HW 9.3.2 Label

For HW Q9.3.2, where we train on the first 4 years of data, what would be our label value for each batch of training data (assuming the batch is 1 year)? Do we have an overall S&P 500 opening stock value that we can use as a label?

I understand for 9.3.3. that we are using the opening stock price for each company as label, but for 9.3.2. it is unclear whether this is the case.

1 Like

Concatenate all opening stock prices at each day (as time step t) for labels.

Do we use the next day’s opening prices as the label? It seems counterintuitive to use the same day’s opening prices as the label

So are we going to have 500 labels concatenated, to all be predicted for each time batch? Or are we just simply averaging them out

Yes, the next day’s opening price, and also yes your output should be a vector of dimension ~500 (or however many securities).