Cecilia De Marinis
IAAC – Master in Robotics and Advanced Construction
Software III
Faculty Assistant: Nikol Kirova


 Photo Credits: Mateusz Zwierzycki



The course will introduce participants to classification and regression using supervised machine learning. The neural networks will be applied to the specific datasets, which the students will pre-and post-process. The following tasks will be demonstrated:

Dataset preparation Supervised Learning requires high-quality, high-volume datasets to work with. Once the dataset is large enough, neural networks can yield robust and accurate results. During the course, the students will use existing, open-source datasets, as well as prepare their own. The part of the course will focus on the quality of the data as well as methods used to generate datasets. 

Classification – understanding patterns Classification will be useful whenever a binary decision must be made – if a pattern seems like something we want to detect. The pattern can be just about anything – a cat in an image, a word in a recording, or a shape in a parametric model. 

Regression – estimating values While classification uses neural networks for pattern recognition, regression will benefit by using that tool for value estimation. Again, the biggest challenge is to obtain a source dataset. Interpreting results Finally the course participants will learn how to interpret the outputs of a neural network. With many parameters, the “knob-turning” task can be very confusing when training a network. By studying the behaviour of small neural networks in-depth, the students will get an understanding of each of the parameters and how to change them in various scenarios.