Choosing Best Kernel and Best Parameters in SVM Module |AI Sangam
Choosing Best Kernel and Best Parameters in SVM Module

Choosing Best Kernel and Best Parameters in SVM Module |AI Sangam


Notice: Undefined variable: fields in /home/vmp7bxbd1swb/public_html/blog/wp-content/plugins/wp-about-author/wp-about-author-admin.php on line 158

Objective of this Article: This blog is about Choosing Best Kernel and Best Parameters in SVM Module

Understanding Parameters associated with Kernel function (Support Vector Machine)

Cost

It is also known as Penality Parameter. It determines influence of misclassification on the objective function. It controls cost of misclassification on the training data. If the value of C is high, the model will choose more data points as support vector while low value of C will make the model to choose less data points as support vector. High value leads to high variance and low bias which leads to problem of overfitting while low value leads to low variance and high bias which leads to the problem of underfitting.

Gamma

It tells influence of data points on the decision boundary. If the value of Gamma is high, decision boundary will depend on data points near to the decision boundary while for low value, decision boundary depends on far away points. When we see different shapes of decision boundary either wiggly or straight line, it depends on Gamma.

Kernel and What it do????

Kernel measures the similarity between two data points. It helps to perform some calculations fast which otherwise will take a lot of time. Sometimes it is not possible to draw the decision boundary for some problem. We need to convert it into higher dimensional space (Feature Space) to find the hyperplane. Kernel helps to find the hyperplane in the high dimensional space without much increase in the computational cost. It transforms one dimensional space into other dimensional space.

Different Types of Kernels (Only two are explained here)

• RBF: Radial Basis Kernel Function is the set of mathematical function. These are suitable when number of observation are larget than number of features.

• Linear Kernels: These are used when number of features is more than number of observations. I will add all the description in the end of the article.

Please see it through the video link Choosing Best Kernel and Best Parameters in SVM Module and I have made more videos on it whose link you can find in the description of video

One Vs All approach with Grid Search in SVM 

Real time prediction for multiclass classification using python and jupyter notebook

Leave a Reply