3 ms·
Interesting, but how does it perform on standard benchmarks like image net and MNIST?
by jxcole 7y ago
Interesting, but how does it perform on standard benchmarks like image net and MNIST?
- p1esk 7y agoThey have some cifar10 results in the paper, but only very small networks.
- JulianWasTaken 7y agoI am way out of my depth so maybe this is pure nonsense, but presumably the goals aren't so much to evolve better performing models directly for some dataset specifically, but to see what kinds of model families it evolves and whether we've thought of all of them? Once we've got one we can then presumably train specific models in a more targeted way.