Concept Of Associative Memory Research Articles

Efficient Memristive Circuit Design of Neural Network-Based Associative Memory for Pavlovian Conditional Reflex.

The brain’s learning and adaptation processes heavily rely on the concept of associative memory. One of the most basic associative learning processes is classical conditioning. This work presents a memristive neural network-based associative memory system. The system can emulate Pavlovian conditioning principles including acquisition, extension, generalization, differentiation, and spontaneous recovery that have not been considered in most of the previous counterparts. The proposed circuit can emulate these principles thanks to the resistance-changing characteristics of the memristor. Generalization has been achieved by providing both unconditional and neutral stimuli to the network to reduce the memristance of the memristor. Differentiation has been attained by employing unconditional and conditional stimuli in a training scheme to obtain a certain memristance that causes the network to respond differently to both stimuli. A revival of an exterminated stimuli is also done by increasing the synaptic weight of the system. Compared to previous designs, the proposed memristive circuit can implement all the functions of conditional reflex. Our rigorous simulations demonstrated that the proposed memristive system can condition neutral stimuli, show generalization between similar stimuli, distinguish dissimilarities between the generalized stimuli, and recover faded stimuli.

Associative Memories With Synaptic Delays.

In this paper, we introduce a new concept of associative memories in which synaptic connections of the self-organizing neural network learn time delays between input sequence elements. Synaptic connections represent both the synaptic weights and expected delays between the network inputs. This property of synaptic connections facilitates recognition of time sequences and provides context-based associations between sequence elements. Characteristics of time delays are learned and are updated each time an input sequence is presented. There are no separate learning and testing modes typically used in other neural networks, as the network starts to predict the next input element as soon as there is no expected input signal. The network generates output signals useful for associative recall and prediction. These output signals depend on the presented input context and the knowledge stored in the graph. Such a mode of operation is preferred for the organization of episodic memories used to store the observed episodes and to recall them if a sufficient context is provided. The associative sequential recall is useful for the operation of working memory in a cognitive agent. Test results demonstrate that the network correctly recognizes the input sequences with variable delays and that it is more efficient than other recently developed sequential memory networks based on associative neurons.

Bidirectional and multidirectional associative memories as models in linkage analysis in data analytics: Conceptual and algorithmic developments

Associative and bidirectional associative memories are examples of associative structures studied intensively in the literature and exhibiting a large volume of applied studies. The underlying idea is to reveal and describe linkages among data and express them in a form of an associative mapping. Such mappings are constructed in a way so that the recall processes (both one-directional and bidirectional) lead to the recalled items characterized by a minimal recall error. Associative memories, morphological memories, and fuzzy associative memories have been studied in numerous areas yielding efficient applications to image recall and enhancements and fuzzy controllers (which can be regarded as one-directional associative memories). In this study, we revisit and augment the concept of associative memories by offering some new conceptual design insights where the corresponding mappings are realized on a basis of a related collection of landmarks (prototypes) over which an associative mapping becomes spanned. In light of the bidirectional character of mappings, we develop an augmentation of the existing fuzzy clustering (Fuzzy C-Means) in the form of a so-called collaborative fuzzy clustering. Here an interaction in the construction of prototypes is optimized so that the bidirectional recall error can be minimized. Further conceptual architectural augmentations are discussed including a relational description of associative memories and linkage analysis accomplished in the presence of explanatory spaces. We generalize the mapping into its granular version in which numeric prototypes formed through the clustering process are made granular so that the quality of the recall can be quantified. Several scenarios of allocation of information granularity aimed at the optimization of the characteristics of recalled results (information granules) quantified in terms of coverage and specificity are proposed. Illustrative examples are presented as well.

Granular Bidirectional and Multidirectional Associative Memories: Towards a Collaborative Buildup of Granular Mappings

Associative and bidirectional associative memories are examples of associative structures studied intensively in the literature. The underlying idea is to realize associative mapping so that the recall processes (one-directional and bidirectional ones) are realized with minimal recall errors. Associative and fuzzy associative memories have been studied in numerous areas yielding efficient applications for image recall and enhancements and fuzzy controllers, which can be regarded as one-directional associative memories. In this study, we revisit and augment the concept of associative memories by offering some new design insights where the corresponding mappings are realized on the basis of a related collection of landmarks (prototypes) over which an associative mapping becomes spanned. In light of the bidirectional character of mappings, we have developed an augmentation of the existing fuzzy clustering (fuzzy c-means, FCM) in the form of a so-called collaborative fuzzy clustering. Here, an interaction in the formation of prototypes is optimized so that the bidirectional recall errors can be minimized. Furthermore, we generalized the mapping into its granular version in which numeric prototypes that are formed through the clustering process are made granular so that the quality of the recall can be quantified. We propose several scenarios in which the allocation of information granularity is aimed at the optimization of the characteristics of recalled results (information granules) that are quantified in terms of coverage and specificity. We also introduce various architectural augmentations of the associative structures.

