# extreme learning machine

> type of artificial neural network

**Wikidata**: [Q21948731](https://www.wikidata.org/wiki/Q21948731)  
**Wikipedia**: [English](https://en.wikipedia.org/wiki/Extreme_learning_machine)  
**Source**: https://4ort.xyz/entity/extreme-learning-machine

## Summary
An extreme learning machine is a type of artificial neural network characterized by a single-layer feedforward architecture with random hidden layer weights and analytically determined output weights.

## Key Facts
- Extreme learning machines are a subclass of feedforward neural networks
- They are also classified as a type of artificial neural network
- The architecture uses random assignment of input weights and hidden layer biases
- Output weights are determined analytically rather than through iterative training
- This approach enables extremely fast learning compared to traditional backpropagation methods

## FAQs

### Q: What makes extreme learning machines different from traditional neural networks?
A: Extreme learning machines use randomly assigned input weights and hidden layer biases, with output weights calculated analytically rather than through iterative training, enabling much faster learning speeds.

### Q: How does the learning process work in extreme learning machines?
A: The learning process involves randomly initializing the input weights and hidden layer biases, then analytically computing the output weights using least-squares methods, eliminating the need for iterative weight updates.

### Q: What are the main advantages of extreme learning machines?
A: The main advantages include extremely fast learning speed, minimal human intervention required, good generalization performance, and the ability to handle various types of data including real, binary, and categorical.

## Why It Matters
Extreme learning machines represent a significant advancement in neural network training methodology by dramatically reducing computational complexity while maintaining competitive performance. The approach addresses one of the fundamental challenges in neural network training - the time-consuming iterative optimization process - by replacing it with analytical solutions. This makes ELM particularly valuable for applications requiring rapid model deployment or real-time learning capabilities. The technique has found applications across various domains including classification, regression, clustering, and feature learning, demonstrating that random feature mapping combined with linear analysis can achieve results comparable to traditional networks while requiring far less computational resources and training time.

## Notable For
- Extremely fast learning speed compared to traditional backpropagation-based networks
- Analytical determination of output weights rather than iterative optimization
- Minimal human intervention required during training process
- Ability to handle multiple types of data including real, binary, and categorical
- Competitive generalization performance despite random hidden layer initialization

## Body
Extreme learning machines employ a single-layer feedforward network architecture where input weights and hidden layer biases are randomly assigned from continuous probability distributions. The output weights are then determined by solving a linear system of equations, typically using Moore-Penrose generalized inverse. This analytical approach eliminates the need for gradient-based optimization methods like backpropagation. The random feature mapping in the hidden layer transforms the original input space into a higher-dimensional feature space where linear separation becomes possible. The method can be extended to various activation functions including sigmoid, sine, and radial basis functions. Research has demonstrated that ELMs can achieve similar or better generalization performance compared to traditional networks while requiring significantly less training time. The technique has been successfully applied to both regression and classification problems across diverse domains.

## Schema Markup
```json
{
  "@context": "https://schema.org",
  "@type": "Thing",
  "name": "extreme learning machine",
  "description": "A type of artificial neural network characterized by random hidden layer weights and analytically determined output weights",
  "additionalType": "feedforward neural network"
}

## References

1. [OpenAlex](https://docs.openalex.org/download-snapshot/snapshot-data-format)