stock-rnn

Why we included this project

Lilian Weng built this as a teaching project rather than a forecasting tool, and it shows its age in the best way: the code is plain enough to read the architecture directly, and the two companion blog posts explain the reasoning behind each piece. The multi-stock setup is the interesting part. Instead of training one model per ticker, you feed embedding vectors for each stock into the network alongside the price series, so a single LSTM learns patterns across many tickers at once. The fetch and train scripts run without much ceremony, and the README walks you through pulling S&P 500 data from Yahoo Finance and starting TensorBoard. If you want to see how multilayer LSTM cells actually fit together in TensorFlow, this is a solid reference to work from.