How Meta Optimizes Their Hyperparameters With Nevergrad

Author(s): Benjamin Bodner

Originally published on Towards AI.

The optimization platform which rules them all

This member-only story is on us. Upgrade to access all of Medium.

The kings of optimization have been working tirelessly to deliver amazing open source libraries to us continuously!

And as always, Meta delivers!

Have you ever spent days — or even weeks — tuning the hyperparameters of your machine-learning model and feeling like you’re blindly throwing darts at a wall? You’re not alone!

Hyperparameter optimization can suck if you’re not systematic or use automated tools.

I won’t like, it’s kinda my guilty pleasure to try out different solutions and see what they do — kinda like watching fire burn. But after a while it gets old and you just want results.

But what if I told you there’s a better way?

Nevergrad — Meta’s open-source hyperparameter optimization tool makes this guesswork (and the grind) obsolete. Meta’s FAIR teams claim to use it in many of their internal use cases, such as job scheduling, reinforcement learning, and image generation.

If it’s good enough for them, it’s probably worth a look, right?

Said no one ever

Ever try to cook something new, perhaps a dish from a different country that you liked in a restaurant, and blindly get all the ingredients and recipes right without looking at the… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI