Neural Network
Layered composition of weighted sums and nonlinearities, trained by gradient descent.
Overview
A neural network is a stack of layers, each applying a weighted sum and a nonlinearity to the previous layer's output, with the weights adjusted by gradient descent. It solves the problem that linear models cannot represent interactions between inputs. It exists because a composition of simple nonlinear steps can approximate very complex functions, and because backpropagation makes training such a composition tractable.
Technology Graph
How Neural Network connects to the rest of the stack.
Uses and depends on
- CNN architecture Algorithms CNN based on Neural Network High ✓ source
- Diffusion technique Techniques Diffusion based on Neural Network High ✓ source
- GAN architecture Algorithms GAN based on Neural Network High ✓ source
- MoE architecture Algorithms MoE based on Neural Network High ✓ source
- RNN architecture Algorithms RNN based on Neural Network High ✓ source
Used by
- Recommendation application Applications Neural Network used by Recommendation High ✓ source
Applied to
- Backpropagation technique Algorithms Backpropagation used for Neural Network High ✓ source