Machine learning models are not functional out of the box, they require a period of training during which they learn to represent the dataset presented to them. This project aims to visualise this training process in the case of a simple linear regression model which trains on a randomly generated dataset consisting of 100 data points. Below you can tweak some of the hyperparameters present in machine learning models.

There are a few things to note before using the tools below. This is a very simple problem and hence sometimes the random initialisation of the model can prove sufficiently good resulting in little change in loss during training, as a result of this, it is recommended to perform a few re-runs and observe the behaviour. Furthermore, the model proves capable of learning the data within the first few iterations because of which it is recommended to explore the low learning rates first to best visualise the convergent behaviour of the model. Finally, at large learning rates, the model may be sufficiently perturbed to cause a divergence resulting in a loss of NAN in which case it is advised to reload the dataset and try again.

To break the start command, simply refresh the page.