Kolmogorov-Arnold Networks (KAN)
- one downside of KANs is that they take longer per parameter to train—in part because they can’t take advantage of GPUs
KAN - Comparison
Link to original
- In traditional multi-layer perceptrons [left]:
- each synapse learns a number called a weight
- each neuron applies a simple function to the sum of its inputs.
- In the new Kolmogorov-Arnold architecture [right]:
- each synapse learns a function
- each neuron sum the outputs of those functions
---cognitive-computing---machine-intelligence/ai---subfields/machine-learning-(ml)---pattern-recognition/ml---models/artificial-neural-networks-(ann)/ann---architecture-comparisons/feed-forward-networks-／-multi-layer-perceptrons-vs-kolmogorov-arnold-networks-(kan)/../../../../../../../../../computer/artificial-intelligence-(ai)---cognitive-computing---machine-intelligence/ai---subfields/machine-learning-(ml)---pattern-recognition/ml---models/artificial-neural-networks-(ann)/ann---architecture-comparisons/feed-forward-networks-／-multi-layer-perceptrons-vs-kolmogorov-arnold-networks-(kan)/kolmogorov-arnold-network.png)