GANs and it’s application in speech technology

AAII Technical Lecture Series (ATLS-2)

Yann LeCun described Generative Adversarial Networks (GAN) as “the most interesting idea in the last 10 years in Machine Learning”. GAN is a class of deep >generative networks, developed by Ian Goodfellow and other researchers at the University of Montreal, including Yoshua Bengio, in 2014. Generative means, in simple >words, is that it can generate new content, e.g., it can create new artwork or a new piece of music that never exists.

To explore this exciting topic, AAII invited Dr. Nagaraj Adiga, a senior research scientist from Zapr Media Labs, who has vast experiences of using GAN in different speech applications.

The outline of the talk was -

• Basics of GAN
• Different types of GANs and their applications
• Applications of GAN
• Demo - How to train a GAN model

Related links:

• Presentation slides
• GAN Paper
• GAN Zoo
• GAN Applications

About speaker: Dr. Nagaraj Adiga is currently working on the GANs and autoregressive architectures for different speech applications like Enhancement, Recognition, Synthesis, and Voice Conversion. Before joining Zapr Media Labs as a research scientist, he worked as a post-doctoral researcher at the University of Crete, Greece, and a Signal Processing Engineer in Apple Inc. in developing neural-based text to speech (TTS) systems in noisy scenarios. Linkedin Google Scholar

Speaker: Nagaraj Adiga (Ph.D., IITG | Post-Doc, University of Crete, Greece), Senior research scientist, Zapr Media Labs, Bangalore
Topic: GANs and it’s application in speech technology
Date: 28 November 2020
Time: 6 PM (IST)