Deep Learning: Classification Problems

Supervised learning where the goal is to predict a continuous output value based on labeled data.

A type of supervised learning where the goal is to predict a value on a continuous scale, given labeled data.

A type of supervised learning where the goal is to predict a value on a continuous scale, given labeled data.

A type of machine learning where the goal is to find patterns and insights from unlabeled data without explicit guidance.

An unsupervised learning technique that groups data points into clusters based on their similarity.

A specific type of clustering algorithm that aims to partition data points into k clusters by minimizing the distance between points within each cluster.

The task of identifying the optimal number of clusters to create in a dataset.

An unsupervised learning technique where the goal is to reduce the dimensionality of data while preserving as much information as possible.

A method for training neural networks where the model's parameters are adjusted in the opposite direction of the gradient of the loss function.

The amount by which the model parameters are updated in each iteration of gradient descent.

A method of updating the model parameters using the gradient of the loss function computed on a small batch of training data.

A method of updating the model parameters using the gradient of the loss function computed on a single training sample.

A method of updating the model parameters using the gradient of the loss function computed on the entire training dataset.

A local minimum in the loss function is a point where the loss is lower than any nearby point, but not necessarily the lowest point overall.

The forward pass involves running the model on input data to obtain predictions.

The backward pass involves calculating the gradient of the loss function with respect to the model's parameters.

A measure of how well a model is performing on a given dataset. It quantifies the difference between the model's predictions and the actual values.

The ability of a function to have its output smoothly change based on small changes in its input. This allows us to calculate the gradient of the function.

This mathematical tool describes the rate of change of a function at a specific point. In machine learning, it tells us how much the loss function changes with respect to each weight in the model.

A scalar coefficient used in a neural network model to adjust the contribution of different input features. These coefficients are typically represented by 'w' in mathematical notation.

The process of iteratively updating the weights in a neural network based on the calculated gradients. This aims to minimize the loss function and improve the model's accuracy.

The process of evaluating a model's performance on a separate dataset that was not used during training. This helps to assess the model's ability to generalize to unseen data.

A technique used in optimization to help the model escape local minima by incorporating the 'momentum' of past parameter updates. It considers not only the current gradient but also the direction of past movements.

The process of calculating the gradient of a function with respect to its parameters. It's essential for optimization algorithms that use gradients to find the best parameter values.

A method for efficiently calculating the gradients of a neural network's loss function with respect to its weights and biases. It uses the chain rule of calculus to compute gradients from the output back to the input.

The process of changing the parameters of a model to minimize the loss function. It typically involves using gradient-based optimization algorithms, like gradient descent.

The process of evaluating the model's performance on a separate dataset (not used for training). Used to assess the model's generalization ability.

The amount by which a value changes as another value changes. It measures the sensitivity of one value to the change in another value.

A representation of a computation, where nodes represent values and edges represent operations that connect those values.

The value that the model tries to predict based on the input data.

The value that the model predicts after processing the input data.

The process of updating the model's parameters to minimize the difference between the model's predictions and the true values.

The values that the model learns from the data to make predictions. They control how the model processes information.

A type of neural network layer that performs a linear transformation on its input, followed by an activation function. It implements the formula: output = activation(dot(W, input) + b), where W and b are the model parameters and activation is an element-wise function.

A function that is applied element-wise to the output of a dense layer, introducing non-linearity and allowing the network to learn more complex patterns. Examples include ReLU, sigmoid, and softmax.

The process of training a model on a dataset multiple times, where each pass through the entire dataset is considered an epoch. This helps in finding better model parameters.

The weighted sum of inputs to a neuron, which determines the neuron's activation value. It represents the linear contribution of each input.