Calling python, machine learning experts

[Spartan]

1 hour ago, Ellen said:

@Spartan is there any limit to choose how much variance should be captured by pca ? For example my data lo first two principal components give about 51%. Ivi use cheskoni train cheste I'm getting warast accuracy. I keep manually changing variance value. Best results are at 95% but then it chooses the no. of components almost equal to number of features which is pointless. So how to pick the range?

 

@Ellen variance distribution purely depends on what ur final model expectation would be. 100% ki try chestam with 0 noise but kudaradu kada.

we just need to keep adjusting those parameters. I think they are working on having a algorithm to define that bias-variance offset 

but still it depends on ur model requirements 

[Ellen]

1 hour ago, Spartan said:

@Ellen variance distribution purely depends on what ur final model expectation would be. 100% ki try chestam with 0 noise but kudaradu kada.

we just need to keep adjusting those parameters. I think they are working on having a algorithm to define that bias-variance offset 

but still it depends on ur model requirements 

True 

That's awesome 👍

So appati daka aite repeat trials a Inka 

[Spartan]

40 minutes ago, Ellen said:

True 

That's awesome 👍

So appati daka aite repeat trials a Inka 

Yes