MLDL-I: Machine Learning and Deep Learning - I
This course is offered at the Institute of Robotics and Automation, BUET (IRAB) by instructors Md Awsafur Rahman and Bishmoy Paul.
Course Overview
The course is designed to provide a comprehensive understanding of ML and DL techniques and their practical applications. Throughout this course, we will dive into the fundamentals of ML and DL, explore various algorithms, and gain practical hands-on experience in implementing them. Please refer to the Course Outline for more details.
Learning Objectives
By the end of this course, we aim to achieve the following learning objectives:
-
Understanding the Basics: We will develop a strong foundation in ML and DL, including their underlying concepts, terminology, and mathematical principles.
-
Exploring Algorithms: We will explore various ML and DL algorithms, such as linear regression, logistic regression, support vector machines, decision trees, neural networks, and convolutional neural networks. We will learn how to select the appropriate algorithm for different tasks.
-
Hands-on Implementation: This course places a strong emphasis on practical implementation. We will gain practical experience by working on real-world datasets, using popular ML and DL libraries such as scikit-learn and TensorFlow.
-
Model Evaluation and Optimization: We will learn how to evaluate ML and DL models using performance metrics and techniques such as cross-validation. Additionally, we will explore strategies for model optimization and hyperparameter tuning.
-
Application Development: Throughout the course, we will work on projects and case studies to apply our knowledge and skills in solving real-world problems. We will learn how to preprocess data, train models, and deploy ML and DL solutions.
Course Structure
The course will be structured as a combination of lectures, hands-on exercises, assignments, and projects. We will have the opportunity to actively engage in discussions, ask questions, and collaborate with fellow students.
Assignments
Each assignment is created as a Kaggle competition so that students can evaluate their scores in real-time and also compare their results with other students. Please note that the initial leaderboard is based on a small portion of the test data (public test), while the final standing will be based on the rest of the test data (private test). Therefore, avoid overfitting on the public test data.
- Assignment-01: Titanic Competition
- Assignment-02: Multiclass Classification with MLP
- Assignment-03: Multiclass Classification with CNN
Prerequisites
To fully benefit from this course, it is recommended to have a basic understanding of programming concepts and mathematics. Familiarity with the Python programming language will be advantageous, as most practical implementations will be done using Python libraries.