Penalty value to increase Accuracy

#1

why are we not concerned about C value (Penalty) to increase Accuracy here

#2

hey @Par1hsharma ,
i am not sure about why it is not concerned in the respective video.

But C parameter , is basically when you see bias and variance tradeoff in the training.
then you majorly need to tune it , and improve performance.

I hope this helps.

#3

what is bias and variance tradeoff and why it is not discussed in the video . why are they leaving such important topics

#4

i guess this topic is raised in some other lecture. But not worries.

So, in simple words bias is training time error and variance is testing time error.
So , for a better model performance both these values need to almost similar to each other.
if variance > bias , the model is overfitted,
if variance < bias , the model is underfitted.

Hence we need to maintain this, and this differences is called bias and variance tradeoff.

I hope this helps.

#5

what do we mean by testing /training time error . is it related to loss function?

#6

Yeah the loss value.
that we get at training time and testing time.