Mathematics preliminaries: Convex optimization, Gradient-based methods, Brief review of linear algebra, Basics of probability, Bayes theorem, Probability distributions.
2. Introduction to various machine learning techniques: Motivation of ML in mechanical engineering, Classification of various learning algorithms, linear regression, Polynomial regression, logistic regression, Principal component analysis, Support vector machines, Introduction to infromation theory, Underfitting, Overfitting, Bias, and Variance.
3. Neural networks: Neural network, Supervised learning with neural network, Forward and backpropagation, Implementation of a neural network to solve a classification problem in Python, Regularization, Hyperparameter optimization, Weight initialization, Implementation in Keras/Tensor flow, Convolution neural network, Recurrent neural network, LSTM, Autoencoders.
If time permits (unlikely), we can cover things like a decision tree and random forest.
4. Project: Application of neural network into mechanical engineering.
There is no prerequisite for the course. However, some basic knowledge of Python might be useful; having said that, we can have some Python sessions conducted by TA. Do not get scared by this requirement just be ready to put slightly more effort.
Attendance is mandatory in the course.
Evaluation: Coding assignment +project (40-50 %), tests and exams (50-60%). I am yet to decide whether to have an exam based on the coding.
The programming assignments should be done in Python.
The grading will be relative to extreme grades (A and F).
The course is going to follow a strict plagiarism policy. Anyone found cheating will be awarded F without any exception.
Deep Learning by Ian Goodfellow, Yoshua Bengio, and Aaron Courville, The MIT press.
Neural Networks and deep learning by Michael Nielsen, ebook.
Pattern recognition and Machine learning by Christopher Bishop
Deep learning by Christopher Bishop