Classes of feedforward neural networks and their circuit complexity
نویسندگان
چکیده
-Th& paper aims to p&ce neural networks in the conte.\t ol'booh'an citz'ldt complexit.l: 1,1~, de/itte aplm~priate classes qlfeedybrward neural networks with specified fan-in, accm'ac)' olcomputation and depth and ttsing techniques" o./commzmication comph:¥ity proceed to show t/tat the classes.fit into a well-studied hieralz'h)' q/boolean circuits. Results cover both classes of sigmoid activation./hnction networks and linear threshold networks. Tiffs provides a much needed theoretical basis./or the study o/the computational power qlilbed[brward neural networks. Keywords--Neural networks, Boolean circuits, Circuit complexity, Communication complexity, Fan-in, Threshold networks, Feedforward, Circuit depth.
منابع مشابه
Placing Feedforward Neural Networks Among Several Circuit Complexity Classes
This paper examines the circuit complexity of feedforward neural networks having sigmoid activation function. The starting point is the complexity class NN defined in [18]. First two additional complexity classes NN∆ k and NN∆,ε k having less restrictive conditions (than NN) concerning fan-in and accuracy are defined. We then prove several relations among these three classes and well establishe...
متن کاملCircuit Complexity and Feedforward Neural Networks
Circuit complexity, a subfield of computational complexity theory, can be used to analyze how the resource usage of neural networks scales with problem size. The computational complexity of discrete feedforward neural networks is surveyed, with a comparison of classical circuits to circuits constructed from gates that compute weighted majority functions.
متن کاملOn Properties of Networks of Neuron-Like Elements
The complexity and computational capacity of multi-layered, feedforward neural networks is examined. Neural networks for special purpose (structured) functions are examined from the perspective of circuit complexity. Known results in complexity theory are applied to the special instance of neural network circuits, and in particular, classes of functions that can be implemented in shallow circui...
متن کاملComplexity of Predictive Neural Networks
We announce some general bounds on the interactions of neural and information complexities of feedforward neural networks using general classes of activation functions. We show that, up to constant factors, neural and information complexities combine in a well-defined way in the determination of the complexity of a network.
متن کاملHardness Results for General Two-Layer Neural Networks
We deal with the problem of learning a general class of 2-layer neural networks in polynomial time. The considered neural networks consist of k linear threshold units on the hidden layer and an arbitrary binary output unit. We show NP-completeness of the consistency problem for classes that use an arbitrary set of binary output units containing a function which depends on all input dimensions. ...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
- Neural Networks
دوره 5 شماره
صفحات -
تاریخ انتشار 1992