Demystifying hyperparameters versus parameters
The key difference between a hyperparameter and a parameter is how its value is generated. A parameter value is generated by the model during the model-training phase. In other words, its value is learned from the given data instead of given by the developer. On the other hand, a hyperparameter value is given by the developer since it can't be estimated from the data.
Parameters are like the heart of the model. Poorly estimated parameters will result in a poorly performing model. In fact, when we said we are training a model, it actually means that we are providing the data to the model so that the model can estimate the value of its parameters, which is usually done by performing some kind of optimization algorithm. Here are several examples of parameters in ML:
- Coefficients (
) in linear regression - Weights (
) in a multilayer perceptron (MLP)
Hyperparameters, on the other hand, are a set of values that support...
