Due on May 2

  1. (15 points) Let's build a U-net like predictor for the stock market data. Please construct the structure as shown below. You may want to trucate the training and test input lengths to factors of 4 to avoid complication from downsampling and upsampling.

    1. (5 points) Plot the the estimated price along with the ground truth for both training and test data in two separate plots.

  2. (20 points, extra-credit) Create an LSTM predictor for the stock market data. Please use two layers of LSTM (each with 50 hidden dimensions) for your predictor. Plot your result as in question 1. For simplicity, you can take the current prices of other stocks as input and the current MSFT price as output. This blog post should be helpful.

alt text