A Novel Multivalued Associative Memory System

Globally coupled map (GCM) model can evolve through chaotic searching into several stable periodic orbits under properly controlled parameters. This can be exploited in information processing such as associative memory and optimization. In this paper, we propose a novel covariance learning rule for multivalue patterns and apply it in memorization of gray scale images based on modified GCM model (S GCM). Analysis of retrieval results are given finally.

DISCRETE BIDIRECTIONAL ASSOCIATIVE MEMORY WITH LEARNING FUNCTION

In this paper we propose a new discrete bidirectional associative memory (DBAM) which is derived from our previous continuous linear bidirectional associative memory (LBAM). The DBAM performs bidirectionally the optimal associative mapping proposed by Kohonen. Like LBAM and NBAM proposed by one of the present authors,the present BAM ensures the guaranteed recall of all stored patterns,and possesses far higher capacity compared with other existing BAMs,and like NBAM, has the strong ability to suppress the noise occurring in the output patterns and therefore reduce largely the spurious patterns. The derivation of DBAM is given and the stability of DBAM is proved. We also derive a learning algorithm for DBAM,which has iterative form and make the network learn new patterns easily. Compared with NBAM the present BAM can be easily implemented by software.

BIDIRECTIONAL ASSOCIATIVE MEMORY ENSEMBLE

The multiple classifier system （MCS）, composed of multiple diverse classifiers or feed-forward neural networks, can significantly improve the classification or generalization ability of a single classifier. Enlightened by the fundamental idea of MCS, the ensemble is introduced into the quick learning for bidirectional associative memory （QLBAM） to construct a BAM ensemble, for improving the storage capacity and the error-correction capability without destroying the simple structure of the component BAM. Simulations show that, with an appropriate ＂overproduce and choose＂ strategy or ＂thinning＂ algorithm, the proposed BAM ensemble significantly outperforms the single QLBAM in both storage capacity and noise-tolerance capability.

Multi-Valued Associative Memory Neural Network

A novel learning method for multi-valued associative memory network is introduced, which is based on Hebb rule, but utilizes more information. According to the current probe vector, the connection weights matrix could be chosen dynamically. Double-valued and multi-valued associative memory are all realized in our simulation experiment. The experimental results show that the method could enhance the associative success rate.

Properties and Stability of Max-Product Fuzzy Bi-Directional Associative Memory

In this paper, a fuzzy operator of max-product is defined at first, and the fuzzy bi-directional associative memory (FBAM) based on the fuzzy operator of max-product is given. Then the properties and the Lyapunov stability of equilibriums of the networks are studied.

A Hopfield-like hippocampal CA3 neural network model for studying associative memory in Alzheimer's disease

Associative memory, one of the major cognitive functions in the hippocampal CA3 region, includes auto-associative memory and hetero-associative memory. Many previous studies have shown that Alzheimer＇s disease （AD） can lead to loss of functional synapses in the central nervous system, and associative memory functions in patients with AD are often impaired, but few studies have addressed the effect of AD on hetero-associative memory in the hippocampal CA3 region. In this study, based on a simplified anatomical structure and synaptic connections in the hippocampal CA3 region, a three-layered Hopfield-like neural network model of hippocampal CA3 was proposed and then used to simulate associative memory functions in three circumstances： normal, synaptic deletion and synaptic compensation, according to Ruppin＇s synaptic deletion and compensation theory. The influences of AD on hetero-associative memory were further analyzed. The simulated results showed that the established three-layered Hopfield-like neural network model of hippocampal CA3 has both auto-associative and hetero-associative memory functions. With increasing synaptic deletion level, both associative memory functions were gradually impaired and the mean firing rates of the neurons within the network model were decreased. With gradual increasing synaptic compensation, the associative memory functions of the network were improved and the mean firing rates were increased. The simulated results suggest that the Hopfield-like neural network model can effectively simulate both associative memory functions of the hippocampal CA3 region. Synaptic deletion affects both auto-associative and hetero-associative memory functions in the hippocampal CA3 region, and can also result in memory dysfunction. To some extent, synaptic compensation measures can offset two kinds of associative memory dysfunction caused by synaptic deletion in the hippocampal CA3 area.

