Hours:
20 hours (5 credits)

Room:
Aula Riunioni del Dipartimento di Ingegneria dell’Informazione, Largo L. Lazzarino 1, 56122 Pisa, Edificio A, piano 6

Short Abstract:
Modern industry and social life has started a quick process of automatization, equipping (wearable) devices with sensors, actuators and artificial intelligences. Any device is supposed to work according to expected functionalities, but sometimes they generate unexpected behaviors due to changes of their environment, external events, age or failures.
In this course, we will discuss on how to find anomalies in IoT Systems (networks of intelligent devices) by exploiting the huge amount of data their sensors produce, by comparing different anomaly detection techniques up to neural networks and deep learning.

Course Contents in brief:

1. Find the unexpected: anomaly detection (4h)
   1. Classification
   2. Clustering
   3. Python for dummies
2. IoT Data Streams (4h)
   1. Definition
   2. Data cleaning and filtering
   3. Data transformation
   4. Using python for data manipulation
3. A gentle introduction to Deep Learning (8h)
   1. Machine/Deep Learning
   2. Why Neural Networks?
   3. Gradient Descent and Back Propagation
   4. Recurrent Neural Networks
   5. Autoencoders
   6. Variational Autoencoders
   7. Pytorch Implementations
4. Finding anomalies in IoT Data Streams with Neural Networks (4)
   1. Supervised approach
   2. Unsupervised approach
   3. Pytorch solutions

Schedule:

1. Find the unexpected: anomaly detection - November 25, 14.30 - 18.30
2. IoT Data Streams - November 26, 14.30 - 18.30
3. A gentle introduction to Deep Learning (part 1) - November 27, 14.30 - 18.30
4. A gentle introduction to Deep Learning (part 2) - November 28, 14.30 - 18.30
5. Finding anomalies in IoT Data Streams with Neural Networks - November 29, 09.00 - 13.00