First application of PyTorch an ML

I have a (relatively small) dataset, 640,000 datapoints. I am able to create a model from this which has 8 inputs and generates 4 outputs.

The model predicts reasonably well, but I would like to further improve the model but I am unsure how to go about setting up the layers, number of neurons and number of layers to change or improve my model.

Research of done on how to set up the layers, etc… is very vague and hasn’t helped shed any light on this for me.

Any guidance would be greatly appreciated.

The architecture of the model depends upon the dataset. If your dataset is complex you need to increase the number of layers in model. This is not the hard and fast rule but it give you the idea. There are also a lot of other parameters as well which need to me adjust or added in model architecture like dropout, batch normalization, regularization, epochs, batch size, learning rate and optimizer as well. You should know what are these and how it will effect the model performances. By doing so you can add the easily improve/made you model from scratch.