arXiv subjects

I. Kanter

Publications and source records attributed to I. Kanter.

26 records · Page 2Linked to original sources

arXiv2000

Dynamics of Interacting Neural Networks

The dynamics of interacting perceptrons is solved analytically. For a directed flow of information the system runs into a state which has a higher symmetry than the topology of the model. A symmetry breaking phase transition is found with increasing learning rate. In addition it is shown that a system of interacting perceptrons which is trained on the history of its minority decisions develops a good strategy for the problem of adaptive competition known as the Bar Problem or Minority Game.

cond-mat.dis-nn↗

arXiv1998

Long-term properties of time series generated by a perceptron with various transfer functions

We study the effect of various transfer functions on the properties of a time series generated by a continuous-valued feed-forward network in which the next input vector is determined from past output values. The parameter space for monotonic and non-monotonic transfer functions is analyzed in the unstable regions with the following main finding; non-monotonic functions can produce robust chaos whereas monotonic functions generate fragile chaos only. In the case of non-monotonic functions, the number of positive Lyapunov exponents increases as a function of one of the free parameters in the model, hence, high dimensional chaotic attractors can be generated. We extend the analysis to a combination of monotonic and non-monotonic functions.

cond-mat.dis-nn↗

arXiv1998

Statistical Properties of Contact Maps

A contact map is a simple representation of the structure of proteins and other chain-like macromolecules. This representation is quite amenable to numerical studies of folding. We show that the number of contact maps corresponding to the possible configurations of a polypeptide chain of N amino acids, represented by (N-1)-step self avoiding walks on a lattice, grows exponentially with N for all dimensions D>1. We carry out exact enumerations in D=2 on the square and triangular lattices for walks of up to 20 steps and investigate various statistical properties of contact maps corresponding to such walks. We also study the exact statistics of contact maps generated by walks on a ladder.

cond-mat.stat-mech↗

arXiv1996

Storage capacity of correlated perceptrons

We consider an ensemble of $K$ single-layer perceptrons exposed to random inputs and investigate the conditions under which the couplings of these perceptrons can be chosen such that prescribed correlations between the outputs occur. A general formalism is introduced using a multi-perceptron costfunction that allows to determine the maximal number of random inputs as a function of the desired values of the correlations. Replica-symmetric results for $K=2$ and $K=3$ are compared with properties of two-layer networks of tree-structure and fixed Boolean function between hidden units and output. The results show which correlations in the hidden layer of multi-layer neural networks are crucial for the value of the storage capacity.

cond-mat.dis-nn↗

arXiv1996

Training a perceptron by a bit sequence: Storage capacity

A perceptron is trained by a random bit sequence. In comparison to the corresponding classification problem, the storage capacity decreases to alpha_c=1.70\pm 0.02 due to correlations between input and output bits. The numerical results are supported by a signal to noise analysis of Hebbian weights.

cond-mat↗

arXiv1996

Cluster Dynamics for Randomly Frustrated Systems with Finite Connectivity

In simulations of some infinite range spin glass systems with finite connectivity, it is found that for any resonable computational time, the saturatedenergy per spin that is achieved by a cluster algorithm is lowered in comparison to that achieved by Metropolis dynamics.The gap between the average energies obtained from these two dynamics is robust with respect to variations of the annealing schedule. For some probability distribution of the interactions the ground state energy is calculated analytically within the replica symmetry assumptionand is found to be saturated by a cluster algorithm.

cond-mat↗

arXiv1996

Mean Field Behavior of Cluster Dynamics

The dynamic behavior of cluster algorithms is analyzed in the classical mean field limit. Rigorous analytical results below $T_c$ establish that the dynamic exponent has the value $z_{sw}=1$ for the Swendsen-Wang algorithm and $z_{uw}=0$ for the Wolff algorithm. An efficient Monte Carlo implementation is introduced, adapted for using these algorithms for fully connected graphs. Extensive simulations both above and below $T_c$ demonstrate scaling and evaluate the finite-size scaling function by means of a rather impressive collapse of the data.

cond-mat↗

arXiv1995

THE BIT GENERATOR AND TIME SERIES PREDICTION

Generation and prediction of time series is analyzed for the case of a Bit-Generator: a perceptron where in each time step the input units are shifted one bit to the right with the state of the leftmost input unit set equal to the output unit in the previous time step. The long-time dynamical behavior of the Bit-Generator consists of cycles whose typical period scale polynomially with the size of the network and whose spatial structure is periodic with a typical finite wave length. The generalization erroron a cycle is zero for a finite training set and global dynamical behaviors such as the cycle period can also be learned in a finite time. Hence, a projection of a rule can be learned in a finite time.

cond-mat↗