GLOBAL DYNAMICS OF DELAYED BIDIRECTIONAL ASSOCIATIVE MEMORY (BAM) NEURAL NETWORKS

Without assuming the smoothness,monotonicity and boundedness of the activation functions, some novel criteria on the existence and global exponential stability of equilibrium point for delayed bidirectional associative memory (BAM) neural networks are established by applying the Liapunov functional methods and matrix_algebraic techniques. It is shown that the new conditions presented in terms of a nonsingular M matrix described by the networks parameters,the connection matrix and the Lipschitz constant of the activation functions,are not only simple and practical,but also easier to check and less conservative than those imposed by similar results in recent literature.

Comments on″Capacity Analysis of the Asymptotically Stable Multi-Valued Exponential Bidirectinal Associative Memory″

Asymptotical stability is an important property of the associative memory neural networks.In this comment,we demonstrate that the asymptotical stability analyses of the MVECAM and MV-eBAM in the asynchronous update mode by Wang et al are not rigorous,and then we modify the errors and further prove that the two models are all asymptotically stable in both synchronous and asynchronous update modes.

Double-pattern associative memory neural network with pattern loop

A double-pattern associative memory neural network with “pattern loop” is proposed. It can store 2N bit bipolar binary patterns up to the order of 2 2N , retrieve part or all of the stored patterns which all have the minimum Hamming distance with input pattern, completely eliminate spurious patterns, and has higher storing efficiency and reliability than conventional associative memory. The length of a pattern stored in this associative memory can be easily extended from 2N to kN.

Using Quantum Associative Memory to Simulate the Brain Functions

To simulate the brain functions,a quantum associative memory combined with information preprocessing by a sparse coding model is presented. The sparse coding scheme is used to simulate the information transformation from retina up to primary visual cortex (V1) along the visual path and the quantum associative memory is used to simulate the pattern processing functions of the brain such as the pattern storing,forgetting and retrieving. Experimental results show that the model exhibits good associative ability on face recognition. Considering the huge storage capacity,mass parallel-distributed processing ability and oscillatory phenomena of the quantum system,this model might be a biological plausible implementation.

An Incremental Time-delay Neural Network for Dynamical Recurrent Associative Memory

An incremental time-delay neural network based on synapse growth, which is suitable for dynamic control and learning of autonomous robots, is proposed to improve the learning and retrieving performance of dynamical recurrent associative memory architecture. The model allows steady and continuous establishment of associative memory for spatio-temporal regularities and time series in discrete sequence of inputs. The inserted hidden units can be taken as the long-term memories that expand the capacity of network and sometimes may fade away under certain condition. Preliminary experiment has shown that this incremental network may be a promising approach to endow autonomous robots with the ability of adapting to new data without destroying the learned patterns. The system also benefits from its potential chaos character for emergence.

Memory Chain in a Chaotic Autoassociative Neural Network

Memory chain is observed in a chaotic autoassociative neural network. The network recalls first stored pattern from a fragment of a memory, stays at this pattern for a while, transits to the second stored pattern that overlaps with the first recalled pattern.Then it stays at the second recalled pattern for a while, transits to the third stored pattern that overlaps with the second recalled pattern, and so on. Thus a memory chain is generated. The memory chain ends with the pattern that overlaps no other stored patten. This phenomenon is similar to the way of recalling process of human beings in some respects.

Almost periodic solutions of memristive multidirectional associative memory neural networks with mixed time delays

Traditional biological neural networks cannot simulate the real situation of the abrupt synaptic connections between neurons while modeling associative memory of human brains.In this paper,the memristive multidirectional associative memory neural networks(MAMNNs)with mixed time-varying delays are investigated in the sense of Filippov solution.First,three steps are given to prove the existence of the almost periodic solution.Two new lemmas are proposed to prove the boundness of the solution and the asymptotical almost periodicity of the solution by constructing Lyapunov function.Second,the uniqueness and global exponential stability of the almost periodic solution of memristive MAMNNs are investigated by a new Lyapunov function.The sufficient conditions guaranteeing the properties of almost periodic solution are derived based on the relevant definitions,Halanay inequality and Lyapunov function.The investigation is an extension of the research on the periodic solution and almost periodic solution of bidirectional associative memory neural networks.Finally,numerical examples with simulations are presented to show the validity of the main results.

Synthesization of high-capacity auto-associative memories using complex-valued neural networks

In this paper, a novel design procedure is proposed for synthesizing high-capacity auto-associative memories based on complex-valued neural networks with real-imaginary-type activation functions and constant delays. Stability criteria dependent on external inputs of neural networks are derived. The designed networks can retrieve the stored patterns by external inputs rather than initial conditions. The derivation can memorize the desired patterns with lower-dimensional neural networks than real-valued neural networks, and eliminate spurious equilibria of complex-valued neural networks. One numerical example is provided to show the effectiveness and superiority of the presented results.

