Alternative learning mechanisms

tags
Machine learning

Many people, including Geoffrey Hinton, have raised concerns about the back-propagation algorithm and the fact that it’s likely not a promising way to achieve Artificial Intelligence (see this Axios blog post).

Alternative mechanisms for learning have been and are currently studied to try and approach the learning problem in a more effective way.

Direct feedback alignment

(Nøkland 2016)

Hebbian learning

The theory is sometimes summarized as “Cells that fire together wire together.”.

Predictive coding

(Whittington and Bogacz 2017; Millidge, Tschantz, and Buckley 2020)

Bibliography

  1. . . "Predictive Coding Approximates Backprop Along Arbitrary Computation Graphs". arXiv:2006.04182 [Cs], October.

  2. . . “An Approximation of the Error Backpropagation Algorithm in a Predictive Coding Network with Local Hebbian Synaptic Plasticity”. Neural Computation 29 (5):1229–62.

  3. . . “Direct Feedback Alignment Provides Learning in Deep Neural Networks”. arXiv:1609.01596 [Cs, Stat], December.

← Back to Notes