OHOmer Hayatinepoch.hashnode.dev·Jul 14 · 14 min readWhy Machine Learning Models Should Be Allowed to Say “I Don’t Know” Confidence, calibration, abstention, and the case for human review in real-world ML systems Many machine-learning systems behave as if every input deserves an answer. We encounter a version of this pr10
OHOmer Hayatinepoch.hashnode.dev·Jun 26 · 10 min readBackpropagation from First Principles: What I Finally UnderstoodBackpropagation becomes much less mysterious once you stop seeing it as “the algorithm that trains neural networks” and start seeing it as the chain rule moving backward through a computation graph. K10
