Get best of the week

ML REPA Meetup at Raffeisenbank: keep broadcasting

Join the ML REPA Meetup broadcast tomorrow, February 27th : weโ€™ll discuss together whether it is possible to consider the Sacred experiment logging library as an alternative to MLflow, how to speed up A / B testing with machine learning, and how to switch from MLflow to MLPanel to support several projects.

We will be in touch at 19:00!


What we will talk about


7 p.m. Sacred - Another Library for Experiment Logging
Andrey Gurevich, Data Scientist, dostavista.global

About Sacred, Karl! A library that is worthy of the attention of the community! About the ease of integration into the code, the wow effect out of the box, the convenience of the web interface and limitations. Can it be considered as an alternative to MLflow? What are the pros and cons? For those who wanted to try and those who hear for the first time.

19:30. Speeding up A / B testing with machine learning
Nerses Bagiyan, Raiffeisenbank

Many have heard about A / B testing, few have done it themselves and few are able to automate it. Let's talk about tasks, formulation and acceleration of A / B experiments. How to reduce the variance? What to do if there are a lot of metrics? How to predict the future assessment of the effect? About this and other interesting aspects of A / B testing.

20:10. From MLflow to MLPanel: support for multiple projects, Model Registry and deployment in one interface
Mikhail Rozhkov, ML-REPA

When you have to develop several projects in parallel, you want to keep everything at hand. Here are my projects, here are the models ... and here, now I click Deploy, and everything will fly! MLflow is awesome, but it doesn't have everything you need. Therefore, the idea of โ€‹โ€‹his upgrade came up. Letโ€™s see what happened, where they bicycled, and whether it is possible to make a complete solution for managing ML projects from this prototype!

More articles:


All Articles