Neural Networks Basics
Artificial neural networks are a group of machine learning models, which are based on an ensemble of artificial neurons. Those neurons, as well as the layer wise ordering, are inspired by real neurons and animal brains. Each of the artificial neurons inside the neural network has a set of parameters which are to be learned during the training. The most common way to learn the parameters is to apply a layer wise form of gradient descend called backpropagation. Unlike most other machine learning models, neural networks are very flexible and can be used for several tasks on different data types e.g. image classification, time series encoding or object detection in point clouds. Nowadays most state of the art models in machine learning are based on neural networks.
Related
- [slides] Convolutional Neural Networks
- [theoretical-background] Optimization Using (Stochastic) Gradient Decent
- [slides] Neural Networks - Basics
- [theoretical-background] Optimization Methods