What we're about
Upcoming events (1)
We would like to thank Pendo for hosting us PHYSICALLY
18:00-18:30 Gathering and snacks
18:30-18:45 Welcome words from our host
18:45-19:15 The notorious POC | Dr. Inbal Budowski-Tal, Sr. Director of ML @ Pendo
19:15-19:45 Deep Learning Approach to Partial Differential Equations-based Problems | Leah Bar, OriginAI and Tel-Aviv University
19:45-20:00 A short break
20:00-20:30 A Subset-Based Strategy for Faster AutoML | Teddy Lazebnik, CTO@DataClue and PostDoc Researcher@UCL
The notorious POC | Dr. Inbal Budowski-Tal, Sr. Director of ML @ Pendo
Ever found yourself in an endless POC that opened more and more research questions? Ever found yourself delivering a POC, and expected to deliver the project to Production a minute later? Or perhaps, presenting a great POC, deploying to Prod, and then struggling with poor results?
The Proof of Concept is the very first stage of every ML project and the most notorious one.
In this talk, I will share my experience with good (and bad!) POCs, and discuss how to avoid some common pitfalls.
Deep Learning Approach to Partial Differential Equations-based Problems | Leah Bar, OriginAI and Tel-Aviv University
We present an unsupervised deep-learning approach for the solution of partial differential equations.
The proposed framework is very general, where boundary conditions and other regularizations can be easily integrated
in the loss function. The method is demonstrated on the eigenvalue problem and Electrical Impedance Tomgraphy application.
A Subset-Based Strategy for Faster AutoML | Teddy Lazebnik, CTO@DataClue and PostDoc Researcher@UCL
Automated machine *AutoML) learning frameworks have become important tools in the data scientists’ arsenal. However, when the dataset is large, the overall AutoML running times become increasingly high. In this lecture, we present AutoML optimization strategy that tackles the data size, rather than configuration space by wrapping existing AutoML tools, and instead of execute them directly on the entire dataset, we find a small yet representative data subset to work with