(Machine Learning) Use SVM from sklearn to classify non-linearly sperable datase
ID: 3604723 • Letter: #
Question
(Machine Learning) Use SVM from sklearn to classify non-linearly sperable datasets. Refer to the example in sklearn http://scikit-learn.org/stable/auto_examples/svm/plot_iris.html, you can use this code or part of it in your solutions. Load (using load_breast_cancer) datasets from sklearn (datasets.load_breast_cancer()):
a. select and evalute the "best kernal SVM" and the "worse kernel SVM" model could fit this dataset (you can empirically select the hyperparameters values), justify your answer.
Explanation / Answer
SVM deals with support vector machines.They are associated with learning algorithms which have supervised learning models.It use the Kernel trick which perform a non-linear classification whose task is to map the inputs into high dimensional fetured spaces.The dataset that we need in it will basically depends on the choice of data function.Then it will be more easy to perform or understand.We may use the ideas from the concentric circles,spiral shaped classes or from the nested banana shaped classes.Iris dataset can also be platform for computing this.SVM is used to classify the pixels of the image.The distances are misclassified in the non-linear datasets.It has a misclassification error but with the smaller margin in it but it is not considered upto that much level but it focuses largely on the hyperplane eith large or big margin.
The kernel either best or worse can be choosen based on selection of cross validation based model and its a very tricky issue.The choice of kernel depends on the magnitude of relative distance between the elements.Basically ,the kernel depends on the distance between the two points.Bochner's theorem can be used for very well explaination for this result.So,here is the homogenious meaning of selecting the kernem that will be fit for this datasel.But hopefully and targetly,the best Kernel SVM will be the appropriate and the best choice or this type of data manipulation.
Related Questions
Navigate
Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.