One-Hot Encoding

First learned about this term through Machine Learning, but surprisingly I hear is as well in Digital Circuits.

One-hot distribution

• This is just the Kronecker Delta

Machine Learning

Where there is categorical data, we use one-hot encoding to convert it into features, where each category is a feature of 1.

Danger

We also need to remove one column because that is simply a duplicate because it is a Dummy Variable (i.e. redundant information).

It seems that Andrej Karpathy didn’t worry about this however.