ResearchProjectFramework

Try to design a better practice.

Design

1. output_data folder will be huge, so keep it in a separate place, so when we copy paste another experiment, we don't need to copy the output data.

2. every experiment need a hypothesis, so we would remember what we are testing.

3. whenever there's a new idea, formulate it as a hypothesis, and copy paste another exp folder for that.

4. common package includes all the tools needed. copy that with the exp folder, so we won't mess up the dependencies of earlier experiments.

5. use start_experiment.sh to start the experiment, so we know what we executed.

6. commit everything except output_data onto github every night when work ends.