Question

Explain the following concepts in the correspondent learning models:

1. What is the C parameter in the logistic regression and the support vector machine? What is the alpha parameter in the neural network (MLPClassifier in Scikit-learn)? What is their relationship with overfitting and underfitting?

2. What is the purpose of kernels in support vector machines?

3. Explain what is the random_state in Sciket-learn machine learning models such as logistic regression, MLPClassifieres.

Answer 1

1. The C paprameter in logistic reression and SVM is a regularization paramerter and is equivalent to inverse of regularization strength, i.e., 1/lambda and smaller the value of C, more it it regularized. Alpha is also a regularization parameter for MLP. Regulariation helps to simplify the model artificially. With the increasing the regularization, the effective model complexity decreases and hence prevents overfitting and decreasing or completely removing Regularization helps the model to learn better.

2. Kernels takes the input and modify the data into required form. Mainly it helps to transforms a non-linear dimention surface to a linear dimention surface in higher dimentions.

3. random_state can be used when we are regularizing the model and it helps to control the random number generation. Using an int will produce the same results across different calls.