Assignment_1_250595_srikar

[tlssrikar07]

No description provided.

[Parth-2412]

So here are some good practices and suggestions you should keep in mind going forward:

1. The parameters of a model should be defined in the class itself as self.parameter_name. In your code, you pass the weights variable back and forth between the train and test function. Instead storing it as self.weights would be better
2. Class names should be defined in capital
3. In your normalization function, you do not need to loop through j from 0 to 3. Numpy can handle that through vectorization.
4. Do not hard code the train test split. Determine it programatically.
5. The train test split is not a part of the model itself. It should be a separate function.
6. Ordering of the influencers is straight up wrong. You either one hot encode the variable (99.99% of the time) or you order them correctly based on the semantic meaning of the values of the influences. For example here if you do want to go the ordering way (which 99.99% you don't) you should order it this way: Mega -> Macro -> Micro -> Nano.
7. Develop a habit of commenting the shapes after each numpy operation.
8. The error seems pretty high and again this is directly related to point no. 6
9. You should implement min-batch/stochastic gradient descents as well. Now you have only batch GD.

[Parth-2412]

IMPORTANT
Update your PR in accordance with the instructions given. It was stated that you should upload your assignment notebook in a folder with your name