Perspective in Development of Shape Memory Materials Associated with Martensitic Transformation

By consideration of the characteristics of martensitic transformation and the derivation from the application of the group theory to martensitic transformation, it may be concluded that the shape memory effect (SME) can be attained in materials through a martensitic transformation and its reverse transformation. only when there forms single or nearly single variant of martensite, with an absence of the factors causing the generation of the resistance against SME. on this principle, various shape memory materials including nonferrous alloys. iron-based alloys and ceramics containjng zirconia are expected to be further developed. A criterion for thermoelastic martensitic transformation is presented, Factors which may act as the resistance against SME in various materials are briefly described

Changes of learning and memory ability associated with neuronal nitric oxide synthase in brain tissues of rats with acute alcoholism

BACKGROUD： Ethanol can influence neural development and the ability of leaming and memory, but its mechanism of the neural toxicity is not clear till now. Endogenous nitric oxide （NO） as a gaseous messenger is proved to play an important role in the formation of synaptic plasticity, transference of neuronal information and the neural development, but excessive nitro oxide can result in neurotoxicity. OBJECTIVE ： To observe the effects of acute alcoholism on the learning and memory ability and the content of neuronal nitric oxide synthase （nNOS） in brain tissue of rats. DESIGN ： A randomized controlled animal experiment. SETTING ： Department of Physiology, Xinxiang Medical College MATERIALS： Eighteen male clean-degree SD rats of 18-22 weeks were raised adaptively for 2 days, and then randomly divided into control group （n = 8） and experimental group （n = 10）. The nNOS immunohistochemical reagent was provided by Beijing Zhongshan Golden Bridge Biotechnology Co.,Ltd. Y-maze was produced by Suixi Zhenghua Apparatus Plant. METHODS ： The experiment was carded out in the laboratory of the Department of Physiology, Xinxiang Medical College from June to October in 2005. ① Rats in the experimental group were intraperitoneally injected with ethanol （2.5 g/kg） which was dissolved in normal saline （20%）. The loss of righting reflex and ataxia within 5 minutes indicated the successful model. Whereas rats in the control group were given saline of the same volume. ② Examinations of learning and memory ability： The Y-maze tests for learning and memory ability were performed at 6 hours after the models establishment. The rats were put into the Y-maze separately. The test was performed in a quiet and dark room. There was a lamp at the end of each of three pathways in Y-maze and the base of maze had electric net. All the lamps of the three pathways were turned on for 3 minutes and then turned off. One lamp was turned on randomly, and the other two delayed automatically. In 5 seconds after alternation, pulsating electric current presented in the base of unsafe area to stimulate rat＇s feet to run to the safe area. The lighting lasted for 15 seconds as one test. Running from unsafe area to safe area at one time in 10 seconds was justified as successful. Such test was repeated for 10 times for each rat and the successful frequency was recorded. The qualified standard of maze test was that the rat ardved in the safe area g times during 10 experiments. The number of trainings for the qualified standard was used to represent the result of spatial learning. ③ Determination of the content of nNOS in brain tissue： After the Y-maze test, the rats were anaesthetized, and blood was let from the incision on right auricle, transcardially perfused via the left ventricle with about 200 mL saline, then fixed by perfusion of 40 g/L paraformaldehyde. Hippocampal CA1 region, corpus striatum and cerebellum were taken to prepare serial freezing coronal sections. The nNOS contents in the brain regions were determined with the immunohistochemical methods to reflect the changes of nitdc oxide in brain tissue. MAIN OUTCOME MEASURES ： The changes of learning and memory ability and the changes of the nNOS contents in the brain tissue of rats with acute alcoholism were observed. RESULTS ： One rat in the experimental group was excluded due to its slow reaction to electdc stimulation in the Y-maze test, and the other 17 rats were involved in the analysis of results. ① The training times to reach qualifying standards of Y-maze in the expedmental group was more than that in the control group [（34.33 ±13.04）, （27.50±8.79） times, P〈 0.05]. ② Forms and numbers of nNOS positive neurons in brain tissue： It could be observed under light microscope that in the hippocampal CA1 region, there were fewer nNOS positive neurons, which were lightly stained, and the processes were not clear enough; But the numbers of the positive neurons which were deeply stained as huffy were obviously increased in the experimental group, the cell body and cyloplasm of process were evenly stained, but the nucleus was not stained. The nNOS positive neurons in corpus stdatum had similar forms and size in the experimental group and control group. The form of the nNOS positive neurons in cerebellum were similar between the two groups. The numbers of nNOS positive neurons in hippocampal CA1 region and corpus striatum in the expedmental group [（18.22±7.47）, （11.38±5.00） cells/high power field] were obviously higher than those in the control group [（10.15±4.24）, （6.15±3.69） cells/high power field. The number of nNOS positive neurons in cerebellum had no significant difference between the two groups [（49.56±18.84）, （44.43±15.42） cells/high power field, P〉 0.05]. CONCLUSION ： Acute alcoholism may impair learning and memory ability, and nitric oxide may be involved in mediating the neurotoxic role of ethanol.

