Michael Zakhary

Using a model’s robustness to outliers to detect them Michael Zakhary · Follow Published in Towards Data Science · 8 min read · Sep 28, 2023 — Photo by Will Myers on Unsplash Problem Statement The problem of outlier detection can be tricky, especially if the ground truth or the description of what is an outlier is ambiguous or based upon multiple factors. Mathematically speaking, an outlier can be defined as data points more than

Is low inductive bias essential for building general-purpose AI? Michael Zakhary · Follow Published in Towards Data Science · 6 min read · 6 days ago — Photo by BoliviaInteligente on Unsplash In today’s ML space, we find ourselves surrounded by these massive transformer models like chatGPT and BERT that give us unbeatable performance on just about any downstream task, with the caveat being the requirement of huge amounts of pre-training on upstream tasks first.