PodParley PodParley

Compressing deep learning models: distillation (Ep.104)

Episode 101 of the Data Science at Home podcast, hosted by Francesco Gadaleta, titled "Compressing deep learning models: distillation (Ep.104)" was published on May 20, 2020 and runs 22 minutes.

May 20, 2020 ·22m · Data Science at Home

0:00 / 0:00

Using large deep learning models on limited hardware or edge devices is definitely prohibitive. There are methods to compress large models by orders of magnitude and maintain similar accuracy during inference. In this episode I explain one of the first methods: knowledge distillation  Come join us on Slack Reference Distilling the Knowledge in a Neural Network https://arxiv.org/abs/1503.02531 Knowledge Distillation and Student-Teacher Learning for Visual Intelligence: A Review and New Outlooks https://arxiv.org/abs/2004.05937

Using large deep learning models on limited hardware or edge devices is definitely prohibitive. There are methods to compress large models by orders of magnitude and maintain similar accuracy during inference.

In this episode I explain one of the first methods: knowledge distillation

 Come join us on Slack

Reference
The Analytics Engineering Podcast dbt Labs, Inc. Tristan Handy has been curating the Analytics Engineering Roundup newsletter since 2015, pulling together the internet's best data science & analytics articles.Tristan and co-host Julia Schottenstein now bring the Roundup to real life, hosting biweekly conversations with data practitioners inventing the future of analytics engineering.You can view full episode summaries and read back issues of the Roundup newsletter at https://roundup.getdbt.com.The podcast is sponsored by dbt labs, makers of the data transformation framework dbt. To reach our team, drop a note to [email protected]. Explicit STEM.queer() Vera Sativa Machine learning, data science, feminismo y queer anarquismo.Episodios cada 2 semanas. Explicit 天方烨谈 基因频道 华大基因专业团队倾情打造,基因科普娓娓道来! Explicit Explorers Wanted 5d20 Media, LLC We are an actual play podcast using the Numenera (http://numenera.com) Discovery and Destiny rules. Set one billion years in the future, we journey across the Ninth World. There have been eight worlds before this, where civilizations rose to intergalactic heights only to fall into ashes, leaving a world of strange relics behind them. Join our ragtag crew of messy adventurers as they navigate weird ruins, contend with criminal intrigue, and ignore their own better judgment... Repeatedly.See more (https://www.explorerswanted.fm/about)Become a Patron!Campaign Two: Hearts in Orbit<img src="https://files.fireside.fm/file/fireside-uploads/images/2/213fef3d-303d-4053-8ec2-96e695eef9f5/mDJd_g4e.png" alt="Three figures, from left: Ezr Explicit
URL copied to clipboard!