arXiv subjects

Laxmidhar Behera

Publications and source records attributed to Laxmidhar Behera.

42 records · Page 3Linked to original sources

arXiv2015

Optimal Convergence Rate in Feed Forward Neural Networks using HJB Equation

A control theoretic approach is presented in this paper for both batch and instantaneous updates of weights in feed-forward neural networks. The popular Hamilton-Jacobi-Bellman (HJB) equation has been used to generate an optimal weight update law. The remarkable contribution in this paper is that closed form solutions for both optimal cost and weight update can be achieved for any feed-forward network using HJB equation in a simple yet elegant manner. The proposed approach has been compared with some of the existing best performing learning algorithms. It is found as expected that the proposed approach is faster in convergence in terms of computational time. Some of the benchmark test data such as 8-bit parity, breast cancer and credit approval, as well as 2D Gabor function have been used to validate our claims. The paper also discusses issues related to global optimization. The limitations of popular deterministic weight update laws are critiqued and the possibility of global optimization using HJB formulation is discussed. It is hoped that the proposed algorithm will bring in a lot of interest in researchers working in developing fast learning algorithms and global optimization.

cs.NE↗

arXiv2012

Optimal migration promotes the outbreak of cooperation in heterogeneous populations

We consider a population of agents that are heterogeneous with respect to (i) their strategy when interacting $n_{g}$ times with other agents in an iterated prisoners dilemma game, (ii) their spatial location on $K$ different islands. After each generation, agents adopt strategies proportional to their average payoff received. Assuming a mix of two cooperating and two defecting strategies, we first investigate for isolated islands the conditions for an exclusive domination of each of these strategies and their possible coexistence. This allows to define a threshold frequency for cooperation that, dependent on $n_{g}$ and the initial mix of strategies, describes the outbreak of cooperation in the absense of migration. We then allow migration of a fixed fraction of the population after each generation. Assuming a worst case scenario where all islands are occupied by defecting strategies, whereas only one island is occupied by cooperators at the threshold frequency, we determine the optimal migration rate that allows the outbreak of cooperation on \emph{all} islands. We further find that the threshold frequency divided by the number of islands, i.e. the relative effort for invading defecting islands with cooperators decreeses with the number of islands. We also show that there is only a small bandwidth of migration rates, to allow the outbreak of cooperation. Larger migration rates destroy cooperation.

physics.soc-ph↗

arXiv2008

Nonlinear Voter Models: The Transition from Invasion to Coexistence

In nonlinear voter models the transitions between two states depend in a nonlinear manner on the frequencies of these states in the neighborhood. We investigate the role of these nonlinearities on the global outcome of the dynamics for a homogeneous network where each node is connected to $m=4$ neighbors. The paper unfolds in two directions. We first develop a general stochastic framework for frequency dependent processes from which we derive the macroscopic dynamics for key variables, such as global frequencies and correlations. Explicite expressions for both the mean-field limit and the pair approximation are obtained. We then apply these equations to determine a phase diagram in the parameter space that distinguishes between different dynamic regimes. The pair approximation allows us to identify three regimes for nonlinear voter models: (i) complete invasion, (ii) random coexistence, and -- most interestingly -- (iii) correlated coexistence. These findings are contrasted with predictions from the mean-field phase diagram and are confirmed by extensive computer simulations of the microscopic dynamics.

cond-mat.stat-mech↗

arXiv2004

Quantum Brain: A Recurrent Quantum Neural Network Model to Describe Eye Tracking of Moving Targets

A theoretical quantum brain model is proposed using a nonlinear Schroedinger wave equation. The model proposes that there exists a quantum process that mediates the collective response of a neural lattice (classical brain). The model is used to explain eye movements when tracking moving targets. Using a Recurrent Quantum Neural Network(RQNN) while simulating the quantum brain model, two very interesting phenomena are observed. First, as eye sensor data is processed in a classical brain, a wave packet is triggered in the quantum brain. This wave packet moves like a particle. Second, when the eye tracks a fixed target, this wave packet moves not in a continuous but rather in a discrete mode. This result reminds one of the saccadic movements of the eye consisting of 'jumps' and 'rests'. However, such a saccadic movement is intertwined with smooth pursuit movements when the eye has to track a dynamic trajectory. In a sense, this is the first theoretical model explaining the experimental observation reported concerning eye movements in a static scene situation. The resulting prediction is found to be very precise and efficient in comparison to classical objective modeling schemes such as the Kalman filter.

q-bio.NC↗

arXiv2003

On Spatial Consensus Formation: Is the Sznajd Model Different from a Voter Model?

In this paper, we investigate the so-called ``Sznajd Model'' (SM) in one dimension, which is a simple cellular automata approach to consensus formation among two opposite opinions (described by spin up or down). To elucidate the SM dynamics, we first provide results of computer simulations for the spatio-temporal evolution of the opinion distribution $L(t)$, the evolution of magnetization $m(t)$, the distribution of decision times $P(τ)$ and relaxation times $P(μ)$. In the main part of the paper, it is shown that the SM can be completely reformulated in terms of a linear VM, where the transition rates towards a given opinion are directly proportional to frequency of the respective opinion of the second-nearest neighbors (no matter what the nearest neighbors are). So, the SM dynamics can be reduced to one rule, ``Just follow your second-nearest neighbor''. The equivalence is demonstrated by extensive computer simulations that show the same behavior between SM and VM in terms of $L(t)$, $m(t)$, $P(τ)$, $P(μ)$, and the final attractor statistics. The reformulation of the SM in terms of a VM involves a new parameter $σ$, to bias between anti- and ferromagnetic decisions in the case of frustration. We show that $σ$ plays a crucial role in explaining the phase transition observed in SM. We further explore the role of synchronous versus asynchronous update rules on the intermediate dynamics and the final attractors. Compared to the original SM, we find three additional attractors, two of them related to an asymmetric coexistence between the opposite opinions.

cond-mat.stat-mech↗

arXiv2002

Evolution of Cooperation in a Spatial Prisoner's Dilemma

We investigate the spatial distribution and the global frequency of agents who can either cooperate or defect. The agent interaction is described by a deterministic, non-iterated prisoner's dilemma game, further each agent only locally interacts with his neighbors. Based on a detailed analysis of the local payoff structures we derive critical conditions for the invasion or the spatial coexistence of cooperators and defectors. These results are concluded in a phase diagram that allows to identify five regimes, each characterized by a distinct spatiotemporal dynamics and a corresponding final spatial structure. In addition to the complete invasion of defectors, we find coexistence regimes with either a majority of cooperators in large spatial domains, or a minority of cooperators organized in small non-stationary domains or in small clusters. The analysis further allowed a verification of computer simulation results by Nowak and May (1993). Eventually, we present simulation results of a true 5-person game on a lattice. This modification leads to non-uniform spatial interactions that may even enhance the effect of cooperation. Keywords: Prisoner's dilemma; cooperation; spatial 5-person game

cond-mat.stat-mech↗