samialabed / stackelberg Goto Github PK

Implement MV LR Leader.

Improvements: Iterate on it and implement a NN MV LR Leader

LSTM​ (Long Short Term Memory): a very small Recurrent Neural Network which
could theoretically work well because the data is sequential. We will see if such an
approach is feasible given the reduced dataset size.

Evaluate and implement LSTM

We have an excel sheet we want to parse and extract relevant information from it.

• Implement a generic DAO for the Excel file

Decide on online learning strategy and implement it.

Relevant method to look at
is

@Override
public void proceedNewDay(int newDate) throwsRemoteException {
  Record newRecord = platform.query(playerType, newDate);

and then access the record to get the information

Implement a MV Non-LR model

We need a class which uses DL4J which does the following:

-extracts features for a given day
-supports the addition of new records and computes features for the newly added days
-returns all the features in some matrix/vector format compatible with DL4J models

This class will basically be the data feeder for our models.

Setup the project with maven and add deeplearning4j.org dependencies

Our DAOs can access the excel sheet, deeplearnin4j requires a custom DataVector; initialise the deeplearning4j leader class with the DataVector using our DAO