The FAM(Fuzzy Asociative Memory)neural network model and its application in earthquake prediction

FAM(Fuzzy Associative Memory) Network Model, FAM Adaptive Learning Algorithm and Principal of FAM Inference Machine are introduced, and successfully application to ″New Generation Expert System for Earthquake Prediction″ (NGESEP). This system has good function for knowledge learning without disadvantages of neural network, which the learned knowledge implied in network is difficult to be understood or interpreted by expert system.

Knowledge Representation Methods in Expert System for Earthquake Prediction ESEP 3.0

Knowledge representation is a key to the building of expert systems. The performance of knowledge representation methods directly affects the intelligence level and the problem-solving ability of the system. There are various kinds of knowledge representation methods in ESEP3.0. In this paper, the authors introduce the knowledge representation methods, such as structure knowledge, seismological and precursory forecast knowledge, machine learning knowledge, synthetic prediction knowledge, knowledge to validate and verify certainty factors of anomalous evidence and support knowledge, etc. and propose a model for validation of certainty factors of anomalous evidence. The knowledge representation methods represent all kinds of earthquake prediction knowledge well.

High-Capacity Quantum Associative Memories

We review our models of quantum associative memories that represent the “quantization” of fully coupled neural networks like the Hopfield model. The idea is to replace the classical irreversible attractor dynamics driven by an Ising model with pattern-dependent weights by the reversible rotation of an input quantum state onto an output quantum state consisting of a linear superposition with probability amplitudes peaked on the stored pattern closest to the input in Hamming distance, resulting in a high probability of measuring a memory pattern very similar to the input. The unitary operator implementing this transformation can be formulated as a sequence of one-qubit and two-qubit elementary quantum gates and is thus the exponential of an ordered quantum Ising model with sequential operations and with pattern-dependent interactions, exactly as in the classical case. Probabilistic quantum memories, that make use of postselection of the measurement result of control qubits, overcome the famed linear storage limitation of their classical counterparts because they permit to completely eliminate crosstalk and spurious memories. The number of control qubits plays the role of an inverse fictitious temperature. The accuracy of pattern retrieval can be tuned by lowering the fictitious temperature under a critical value for quantum content association while the complexity of the retrieval algorithm remains polynomial for any number of patterns polynomial in the number of qubits. These models thus solve the capacity shortage problem of classical associative memories, providing a polynomial improvement in capacity. The price to pay is the probabilistic nature of information retrieval.

μ-stability of multiple equilibria in Cohen-Grossberg neural networks and its application to associative memory

In this paper, the μ-stability of multiple equilibrium points(EPs) in the Cohen-Grossberg neural networks(CGNNs) is addressed by designing a kind of discontinuous activation function(AF). Under some criteria, CGNNs with this AF are shown to possess at least 5^(n)EPs, of which 3^(n)EPs are locally μ-stable. Compared with the saturated AF or the sigmoidal AF, CGNNs with the designed AF can produce many more total/stable EPs. Therefore, when CGNNs with the designed discontinuous AF are applied to associative memory, they can store more prototype patterns. Moreover, the AF is expanded to a more general version to further increase the number of total/stable equilibria. The CGNNs with the expanded AF are found to produce(2k+3)^(n)EPs, of which (k+2)^(n)EPs are locally μ-stable. By adjusting two parameters in the AF, the number of sufficient conditions ensuring the μ-stability of multiple equilibria can be decreased. This finding implies that the computational complexity can be greatly reduced.Two numerical examples and an application to associative memory are illustrated to verify the correctness of the obtained results.