27.4 Experience and new algorithms

Newer algorithms there are listed below

Newer algorithms are:

  • Evolutionary Approaches
  • Transformer Networks (BERT, gpt-2, etc)
  • Generative Adversarial Networks (GAN)

where evolutionary approaches have been around for quite some time but the usage of them in machine learning is rather recent.



Splitting the graphs up for each category of experience and plotting the percentage of usage for the given experience level gives an insight into how the usage of new algorithms differs over levels of experience



From the above graph it can be deducted that:

Findings:

  • Very experienced use new algorithms less often
  • Newbies embrace them
  • Evolutionary approaches are popular for medium experienced