Publication View

Activation Functions z (2008)

Abstract
We deal with computational issues of loading a xed-architecture neural network with a set of positive and negative examples. This is the rst result on the hardness of loading a simple 3-node architecture which do not consist of the binary-threshold neurons, but rather utilize a particular continuous activation function, commonly used in the neural network literature. We observe that the loading problem is polynomial-time if the input dimension is constant. Otherwise, however, any possible learning algorithm based on particular xed architectures faces severe computational barriers. Similar theorems have already been proved by Megiddo and by Blum and Rivest, to the case of binary-threshold networks only. Our theoretical results lend further suggestion to the use of incremental (architecture-changing) techniques for training networks rather than xed architectures. Furthermore, they imply hardness of learnability intheprobably-approximately-correct sense as well. 1

Publication details
Download http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.74.3347
Source http://crab.rutgers.edu/~bhaskar/resume/publ/papers/tr2.ienn.rev.pdf
Contributors CiteSeerX
Repository CiteSeerX - Scientific Literature Digital Library and Search Engine (United States)
Type text
Language English
Relation 10.1.1.42.6662, 10.1.1.79.5078, 10.1.1.66.1613, 10.1.1.31.814, 10.1.1.38.6275, 10.1.1.31.2967, 10.1.1.52.263, 10.1.1.12.9839