CarlosGG's Knowledge Garden 🪴

Search

Search IconIcon to open search

KANs

Last updated Nov 14, 2024 Edit Source

Type of neural network inspired by the Kolmogorov-Arnold representation theorem. Unlike traditional multi-layer perceptrons (MLPs), KANs use learnable activation functions located on the network’s edges (weights) rather than on the nodes (neurons). This approach aims to improve the flexibility and performance of neural networks

# Resources

# Code

# References