Pokemon goa trip

#1

What is the max achievable accuracy (with SVM) for this challenge?

#2

Hey @yashikakhurana00 ,
with SVM generally we have seen a score between 40-50% but if some preprocessing on data or other parameters is done , this score might reach too 60-65% .
I have never seen SVM scoring more than that on this dataset.

I hope this helps.
Thank You. :slightly_smiling_face:.

#3

I have a score of 53 pct. Can you suggest what more can I do ?

#4

hey @yashikakhurana00 ,
I SVM you can’t do things as much , as there aren’t many parameters to be altered.
What you can try is ,

  1. Augmenting your data to generate more samples. Augmentations like , brightness change , contrast change , a bit of zoom , shifting images etc.
  2. You can normalize / scale image pixels by dividing them with 255. To get all the pixel values between 0-1.
  3. Try playing with different values of learning rate.
  4. Change your algorithm.

I hope these might help you improve your performance.
Thank You and Happy Learning :slightly_smiling_face:.

1 Like
#5

Thank you for specifying these noteworthy points.
By augmentation of data, do you mean I should make changes in brightness etc. and add them as new images to increase the training data?

#6

That is your choice.
Most times its better to add that newly generated augmented data.
But you need to keep in mind that , you dont increase the data size so much that , your model gets easily overfit on it.
So , if you are generating only 1 or 2 new records of each image , then it might increase your performance.

1 Like
#7

I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.

On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.