Investigation on Dynamic Speech Emotion from the Perspective of Brain Associative Memory

Many researchers have studied speech emotion for years from the perspective of psychology to engineering. To date, none has made the speech emotion recognition system intuitive enough in such a way that it can be embedded in automatic answering machines that can effectively detect the various affective states of human verbal communication. In most cases the underlying emotional information was misinterpreted thus resulting in wrong feedbacks and responses. The complexity of understanding and analyzing speech emotion is presented in the dynamics of the emotion itself. Emotion is dynamic and changeable over time. Hence, it is imperative to cater for this parameter to boost the performance of the speech emotion recognition system. In this paper, values of Valence (V) and Arousal(A) are used to generate a recalibrated affective space model. Such approach is adopted from psychologists’ understanding that emotion can be represented using emotion primitives’ values. The VA approach is then coupled with the brain associative memory concept that can provides a better means in understanding the dynamics of speech emotion. Results of such analysis tallies with the psychological findings and has its practical implementation.

A prototype of programmable associative memory for track finding

We present a device, based on the concept of associative memory for pattern recognition, dedicated to on-line track finding in high-energy physics experiments. A large pattern bank, describing all possible tracks, can be organized into field programmable gate arrays where all patterns are compared in parallel to data coming from the detector during readout. Patterns, recognized among 2/sup 66/ possible combinations, are output in a few 30 MHz clock cycles. Programmability results in a flexible, simple architecture and it allows one to keep up smoothly with technology improvements. A 64 PAM array has been assembled on a prototype VME board and fully tested up to 30 MHz.

A programmable associative memory for track finding

We present a device, based on the concept of associative memory for pattern recognition, dedicated to on-line track finding in high-energy physics experiments. A large pattern bank, describing all possible tracks, can be organized into Field Programmable Gate Arrays where all patterns are compared in parallel to data coming from the detector during readout. Patterns, recognized among 2 66 possible combinations, are output in a few 30 MHz clock cycles. Programmability results in a flexible, simple architecture and it allows to keep up smoothly with technology improvements.

Detecting chaos with local associative memories

A technique to discriminate complex signals associated with deterministic chaos from those of random origin is presented. The method applies a new, computationally efficient, prediction procedure based on the associative memory concept. This procedure was used as a method to model non-linear series. Its performance was analyzed for several time series, including simulated numerical data from the logistic map and from the Mackey-Glass delay equation. Also experimental data from a dripping faucet in a chaotic regime and from low-temperature thermal fluctuations of the voltage across a resistor are considered.

A parallel architecture for textureanalysis based on the concept of associative memory

A parallel architecture suited for texture analysis is presented. The architectureis based on the concept of associative memory, which has learning and recognizing capabilities. With image preprocessing, over 90% correct classification was obtained.

Associate memories for system identification: Inverse problems in remote sensing

The problems of remote sensing of physical parameters from satellites is an important inverse, or system identification problem. Progress has been made in several phases. During the 1960's basic equations of radiative transfer in various types of media (inhomogeneous, emitting, anisotropic, polarized, etc.) were derived for the treatment of direct problems, i.e., problems in which the parameters were known. In order to avoid instabilities in numerically obtaining solutions, nonlinear Cauchy problems for systems of ordinary differential equations were developed. This formulation aided in the treatment of inverse problems, where the parameters are to be estimated based on observations of external or internal radiant energy. System identification techniques such as quasilinearization were found to be effective, when good preliminary estimates were available. Recently, a systematic procedure for obtaining the needed initial estimates has been investigated. This approach, based on the concept of associative memories, has already produced excellent computational results.

VLSI structures for track finding

We discuss the architecture of a device based on the concept of associative memory designed to solve the track finding problem, typical of high energy physics experiments, in a time span of a few microseconds even for very high multiplicity events. This “machine” is implemented as a large array of custom VLSI chips. All the chips are equal and each of them stores a number of “patterns”. All the patterns in all the chips are compared in parallel to the data coming from the detector while the detector is being read out.