Convolutional Neural Networks (CNN)

Status: public · Confidence: medium (0.76) · Basis: verified_sources

## TL;DR
Convolutional neural networks are neural networks designed for grid-like data such as images. They use local filters, shared weights, nonlinear layers, and often pooling or striding to learn visual features across spatial positions.

## Core Explanation
Early layers often detect local patterns, while deeper layers combine them into more task-specific features. CNN history includes document-recognition systems, ImageNet-scale classifiers, biomedical segmentation networks, and residual architectures that made very deep networks easier to train.

## Detailed Analysis
CNNs are still widely used because convolution builds in useful assumptions about locality and translation. Modern systems may combine CNNs with attention, transformers, pretraining, or task-specific heads, so claims should specify the architecture and task rather than treating all visual AI as one model family.

## Further Reading
- LeCun et al. on document recognition
- AlexNet
- ResNet

## Related Articles

- [Activation Functions in Neural Networks](../activation-functions.md)
- [AI for Fraud Detection: Graph Neural Networks, Anti-Money Laundering, and Financial Crime](../ai-for-fraud-detection.md)
- [Graph Neural Networks: Message Passing, Applications, and Frontiers](../graph-neural-networks-message-passing-applications-and-frontiers.md)