The variable thresholds trick
There are many ways to improve a classifier, but the most inspiring way to improve it is to really think hard on how you want to apply your model. The reason is because there might just be an amazing opportunity to use variable thresholds, which can really make the model more flexible in production. 00:00 Introducing variable thresholds 01:05 Business case for variable thresholds 04:01 Understanding the maths 07:13 Towards a notebook 11:23 Measure the bonus Relevant links: https://probabl-ai.github.io/calibrat... https://github.com/probabl-ai/youtube... Website: https://probabl.ai/ LinkedIn: / probabl Twitter: https://x.com/probabl_ai Bluesky: https://bsky.app/profile/probabl.bsky... Discord: https://discord.probabl.ai We also host a podcast called Sample Space, which you can find on your favourite podcast player. All the links can be found here: https://rss.com/podcasts/sample-space/ #probabl

Generating Periodic Features for Seasonal Timeseries

Accuracy rarely matters

AlphaFold - The Most Useful Thing AI Has Ever Done

Connect-4 Classification in Python: Data Preparation, Modeling & Evaluation Explained

Automatically tuning thresholds

When calibration beats metrics

How to remedy a badly calibrated machine learning model

How To Think SO CLEARLY People Assume You're A Genius

What Nobody Tells You About Being a Quant

Turing Award Winner: Disagreeing with Google, Postgres, Future Problems | Mike Stonebraker

Metadata routing in scikit-learn

If you need calm, you'll feel this on your skin (comfort for restless minds)

Your Accuracy Is a Lie — Here's How to Fix It (The Architect's Guide to Robust Model Validation)

Japan – Schweden Highlights | Gruppe F, FIFA WM 2026 | sportstudio

الرقية الشرعية للشفاءمن السحروالعين والحسد حصن من الشيطان رقية البيت والاولاد بصوت القارئ سعيد حمدان

Decision and Classification Trees, Clearly Explained!!!

How To Learn So Fast It’s Almost Unfair

Why Smart People Struggle To Articulate Themselves Clearly

Understanding Thresholds in Machine Learning

