Skip to content

Details

Welcome to the seventh DataTalks HFA meetup!

DataTalks HFA will bring together local data scientists, researchers, statisticians and data enthusiasts around advanced data science, machine learning and AI.

Until further notice, all of the community's events will be held online.
Our seventh meetup will focus on GANs (generative adversarial networks).
Kfir Levy, assistant professor at the Technion's Electrical Engineering department, will present his research on how to train and evaluate GANs.
Tomer Golany, PhD candidate at the Technion's Computer Science department and a research software engineer at Google will present his work on using GANs for ECG synthesis with the goal of improving heartbeat classification.

♦ Time: June 24th, 18:00
♦ Location: Online
♦ Background: Basic knowledge in data science and machine learning is advised

Abstracts for the talks:
** Training and Evaluating Generative Adversarial Networks (GANs) **
We consider the problem of training and evaluating generative models with a Generative Adversarial Network (GAN). Although GANs can accurately model complex distributions, they are known to be difficult to train due to instabilities caused by a difficult minimax optimization problem, and it is not trivial to evaluate them. In this work, we investigate the problem of training GANs through the lens of game theory.
This leads to a new training and evaluation methods that build upon ideas from game theory and online learning.
We provide theoretical guarantees, and develop an efficient heuristic guided by our theoretical results, which we apply to commonly used GAN architectures. On several tasks our approach exhibits improved stability and performance compared to standard GAN training.

** SimGANs: Simulator-Based GANs for ECG Synthesis to Improve Deep ECG Classification **
Generating training examples for supervised tasks is a long sought after goal in AI. In this talk I will explore the problem of heart signal electrocardiogram (ECG) synthesis for improved heartbeat classification.
ECG synthesis is challenging: the generation of training examples for such biological-physiological systems is not straightforward, due to their dynamic nature in which the various parts of the system interact in complex ways. However, an understanding of these dynamics has been developed for years in the form of mathematical process simulators. We study how to incorporate this knowledge into the generative process by leveraging a biological simulator for the task of ECG heartbeat classification. Specifically, we use a system of ordinary differential equations (ODE) representing heart dynamics, and incorporate this ODE system into the optimization process of a generative adversarial network to create biologically plausible ECG training examples.
We will show empirical evaluation and show that heart simulation knowledge during the generation process improves ECG classification.

You may also like