Q.1. Study the below data and answer the following questions. Marks] Note: Positive class label is...

Transcribed Image Text:

Q.1. Study the below data and answer the following questions. Marks] Note: Positive class label is depicted by red circles in the below figure. Negative class label is depicted by blue triangles in the below figure. Round all the calculation to 4 decimal places. -4 -3 (-2,2) (-2,0) -2 (-2,-1) (-1,3) regularization is used. (-1,0) -1 X₂ 4 (0,4) 3 (0,3) 2+ 1 (0,1) -1 -2- (0,-2) -3- -4 0 (0,-1) (1,2) (1,0) 1 (2,0) 2 [3+1.5+2.5 = 7 f 3 (3,0) 4 X₁ A. Apply the logistic regression with batch gradient descent for two iterations. For first batch iteration use the points on and within the positive (first) quadrant where both features are positive. For second batch iteration use the points on and within the other three quadrants where any feature may take negative values. Use learning rate = 0.7, Wo = W2 X1 0, W₁ =0, W₂ = 0, WX= Wo + W₁ X₁ + W2 X2 and assume no E B. Compute the cost(loss) after every batch learning and state your observation. Assume the second iteration results obtained in previous part A. as the final weights using which predict the class of (X₁, X2) = (1,-1). Use default 0.5 as the threshold. C. Construct the confusion matrix for given first batch of training data in the figure w.r.t below two experiments and comment your choice on the suitability of the threshold using the True Positive Rate Vs False Positive rate: i) Use the decision threshold of 0.6. If h(xi) >=0.6, then map xi to positive class. ii) Use the decision threshold of 0.45. If h(x) <0.45 then map xi to negative class. Q.1. Study the below data and answer the following questions. Marks] Note: Positive class label is depicted by red circles in the below figure. Negative class label is depicted by blue triangles in the below figure. Round all the calculation to 4 decimal places. -4 -3 (-2,2) (-2,0) -2 (-2,-1) (-1,3) regularization is used. (-1,0) -1 X₂ 4 (0,4) 3 (0,3) 2+ 1 (0,1) -1 -2- (0,-2) -3- -4 0 (0,-1) (1,2) (1,0) 1 (2,0) 2 [3+1.5+2.5 = 7 f 3 (3,0) 4 X₁ A. Apply the logistic regression with batch gradient descent for two iterations. For first batch iteration use the points on and within the positive (first) quadrant where both features are positive. For second batch iteration use the points on and within the other three quadrants where any feature may take negative values. Use learning rate = 0.7, Wo = W2 X1 0, W₁ =0, W₂ = 0, WX= Wo + W₁ X₁ + W2 X2 and assume no E B. Compute the cost(loss) after every batch learning and state your observation. Assume the second iteration results obtained in previous part A. as the final weights using which predict the class of (X₁, X2) = (1,-1). Use default 0.5 as the threshold. C. Construct the confusion matrix for given first batch of training data in the figure w.r.t below two experiments and comment your choice on the suitability of the threshold using the True Positive Rate Vs False Positive rate: i) Use the decision threshold of 0.6. If h(xi) >=0.6, then map xi to positive class. ii) Use the decision threshold of 0.45. If h(x) <0.45 then map xi to negative class.

Answer rating: 100% (QA)

BWe will apply the following formula to determine the cost loss f... View the full answer