding and implementation of Generative Adversarial Networks on benchmark datasets.
10. Graph Neural Network
Are you ready to take your machine learning to the next level? Whether you want to build a recommender system for social media platforms or do drug prediction in biomedical, GNN has your back. We will see the Extension of Deep Learning on Graphs (GNN).
Introduction to several GNN variants GCN, GraphSage, etc
11. Natural Language Processing
Text Summarisation
12. Generative AI
Fine-Tuning SLMs and LLMs and Their Integration with Downstream Tasks
13. Course Project
Build your own recommender system using any of the discussed techniques (GNN, CNN, LSTM, classical ML, etc.)
