Bounds on the degree of high order binary perceptrons
نویسنده
چکیده
High order perceptrons are often used in order to reduce the size of neural networks The complexity of the architecture of a usual multilayer network is then turned into the complexity of the functions performed by each high order unit and in particular by the degree of their polynomials The main result of this paper provides a bound on the degree of the polynomial of a high order perceptron when the binary training data result from the encoding of an arrangement of hyperplanes in the Euclidian space Such a situation occurs naturally in the case of a feedforward network with a single hidden layer of rst order perceptrons and an output layer of high order perceptrons In this case the result says that the degree of the high order perceptrons can be bounded by the minimum between the number of inputs and the number of hidden units Acknowledgements This work was initiated while I was a postdoctoral fellow supported by RUTCOR and DIMACS Center for Discrete Mathematics and Theoretical Computer Science I am particularly thankful to my colleague Dr Motakuri Ramana for helpful discussions on this matter
منابع مشابه
Bounds on the Degree of High Order Binary Perceptrons Bounds on the Degree of High Order Binary Perceptrons
High order perceptrons are often used in order to reduce the size of neural networks. The complexity of the architecture of a usual multilayer network is then turned into the complexity of the functions performed by each high order unit and in particular by the degree of their polynomials. The main result of this paper provides a bound on the degree of the polynomial of a high order perceptron,...
متن کاملNew bounds on proximity and remoteness in graphs
The average distance of a vertex $v$ of a connected graph $G$is the arithmetic mean of the distances from $v$ to allother vertices of $G$. The proximity $pi(G)$ and the remoteness $rho(G)$of $G$ are defined as the minimum and maximum averagedistance of the vertices of $G$. In this paper we investigate the difference between proximity or remoteness and the classical distanceparameters diameter a...
متن کاملON THE CHARACTERISTIC DEGREE OF FINITE GROUPS
In this article we introduce and study the concept of characteristic degree of a subgroup in a finite group. We define the characteristic degree of a subgroup H in a finite group G as the ratio of the number of all pairs (h,α) ∈ H×Aut(G) such that h^α∈H, by the order of H × Aut(G), where Aut(G) is the automorphisms group of G. This quantity measures the probability that H can be characteristic ...
متن کاملOn the eigenvalues of some matrices based on vertex degree
The aim of this paper is to compute some bounds of forgotten index and then we present spectral properties of this index. In continuing, we define a new version of energy namely ISI energy corresponded to the ISI index and then we determine some bounds for it.
متن کاملOn the first variable Zagreb index
The first variable Zagreb index of graph $G$ is defined as begin{eqnarray*} M_{1,lambda}(G)=sum_{vin V(G)}d(v)^{2lambda}, end{eqnarray*} where $lambda$ is a real number and $d(v)$ is the degree of vertex $v$. In this paper, some upper and lower bounds for the distribution function and expected value of this index in random increasing trees (rec...
متن کامل