0 to 10

-1 to 1

0.0 to 1.0

1 to 100

It can handle multiple classes easily

It can model probabilities for binary outcomes

It is faster than other algorithms

It does not require any input variables

To map z to a float bounded between 0.0 and 1.0

To increase the learning rate

To reduce the number of parameters

To make the network deeper

y = $$\frac{1}{1 + e^{-z}}$$

$$y = z^2 + b$$

$$y = w^T x + b$$

y = log(z)

Mean Squared Error (MSE)

Binary Cross-Entropy (BCE)

Hinge Loss

Categorical Cross-Entropy

To increase the number of input variables

To estimate the values of the parameters

To reduce the number of output variables

To increase the training sample size

A categorical value between 1 and 10

A float value between 0 and 1

An integer value only

A binary value of 0 or 2

$$y^n + (1 - y^n)$$

$$y^n - (1 - y^n)$$

It is always positive and less than 1

It is always negative

It is always greater than 1

It is always zero

$$(y^n - t^n)$$

$$(y^n - t^n)x_i^n$$

$$y^n(1-y^n)x_i^n$$

$$(y^n - t^n)x_i$$

η

w

b

N

To estimate the output of the learning algorithm

To define the threshold for decision

To use them at test time for prediction

To calculate the loss function

w₁x₁ + w₂x₂ + b = 0

w₁x₁ + w₂x₂ = 1

w₁x₁ - w₂x₂ + b = 0

w₁x₁ + w₂x₂ - b = 0

0

1

0.5

2

The model's ability to minimize false positives

The model's ability to distinguish between classes across various threshold settings

The model's ability to maximize true negatives

The model's ability to increase recall only

FPR = FP / (FP + TN)

FPR = TP / (TP + FN)

FPR = TN / (TN + FP)

FPR = TP / (TP + FP)

To start plotting the ROC curve from the origin

To maximize the AUC

To minimize the error rate

To calculate the F1-score

It computes the area for each trapezoid under the ROC curve and sums them up

It calculates the mean squared error

It finds the maximum value of TPR

It multiplies precision and recall

The classifier is making random guesses.

The classifier is a perfect classifier.

The classifier cannot separate positive and negative classes.

The classifier labels all negative instances as positive.

The classifier is a perfect classifier.

The classifier is making random guesses.

The classifier is likely to effectively separate positive and negative classes.

The classifier labels all positive instances as negative.

To store the predicted probabilities.

To define the range of values for calculating TPR and FPR.

To store the true labels of the data.

To calculate the area under the ROC curve.

Calculate the confusion matrix

Split the dataset into training, development, and test sets

Set the threshold to 0.5

Plot the ROC curve

0.2

0.4

0.6

0.9

To map input values to a range between 0 and 1

To calculate the mean squared error

To initialize the weights

To plot the decision boundary

0.01

0.1

1.0

0.001

To plot the F1 score against thresholds using a binary classifier's probability predictions.

To train a neural network for image classification.

To calculate the mean squared error for regression.

To visualize clustering results.

Because it balances both precision and recall.

Because it only considers accuracy.

Because it ignores false positives.

Because it maximizes the number of predictions.

It shows how the balance between precision and recall changes.

It increases the number of features in the model.

It reduces the training time.

It eliminates the need for validation data.

The output is 1 only when both x1 and x2 are 1.

The output is 1 when either x1 or x2 is 1.

The output is always 0.

The output alternates between 0 and 1.