3 ms·
I think neural nets are approved that with more than 2 hidden layers, it can approximate any functuons. Is there any literature for other algorithms as well?
by eva1984 10y ago
I think neural nets are approved that with more than 2 hidden layers, it can approximate any functuons. Is there any literature for other algorithms as well?
- yzmtf2008 10y agoIt is true that neural networks can approximate any function. But what are feasible ways to construct such networks? Part of the point of having "deep" networks is that it's easier to train that way. Scalability does not only refer to the scale of the model itself, but also the difficulty in building such model.