michael_nielsen

New working notes: "Reduced MNIST: how well can machines learn from small data?" http://cognitivemedium.com/rmnist 

I think there are some benchmarks of human performances on 'alien' alphabets somewhere in the zero-shot transfer learning (eg Omniglot) and Bayesian program synthesis literature.

Any keywords that might help me find them? Tried "human learning alien alphabets bayesian program [/ zero-shot]" and a couple of variants in Scholar, got nothing promising.

The only one that comes to mind was a Bayesian program synthesis paper for digit recognition in, I think, Nature sometime n past few years ago claiming human-level performance (based on a comparison with human performance, obvs).

see e.g. https://arxiv.org/pdf/1605.06065.pdf … and references to/from it

Thanks!