Where to get a professional to help with my notebook?

1

I have tried all the tricks I know on my fairly simple classification problem and it still always predicts 1 (there are only two classes). Perhaps my data is not good enough, but I want to rule out stupid mistakes. I’m out of questions and I don’t expect anyone to read my code (although here it is on Kaggle).

So where/who can I turn to to hire someone to read my notebook?

2

You could create a post in the “Jobs” category here with your search for an engineer.

That being said, I would generally recommend to run a quick test by overfitting the model to a small dataset (e.g. just 10 samples) and make sure your model is able to perfectly fit it.
If that’s not the case, you can quickly iterate and check the code for any issues.

Just by skimming through the code it seems that your model is not using any activation functions:

class Net(nn.Module):
    
    def __init__(self):
        super(Net, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(201, 400),
            nn.Linear(400, 200),
            nn.Linear(200, 50),
            nn.Linear(50, 2)
        )

    def forward(self, x):
        return self.model(x)

and can thus be seen as a single linear layer.

3

thanks for checkout it out.

I have tried activation functions, but it doesn’t help.

Overfitting a small dataset works. Does this mean the problem is data?

4

This would at least mean that the current training code seems to properly update the model.
Once you were able to overfit the data, you could scale up the problem slowly (i.e. add more data stepwise) and tune the hyperparameters of your model with it.

5

If I understand correctly this just means learning rate, loss function, layers and so on? If so, this is what I’m doing. Maybe I’ll get to a result eventually.