The Concept of Forward-Forward Learning Applied to a Multi Output Perceptron

2023/04/06 by Klas-Göran Karlsson, Karlsson, K. Fredrik
Computer Science · Engineering · #Control Systems and Identification #FOS: Computer and information sciences #I.2.6 #Machine Learning (cs.LG) #Machine Learning and ELM #Neural Networks and Applications #Neural and Evolutionary Computing (cs.NE)

paper · pdf · doi:10.48550/arxiv.2304.03189

openalex publication_date 2023/04/06 · openalex created_date 2023/04/09 · openalex updated_date 2026/07/28

Abstract

The concept of a recently proposed Forward-Forward learning algorithm for fully connected artificial neural networks is applied to a single multi output perceptron for classification. The parameters of the system are trained with respect to increased (decreased) "goodness" for correctly (incorrectly) labelled input samples. Basic numerical tests demonstrate that the trained perceptron effectively deals with data sets that have non-linear decision boundaries. Moreover, the overall performance is comparable to more complex neural networks with hidden layers. The benefit of the approach presented here is that it only involves a single matrix multiplication.

Related

• Forward Thinking: Building and Training Neural Networks One Layer at a Time
• The Forward-Forward Algorithm: Some Preliminary Investigations
• The Integrated Forward-Forward Algorithm: Integrating Forward-Forward and Shallow Backpropagation With Local Losses
• Backward-Forward Algorithm: An Improvement towards Extreme Learning Machine
• One-class systems seamlessly fit in the forward-forward algorithm