GHA update rule
From Piki
The Generalized Hebbian Algorithm (GHA) update rule (also known as Sanger's rule in the literature) is a Hebbian type update rule that performs PCA. It is similar to Oja's rule, but converges to ordered principal components.
| GHA | |
| Family | Hebbian |
| Direction | Forward |
| Deployable | Yes |
| Supervised | No |
Contents |
Usage
The Generalized Hebbian Algorithm tunes a Hebbian layer so that its weights form ordered principal components. The principal components are basis vectors that are aligned so that the greatest variance by any projection of the data comes to lie on the first component, the second greatest variance on the second component and so on.
Algorithm
where 
is the learning rate.
Settings
The settings can be modified using the settings browser.
| GHA update rule settings |
|---|
|
|
See also
- Update rule - General article on update rules.
- List of update rules - List of all update rules.
- Hebbian layer - A Synapse block that uses Hebbian updates.
- Hebbian learning - The Hebbian update rule that is the basis for the GHA update rule.
- The talented dr Hebb part 1 - Blog tutorial on using GHA for PCA.