Table 1 Comparisons against existing solutions

From: Optical neural network via loose neuron array and functional learning

(a)

    
 

Forward Model

Genetic Algorithm

Finite Difference

Functional Learning

Gradient

explicit

none

measured

implicit

1-layer MNIST

23.50%

14.06%

8.594%

90.78%

(b)

    
 

Data

Network

Mechanism

Applications

Supervised Learning

labeled

explicit

target learning

regression and prediction

Unsupervised Learning

unlabeled

explicit

self-structuring

data structure learning

Reinforcement Learning

decision process

explicit

trial-and-error

decision making

Functional Learning

z: functional response

z: explicit

z: induction

hardware, chip,

 

p: labeled

p: implicit

p: deduction

and system control

  1. (a) Classification accuracy of training the 1-layer LFNN in the MNIST dataset using different training paradigms. The 1-layer LFNN consists of 2 LC panels. Genetic algorithm and finite difference are measured using approximately equal training time of functional learning. The forward model is measured using an equal number of epochs of functional learning. Detailed discussion is in the supplementary document.
  2. (b) Different intuitions of existing learning paradigms.
Back to article page