Search

Ternary Content Addressable Memory (TCAM) is an important hardware used to store route entries in routers, which is used to assist routers to make fast decision on forwarding packets. In order to cope with the explosion of route entries due to massive IP terminals brought by 5G and the Internet of Things (IoT), today’s commercial TCAM has to keep the corresponding growth in capacity. But large TCAM capacity is causing many problems such as circuit design difficulties, production costs, and high energy consumption, so it is urgent to design a lightweight TCAM with small capacity while still maintains the original query performance.Designing such a TCAM faces two fundamental challenges. Firstly, it is essential to accurately predict the incoming flows in order to cache correct entries in limited TCAM capacity, but prediction on aggregated time-sequential data is challenging in the massive IoT scenarios. Secondly, the prediction algorithm needs to be real-time as the lookup process is in line-rate. In order to address the above two challenges, in this paper, we proposed a lightweight AI-based solution, called AIR, where we successfully decoupled the route entries and designed a parallel-LSTM prediction method. The experiment results under real backbone traffic showed that we successfully achieved comparable query performance by using just 1/8 TCAM size.

Content-addressable memory (CAM) has been a critical component in pattern matching and also machine-learning applications. Recently emerged CAM that is capable of delivering multi-level distance calculation is promising for applications that need matching results beyond Boolean results of ?matched" and ?not matched". However, existing multi-level CAM designs are constrained by the bit-cell device discharging current mismatch and the strict timing of sensing operations for distance calculation. This fact results in the challenge of further improving the accuracy and scalability towards higher-resolution and higher-dimension matching. This work presents a multi-level CAM design that is capable of delivering high-accuracy and high-scalability search, which is immune to the discharging device mismatch and needs no strict timing for result sensing. The inherent enabler is the charge-domain computing mechanism. This work will present the operating mechanisms, the circuit simulation, and content-matching evaluation results, showing the promise towards high reliability, high energy efficiency, and high scalability.

Embedded/edge computing comes with a very stringent hardware resource (area) budget and a need for extreme energy efficiency. This motivates repurposing, i.e., reconfiguring hardware resources on demand, where the overhead of reconfiguration itself is subject to the very same tight budgets in area and energy efficiency. Numerous applications running on resource constrained environments such as wearable devices and Internet-of-Things incorporate CAM (Content Addressable Memory) as a key computational building block. In this paper we present CAMeleon -- a novel energy-efficient compute substrate which can seamlessly be reconfigured to perform CAM operations in addition to logic and memory functions. CAMeleon has a similar level of latency to conventional CAM designs based on SRAM and emerging memory technologies (such as STT-MTJ, ReRAM and PCM), however, performs CAM operations more energy-efficiently, consumes less area, and can support traditional logic and memory functions beyond CAM operations on demand thanks to its reconfigurability.

Content addressable memory (CAM) for search and match operations demands high speed and low power for near real-time decision-making across many critical domains. Resistive RAM (RRAM)-based in-memory computing has high potential in realizing an efficient static CAM for artificial intelligence tasks, especially on resource-constrained platforms. This paper presents an XNOR-based RRAM-CAM with a time-domain analog adder for efficient winning class computation. The CAM compares two operands, one voltage and the second one resistance, and outputs a voltage proportional to the similarity between the input query and the pre-stored patterns. Processing the summation of the output similarity voltages in the time-domain helps avoid voltage saturation, variation, and noise dominating the analog voltage-based computing. After that, to determine the winning class among the multiple classes, a digital realization is utilized to consider the class with the longest pulse width as the winning class. As a demonstrator, hyperdimensional computing for efficient MNIST classification is considered. The proposed design uses 65 nm CMOS foundry technology and realistic data for RRAM with total area of 0.0077 mm2, consumes 13.6 pJ of energy per 1 k query within 10 ns clock cycle. It shows a reduction of ~ 31 × in area and ~ 3 × in energy consumption compared to fully digital ASIC implementation using 65 nm foundry technology. The proposed design exhibits a remarkable reduction in area and energy compared to two of the state-of-the-art RRAM designs.

This paper presents a 2G-search/s embedded Ternary Content Addressable Memory (TCAM) design in 5nm Fin-FET technology with the ability to update both SRAM words in a TCAM entry in a single clock cycle. This reduces TCAM update latency by 50% for data center Application Specific Integrated Circuits (ASICs) with only 1% area overhead and no search power penalty. We present a novel time multiplexed input bus interface on a single port TCAM cell array and new architecture to enable fast updates. Silicon measurement shows the highest reported search rate of 2G-search/s at a 3.48Mb/mm2 memory density including all global peripheral circuitry for a 512 entry, 220-bit wide, 110Kb TCAM.

