Regularization weight in train_svm function

Hello, 

I was wondering, since you specify (in Support_Vector_Machine_from_Scratch.ipynb) that "lambda is the 1/epoch (change rate gets smaller while it pass epochs)", shouldn't lambda be 1/epoch instead of 1/epochs? 

