AI & Machine Learning

Part 2: Derivatives and Backpropagation — Making the Graph Learn Part 2 of 3 in a code-first walkthrough of micrograd, based on Andrej Karpathy’s The spelled-out intro to neural networks and backpropagation. In Part 1, we built a Value class that wraps floats and tracks how they were produced - a computation graph. But the graph alone doesn’t help us train anything. We need to know: if I nudge this input, how does the output change? ...

Part 1 of 3: Building the computation graph — the Value class, forward pass, and how Python operators silently construct the graph that makes backpropagation possible.