# Martijho-PathNet-search-plots

(Forskjeller mellom versjoner)
Gå til: navigasjon, søk
 Versjonen fra 28. feb 2018 kl. 15:58 (vis kilde) (→Diversity metrics)← Eldre redigering Versjonen fra 28. feb 2018 kl. 15:58 (vis kilde) (→Population Diversity)Nyere redigering → Linje 12: Linje 12: - $a = \frac{1}{2\pi^{\frac{n}{2}}*|\Sigma|^{\frac{1}{2}}}*exp\bigg((-0.5)*\frac{(I(n,m) - \mu)'}{\Sigma}*(I(n,m) - \mu)\bigg)$ + $\epsilon = \frac{1}{2\pi^{\frac{n}{2}}*|\Sigma|^{\frac{1}{2}}}*exp\bigg((-0.5)*\frac{(I(n,m) - \mu)'}{\Sigma}*(I(n,m) - \mu)\bigg)$ === Home-made for each layer === === Home-made for each layer ===

## Population Diversity

### Diversity metrics

Three different diversity calculations have been performed, which resulted in three similar diversity messures:  Pairwise Hamming distance calculated for each layer, and then averaged across each path. This plot took 22 minutes and 40 seconds to generate.  This diversity metric resulted in a scaled version of Pairwise Hamming, but took only 36 seconds to generate. See description below  Calculated by averaging the euclidean distance between every genotype in a generation, and a "centroid genotype". The plot was generated in 40 seconds
Pair-wise Hamming distance

Pair-wise hamming distance is calculated by counting the number of similar genes between every pair of genomes in a population. $\epsilon = \frac{1}{2\pi^{\frac{n}{2}}*|\Sigma|^{\frac{1}{2}}}*exp\bigg((-0.5)*\frac{(I(n,m) - \mu)'}{\Sigma}*(I(n,m) - \mu)\bigg)$

### Home-made for each layer  Average population diversity for each layer in generation. Experiment: Recomb  Average population diversity for each layer in generation. Experiment: Low  Average population diversity for each layer in generation. Experiment: High  Average population diversity for each layer in generation. Experiment: Low to High  Average population diversity for each layer in generation. Experiment: High to Low

## Path size  The change in average path size for each generation.  The average size of each layer in each optimal path found.

## Capacity and Reuse  Average number of modules used during the different searches plotted consecutively. The gray dotted line is what number of modules would be used if modules were selected randomly for each task.  Used capacity in a PathNet plotted against that networks average classification accuracy for all tasks  Average reuse for each generation plotted for each experiment type. The gray dotted line is what reuse would be reached by random module selection

## Training  Average training accuracy for each of the five search algorithms during the search for each task. The dotted line is the validation accuracy reached after the search was completed.  Number of training iterations in total for all modules in the PathNet plotted against the cumulative validation accuracy reached for all tasks. The size of the circle corresponds to the amount of modules used.  Plot of the ratio of unit training in locked and stored modules on the total number of unit training in all modules for each task