Deep Generative Models: Teaching Machines to Paint, Write and Compose

From Theory to Practice

Master Programme: Signal Processing and Machine Learning for Big Data

ETSI Telecomunicación - UPM

This will be a two-day Seminar on Deep Generative Models

Day 1: (May 5th, 2021, from 17:00 to 20:00)

Prof. Santiago Zazo (ETSIT.UPM) will present the mathematical fundamentals of Deep Generative Models (DGM). We will address a proper definition involving three main aspects: modelling, inference / learning and generation. Boltzmann machines, normalizing flows, Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs) will be introduced.

Day 2: (May 19th, 2021, from 15:00 to 18:00)

Prof. Juan Parras (ETSIT.UPM) will present some practical examples (using Google Colab). We will introduce three illustrative examples: CycleGAN for painting, basic LSTM for writting generation and the attention mechanism for MIDI music composition.

This open seminar will be in English, and sessions will be on-line using Ms Teams

Day 1: link to Microsoft Teams