4. In logistic regression, we model the class probability p(C = 0lx) = o(wx + wo)...

 

Transcribed Image Text:

4. In logistic regression, we model the class probability p(C = 0lx) = o(wx + wo) where a(z) = is the sigmoid function. 1 1+exp(-2) Assume that we have the following training data: Training example X 1 where 1 2 3 4 5 0 2 1 -1 x2 0 1 0 -1 1 t 0 1 0 p(C = 1x); w) = 0 Suppose we have already applied training examples 1 to 4 to the training, and the weight vector at this time is known to be w = [2] with wo = -1. We are ready to apply our next training sample (training example 5) to update the weights. Use stochastic gradient descent, 1 w; (t+1) < w; (t) 2x; (i) (t() p (C = 1|x); w)) - 1 exp(-w7x-wo) 1+exp(-w7x-wo) 1+exp(wx+wo) find the new weight vector w after this update if the learning rate = 0.5. 4. In logistic regression, we model the class probability p(C = 0lx) = o(wx + wo) where a(z) = is the sigmoid function. 1 1+exp(-2) Assume that we have the following training data: Training example X 1 where 1 2 3 4 5 0 2 1 -1 x2 0 1 0 -1 1 t 0 1 0 p(C = 1x); w) = 0 Suppose we have already applied training examples 1 to 4 to the training, and the weight vector at this time is known to be w = [2] with wo = -1. We are ready to apply our next training sample (training example 5) to update the weights. Use stochastic gradient descent, 1 w; (t+1) < w; (t) 2x; (i) (t() p (C = 1|x); w)) - 1 exp(-w7x-wo) 1+exp(-w7x-wo) 1+exp(wx+wo) find the new weight vector w after this update if the learning rate = 0.5.

Answer rating: 100% (QA)

In this problem we are asked to update the parameter vector w using the stochastic gradient descent ... View the full answer