All the official information is available here: https://corsidilaurea.uniroma1.it/user/13070
Main website of the course: https://sites.google.com/a/uniroma1.it/danielaiacoviello/didattica
Module 3 schedule: Every Thursday and Friday, starting from October 28 until the end of the lessons:
Thursday (8:30-10:00), Room 15 in san Pietro in Vincoli
Friday (8:30-10:00), Room 15 in san Pietro in Vincoli
Join the google classroom of the course for live updates on the class schedule and materials.
(if you have any problem in finding the invitation link just send me an email from your student email)
Office hours: Please send me an email to book an appointment on google meet or at DIAG.
Introduction to Machine Learning problems in process identification and control
Introduction to Deep Learning and its possible roles in control schemes
CNN, RNNs and their applications for process identification and control tasks
Markov Decision Processes, examples
Value Functions
Reinforcement Learning, examples
Q-Learning
introduction to DeepRL, DQN, DDPG
Introduction to Keras and Gym programming for process control tasks
Suggested Readings
[1] Goodfellow, Ian, Yoshua Bengio, and Aaron Courville. Deep learning. MIT press, 2016. https://www.deeplearningbook.org/
[2] Sutton, Richard S., and Andrew G. Barto. Reinforcement learning: An introduction. MIT press, 2018. http://incompleteideas.net/book/the-book.html
LECTURE PRESENTATIONS are available on google classroom
DIARY OF THE LESSONS (WILL BE UPDATED DURING THE COURSE)
29/09/2022 Introduction to the course, introduction to module 3. The role of ML in process identification and control. Supervised, Unsupervised and Reinforcement Learning.
30/09/2022 The concept of capacity, training and generalization. Fundamental elements of a ML model. First examples of regression and classification tasks in PIC.
13/10/2022 First examples in Colab. Basics of Jupyter and Python. Overfit and Underfit. Feed Forward Neural Networks with Keras. Dimensioning of input, output and hidden layers. Introduction to regularization techniques.
14/10/2022 Introduction to CNN for classification and regression tasks. convolution operation, stride, padding. Pooling (max, average) CNN for vision-related tasks.
20/10/2022 Examples of CNNs for PCBA assembly and cell counting. Guide to DNN design.
21/10/2022 Introduction to RNN. Regularization. Transfer learning.
27/10/2022 Federated Learning. Introduction to Markov Decision Processes.
28/10/2022 Examples of MDPs. Concept of state, action, reward, transition function.