4 ms·
Our lab works on changing this. I think it might still take some years for a full solution, but so far we are successful with NNs for small datasets with meta-l
by ersiees 4y ago
Our lab works on changing this. I think it might still take some years for a full solution, but so far we are successful with NNs for small datasets with meta-learning (https://arxiv.org/abs/2207.01848 https://arxiv.org/abs/2207.01848) and large datasets with regularisation (https://arxiv.org/abs/2106.11189 https://arxiv.org/abs/2106.11189). The first is very new, but the second is also cited in this paper, but they didn’t run it as a baseline.
- freemint 4y agoTree based algorithms have the advantage that global or robust optimization of them is possible. Do your approaches offer similar guarantees?
- ersiees 4y agoThe first work, I linked, arguably gets around this issue completely. There we train a neural network to approximate Bayesian prediction directly. It accepts the dataset as input. Thus, there is no optimisation procedure per dataset, just a forward pass. So far, this seems to be pretty stable.
- freemint 4y agoJust because you optimize network parameters to fit a surrogate model of the data doesn't mean you are not doing local optimization without guarantees on robustness or global optimality w.r.t. network parameters to fit the surrogate. Am i missing something?
- civilized 4y agoWhy not run your own method on their benchmark? If your method is general-purpose, it should be very easy. And it would be very impressive to see your method succeed on a task that was not chosen by you to write your paper.
- ersiees 4y agoYes, good point! The paper I am responsible for is targeted at smaller datasets, but I will propose this to the authors of the second paper :)