Deep Learning

Deep learning (DL) is a type of machine learning that uses neural networks with many layers to learn and extract useful patterns from large quantities of data. The defining feature of deep learning models is their ability to learn hierarchical representations that allow them to learn abstract concepts and develop sophisticated models of the world.

‍

MLPs, as a type of neural network, consist of layers of neurons that take in a weighted sum of their inputs, apply an activation function to transform it, and generate an output signal that is passed along to other neurons. In these models, the input layer is the first layer in the network. Data enters the network through this layer and is passed along to the next layer through its nodes. This signal will flow through the network until it reaches the final (output) layer, which contains the results of the computations performed within the intermediate (hidden) layers between input and output network. Each hidden layer is a collection of neurons that receive outputs from the previous layer, perform a computation, and pass the results to the next layer. Weights are numerical values that represent the strength of the connections between two neurons. They determine how much the input signal from a given neuron will influence the output of the next neuron. In addition to weights, every neuron has an additional parameter called a bias that shifts the activation function along the input axis. Activation functions are nonlinear functions applied to the weighted input sum of each neuron within a layer of a neural network.

‍

Loss is a measurement of model error. It is a value calculated by a loss function that compares target and predicted values to quantify model fit. In deep learning, loss is essential to the training process, as it aims to minimize the loss function.

‍