LSTM: Mastering Sequences Over Time

End-to-End LSTM Application: Multi-Dimensional Forecasting and Textual Sequence Analysis

In this final assignment, students are required to build a robust Long Short-Term Memory (LSTM) network to solve a complex sequence prediction task. You will choose between two tracks: 1) A multivariate time-series forecasting model using real-world financial or weather data, or 2) A sequence-to-sequence model for Natural Language Processing. The project must demonstrate proficiency in data normalization, handling vanishing gradients, implementing dropout for regularization, and fine-tuning hyperparameters such as look-back windows and hidden unit counts. You are expected to provide a Jupyter Notebook containing your data pipeline, model definition, training loops, and a comprehensive analysis of the model's performance compared to a baseline such as a SimpleRNN or GRU.