A non-linear learning & classification algorithm that achieves full training accuracy with stellar classification accuracy

TL;DR

This paper introduces the Reverse Ripple Effect (R.R.E), a non-linear, non-iterative learning algorithm that achieves perfect training accuracy and stellar classification performance, validated through simulations and comparisons with existing methods.

Contribution

The paper presents the R.R.E algorithm, a novel deterministic approach that superimposes Gaussian functions for classification, achieving full training accuracy and outperforming traditional methods.

Findings

R.R.E achieves 100% training accuracy.

R.R.E performs well on linearly and non-linearly separable data.

R.R.E outperforms traditional algorithms like SVM and neural networks in certain scenarios.

Abstract

A fast Non-linear and non-iterative learning and classification algorithm is synthesized and validated. This algorithm named the "Reverse Ripple Effect(R.R.E)", achieves 100% learning accuracy but is computationally expensive upon classification. The R.R.E is a (deterministic) algorithm that super imposes Gaussian weighted functions on training points. In this work, the R.R.E algorithm is compared against known learning and classification techniques/algorithms such as: the Perceptron Criterion algorithm, Linear Support Vector machines, the Linear Fisher Discriminant and a simple Neural Network. The classification accuracy of the R.R.E algorithm is evaluated using simulations conducted in MATLAB. The R.R.E algorithm's behaviour is analyzed under linearly and non-linearly separable data sets. For the comparison with the Neural Network, the classical XOR problem is considered.