🔹Yes we can: estimating model performance without target data
🔹Wojtek Kuberski
Learn about two new algorithms that tackle an unsolved problem: how to estimate model performance without target data. We're introducing Confidence-Based Performance Estimation (CBPE) and Direct Loss Estimation (DLE), which offer first-of-their-kind solutions. We'll start by discussing why performance estimation is important and what can affect a model after it's deployed. The talk concludes with a detailed look at CBPE and DLE, showing why they're important new tools for data scientists.