@glemaitre58 Let me know your thoughts on this proposed wording related to hyperparameters.
Current wording:
In the previous modules, we show how to create, train, predict, and even evaluate a predictive model. However, we did not change on the models’ parameters that can be given when creating an instance. Indeed, we used the default values chosen by scikit-learn.
These parameters are called hyperparameters : they are parameters used to control the learning process, for instance the parameter
k
of the k-nearest neighbors. They should not be confused with the other parameters that are inferred during the training process. These parameters define the model itself, for instancecoef_
for the linear models.
Can we edit, for clarity:
In the previous modules, we showed how to create, train, predict, and even evaluate a predictive model. We used the models’ default parameters, as chosen by scikit-learn, when creating an instance of the model. For example, for k-nearest neighbors, we initially used this default parameter:
'kneighborsclassifier__n_neighbors': 5
before trying other model parameters.
These parameters are called hyperparameters: they are parameters used to control the learning process, for instance the parameter
k
of the k-nearest neighbors. They are choices about which numbers we pass to the model for fitting the function. They should not be confused with the other parameters that are inferred during the training process. These parameters define the model itself, for instancecoef_
for the linear models.