On sample complexity of neural networks


الملخص بالإنكليزية

We consider functions defined by deep neural networks as definable objects in an o-miminal expansion of the real field, and derive an almost linear (in the number of weights) bound on sample complexity of such networks.

تحميل البحث