Webinar on Compression of Deep Neural Networks by IBM [Sep 28, 4:30 PM]: Registration Open

Webinar on Compression of Deep Neural Networks by IBM [Sep 28, 4:30 PM]: Registration Open

IBM Webinar
IBM Webinar
About IBM

International Business Machines Corporation is an American multinational technology company headquartered in Armonk, New York. It was founded in 1911 in Endicott, New York, as the Computing-Tabulating-Recording Company and was renamed “International Business Machines” in 1924.

About the Webinar

The Webinar on Compression of Deep Neural Networks is organised by IBM on Sep 28 at 4:30 PM.
The large number of parameters in most of the state-of-the-art DNNs make them compute-intensive, resulting in high infrastructure demands and latency. Hence, it is crucial to reduce the computational demands of DNNs to successfully meet the latency and resource requirements of a production environment (which can be a shared resource like cloud or a low resource system like mobile phones).
In this talk, we shall be mostly focussing on CNNs and transformer architectures like BERT. We shall begin with the common state of the art compression mechanisms for CNNs – these include a filter and weight pruning, quantization, encoding, low-rank decompositions etc. In the latter half of the presentation, we shall cover more advanced compression techniques for BERT like distillation, word vector elimination.

About the Speaker
  • Anamitra Roy Choudhury is a Research Staff Member at the Learning and Reasoning group in IBM Research-India. Currently, he is looking at efficient methods of compression of deep neural models so as to reduce the inference time and memory footprint. This will enable complex models to be run efficiently on low resource systems like mobile/edge devices, or in a shared environment like a cloud. Prior to this,  he was involved in optimization and parallelization of different scientific applications on massively parallel architectures – these include designing kernels for the Exascale architecture, optimization of Graph500 and HPCC benchmarks (in particular RandomAccess) on Blue Gene/Q,  parallelization of financial engineering applications etc. He is a PhD in CSE from IIT Delhi and his thesis is in approximation algorithms for job scheduling problems.
  • Saurabh Goyal is a Research Engineer at the Learning and Reasoning group at IBM Research-India. He is primarily involved in developing novel methods for optimizing the DNN models commonly used for Computer Vision and NLP tasks which helps in reducing their model size, compute resources and inference time. Prior to that he has worked on developing Machine Learning algorithms that can be deployed on severly resource-constraint IoT devices. His research work has led to publications in top-tier conferences like ICML & IEEE Cloud. He is a MS(Research) in CSE from IIT Delhi and a B.Tech from IIT Kanpur.
How to Register?

Interested participants can register for the webinar through this link.
For more details, click the link below.

Webinar on Compression of Deep Neural Networks by IBM


Subscribe to our newsletter

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *