TTharunintharunpandya.hashnode.dev·5d ago · 9 min readFrom Lines to Decisions: Why Regression Fails at ClassificationLast time, right at the end, I asked a question and then walked away from it. Will this order be late, or won't it? The final answer isn't a number on a continuous scale. It's a class: yes or no. Ever10
TTharunintharunpandya.hashnode.dev·Aug 22 · 10 min readPolynomial Regression: Teaching a Line to BendTake our best straight-line model so far, the one that fit our five-order table perfectly. Now widen the range a bit. Some orders close by, a couple kilometers away. Some far, ten or twelve kilometers10
TTharunintharunpandya.hashnode.dev·Aug 19 · 10 min readFeature Scaling in Machine Learning: Why It MattersTake the exact same model, the exact same data, the exact same code. Change one thing: write the distance column in meters instead of kilometers. Nothing else moves. Run it, and the model doesn't just10
TTharunintharunpandya.hashnode.dev·Aug 16 · 9 min readMultiple Linear Regression: When One Input Isn't EnoughPicture two different orders. Both are exactly 3 kilometers away from the restaurant. One arrives in 18 minutes. The other takes 35 minutes. If distance is the only thing our model looks at, it has no10
TTharunintharunpandya.hashnode.dev·Aug 13 · 10 min readGradient Descent: Walking Downhill BlindfoldedIn the last post, we built a way to measure exactly how wrong a guess is, using the cost function. Given any w and b, we can now compute a single number that tells us how badly that line fits our data10
