Neural Networks: How do they work?

Neural networks are versatile models that can learn just about any complex pattern. These powerful models are the core of deep learning consisting of multi-layer perceptrons, convolutional networks, sequence models and many more. In this brief project, I will explore the CIFAR-10 dataset and implement a simple neural network (multi-layer perceptron).

The concept of a neural network is actually quite simple. Similar to how neurons fire or activate in the human brain, the neurons within a layer in a neural network are activated through an activation function. This process returns output that will be passed on to the next layer of the neural network and the cycle is repeated until the end of the neural network. This process is known as the forward pass where your data is fed forward through the network after applying weights and an activation function. Depending on whether you are solving a regression or classification problem, the final output layer of your neural network will output one node for regression problems and multiple nodes for classification problems. In this project, we will be classifying images so the output from the neural network will have one node per class which will go through the softmax function to obtain the final prediction.

#data-science #stochastic-gradient #neural-networks #pytorch #python

10.95 GEEK