Introduction to Machine Learning: Linear Models

<p>Selecting training examples randomly. (D)</p> Signup and view all the answers

<p>Using a randomized small group of training examples. (D)</p> Signup and view all the answers

<p>The parameters stop updating. (C)</p> Signup and view all the answers

<p>Batch Gradient Descent is more efficient for small datasets. (C)</p> Signup and view all the answers

<p>To avoid patterns and introduce randomness in training. (C)</p> Signup and view all the answers

<p>The cost or penalty for incorrect predictions (A)</p> Signup and view all the answers

<p>Gradient Descent (A)</p> Signup and view all the answers

<p>Mean Squared Error (MSE) (C)</p> Signup and view all the answers

<p>It is less sensitive to outliers (D)</p> Signup and view all the answers

<p>They evaluate how well predictions match actual data (A)</p> Signup and view all the answers

<p>It is differentiable (A)</p> Signup and view all the answers

<p>A type of loss that is insensitive to outliers up to a certain threshold (B)</p> Signup and view all the answers

<p>Mean Squared Error (MSE) (C)</p> Signup and view all the answers

<p>Overshooting the minimum (B)</p> Signup and view all the answers

<p>Universal Approximation Theorem (C)</p> Signup and view all the answers

<p>Processes input through weighted connections and activation functions (D)</p> Signup and view all the answers

<p>Using learning rate scheduling (B)</p> Signup and view all the answers

<p>A composition of linear transformations and activation functions (C)</p> Signup and view all the answers

<p>An appropriate activation function (A)</p> Signup and view all the answers

<p>The solution may be suboptimal (C)</p> Signup and view all the answers

<p>Combination of weighted linear transformations and non-linear activation functions (C)</p> Signup and view all the answers

<p>It maximizes the separation margin between two classes. (C)</p> Signup and view all the answers

<p>It is considered a support vector and is crucial for defining the hyperplane. (A)</p> Signup and view all the answers

<p>SVM focuses on maximizing the margin by ignoring outliers. (B)</p> Signup and view all the answers

<p>A margin that allows some data points to violate the separation rule. (B)</p> Signup and view all the answers

<p>(1/margin + ∑penalty). (B)</p> Signup and view all the answers

<p>It creates new variables using a kernel function. (C)</p> Signup and view all the answers

<p>A measure of how far any data point violates the margin. (A)</p> Signup and view all the answers

<p>It maximizes the distance from the hyperplane to the nearest points of each class. (A)</p> Signup and view all the answers

<p>To map output values between specific ranges (D)</p> Signup and view all the answers

<p>The strength of the input node (A)</p> Signup and view all the answers

<p>∑wi*xi (D)</p> Signup and view all the answers

<p>To shift the activation function curve (D)</p> Signup and view all the answers

<p>When outcomes are linearly separable (C)</p> Signup and view all the answers

<p>The output value is +1 (C)</p> Signup and view all the answers

<p>Single-layer Perceptron (C)</p> Signup and view all the answers

<p>Bias (D)</p> Signup and view all the answers

<p>A single-layer perceptron can only process linear patterns, while a multi-layer perceptron can process both linear and non-linear patterns. (A)</p> Signup and view all the answers

<p>Requires minimal training data to achieve high accuracy (A)</p> Signup and view all the answers

<p>The stage where weights and biases are adjusted based on the difference between the actual output and the desired output. (A)</p> Signup and view all the answers

<p>All of the above (D)</p> Signup and view all the answers

<p>Complexity and computational cost of training (A)</p> Signup and view all the answers

<p>Linear (C)</p> Signup and view all the answers

<p>Both A and B (B)</p> Signup and view all the answers

<p>They allow the network to learn complex non-linear relationships. (C)</p> Signup and view all the answers

A flat affine subspace that separates data points in SVM.

The hyperplane that maximizes the separation margin between classes.

The distance between the hyperplane and the nearest data points from each class.

The hyperplane that maximizes the separation margin; also called hard margin.

SVM can ignore outliers and find the best hyperplane.

Allows some points to violate the margin; used in less clear data cases.

A penalty used in SVM to measure violations of the margin.

A method in SVM to handle non-linearly separable data by transforming features.

A single-layer neural network for binary classification.

The data points fed into the perceptron model.

Parameters that determine the strength of input values in perceptron.

A term that shifts the activation function curve.

The total derived from multiplying inputs and weights.

A function that determines the output based on the weighted sum.

The simplest form of perceptron with one layer, analyzing linear data.

A perceptron model with multiple layers for complex patterns.

The phase where activation functions run from input to output layer.

The phase where weights and biases are adjusted based on error.

Challenges that can't be solved using linear models alone.

Quick predictions and handles large/small data effectively.

Computations can be complex and time-consuming.

An optimization algorithm used to minimize a function by moving in the opposite direction of the gradient.

A type of gradient descent that uses the entire dataset to compute the gradient at each iteration.

A variant of gradient descent that uses one random example at each iteration to compute gradients, enhancing efficiency.

A gradient descent variant that splits the dataset into small batches and uses these to compute the gradient for each update.

The first step in the SGD process where model parameters are randomly set before training.

A hyperparameter in the gradient descent algorithm that determines the size of the steps taken towards the minimum.

The process of randomly rearranging the training examples to ensure diversity in updates during SGD.

The point at which the algorithm has sufficiently minimized the cost function and stops updating model parameters.

A hyperparameter that controls the step size during optimization.

A theorem stating a single hidden layer neural network can approximate any continuous function.

A layer in a neural network that processes inputs through weights and activation functions.

The final layer in a neural network that provides the predicted output.

Parameters in neural networks that are adjusted through training to minimize error.

The process of an algorithm reaching a stable solution over iterations.

Quantifies the error as a cost for incorrect predictions.

The function that algorithms aim to minimize, typically involving the loss function.

Calculates the average of the squared differences between predicted and actual values, widely used for regression tasks.

Averages the absolute differences between predicted and actual values, less sensitive to outliers than MSE.

A robust loss function that combines MSE and MAE, less sensitive to outliers.

A loss function that is smoother than MSE, still sensitive to outliers but less so than MSE.

Different loss functions are suited for different types of prediction problems in regression.