I have tried SVM and multilayer neural perceptron on this dataset but can’t get accuracy more than 39%… Why?
Pokemon_Classification
hey @manishasingla67 ,
Performance of any model depends on various different factors.
For Example with multilayer perceptron it can depend upon the number of layers , number of nodes in each layer , activation, optimizer, loss function , input format , is normalization needed or not , learning rate , etc.
So you can’t that if you change you can increase your accuracy , it just kind of mere experiment which you need to perform and check whether the performance improves or not.
In this challenge , SVM’s wont be able to provide you with good results as there are images and for training them into a SVM , you will be getting a high dimensional data and SVM won’t be able perform that good.
Go with neural networks , tune them , use CNN’s as they work really good with images , try with adam / rmsprop optimizers, take loss as categorical crossentropy with softmax as the output layer activation and try varying the learning rate .
I hope this would increase your model performance.
Thank You and Happy Learning 