We introduce a near-term experimental platform for realizing an associative memory. It can simultaneously store many memories by using spinful bosons coupled to a degenerate multimode optical cavity. The associative memory is realized by a confocal cavity QED neural network, with the cavity modes serving as the synapses, connecting a network of superradiant atomic spin ensembles, which serve as the neurons. Memories are encoded in the connectivity matrix between the spins, and can be accessed through the input and output of patterns of light. Each aspect of the scheme is based on recently demonstrated technology using a confocal cavity and Bose-condensed atoms. Our scheme has two conceptually novel elements. First, it introduces a new form of random spin system that interpolates between a ferromagnetic and a spin-glass regime as a physical parameter is tuned---the positions of ensembles within the cavity. Second, and more importantly, the spins relax via deterministic steepest-descent dynamics, rather than Glauber dynamics. We show that this nonequilibrium quantum-optical scheme has significant advantages for associative memory over Glauber dynamics: These dynamics can enhance the network's ability to store and recall memories beyond that of the standard Hopfield model. Surprisingly, the cavity QED dynamics can retrieve memories even when the system is in the spin glass phase. Thus, the experimental platform provides a novel physical instantiation of associative memories and spin glasses as well as provides an unusual form of relaxational dynamics that is conducive to memory recall even in regimes where it was thought to be impossible., Comment: 35 pages, 16 figures, 6 appendices

Key areas of the episodic memory (EM) network demonstrate changing structure and volume during adolescence. EM is multifaceted and yet studies of EM thus far have largely examined single components, used different methods and have unsurprisingly yielded inconsistent results. The Treasure Hunt task is a single paradigm that allows parallel investigation of memory content, associative structure, and the impact of different retrieval support. Combining the cognitive and neurobiological accounts, we hypothesized that some elements of EM performance may decline in late adolescence owing to considerable restructuring of the hippocampus at this time. Using the Treasure Hunt task, we examined EM performance in 80 participants aged 10–17 yr. Results demonstrated a cubic trajectory with youngest and oldest participants performing worst. This was emphasized in associative memory, which aligns well with existing literature indicating hippocampal restructuring in later adolescence. It is proposed that memory development may follow a nonlinear path as children approach adulthood, but that future work is required to confirm and extend the trends demonstrated in this study.

The hippocampus is crucial for forming associations between environmental stimuli. However, it is unclear how neural activities of hippocampal neurons dynamically change during the learning process. To address this question, we developed an associative memory task for rats with auditory stimuli. In this task, the rats were required to associate tone pitches (high and low) and ports (right and left) to obtain a reward. We recorded the firing activity of neurons in rats hippocampal CA1 during the learning process of the task. As a result, many hippocampal CA1 neurons increased their firing rates when the rats received a reward after choosing either the left or right port. We referred to these cells as “reward-direction cells.” Furthermore, the proportion of the reward-direction cells increased in the middle-stage of learning but decreased after the completion of learning. This result suggests that the activity of reward-direction cells might serve as “positive feedback” signal that facilitates the formation of associations between tone pitches and port choice.

Local circuits in different cortical areas and animal species share ubiquitous features of synaptic connectivity and neural spiking activity. For example, neurons in local cortical circuit are more likely to be bidirectionally connected and to exhibit particular connectivity patterns among three or

Deep neural networks are efficient at learning from large sets of labelled data, but struggle to adapt to previously unseen data. In pursuit of generalized artificial intelligence, one approach is to augment neural networks with an attentional memory so that they can draw on already learnt knowledge patterns and adapt to new but similar tasks. In current implementations of such memory augmented neural networks (MANNs), the content of a network’s memory is typically transferred from the memory to the compute unit (a central processing unit or graphics processing unit) to calculate similarity or distance norms. The processing unit hardware incurs substantial energy and latency penalties associated with transferring the data from the memory and updating the data at random memory addresses. Here, we show that ternary content-addressable memories (TCAMs) can be used as attentional memories, in which the distance between a query vector and each stored entry is computed within the memory itself, thus avoiding data transfer. Our compact and energy-efficient TCAM cell is based on two ferroelectric field-effect transistors. We evaluate the performance of our ferroelectric TCAM array prototype for one- and few-shot learning applications. When compared with a MANN where cosine distance calculations are performed on a graphics processing unit, the ferroelectric TCAM approach provides a 60-fold reduction in energy and 2,700-fold reduction in latency for a single memory search operation. A compact ternary content-addressable memory cell, which is based on two ferroelectric field-effect transistors, can provide memory augmented neural networks with improved energy and latency performance compared with traditional approaches based on graphics processing units.

Successful episodic memory requires binding of event details across spatial and temporal gaps. The neural processes underlying mnemonic binding, however, are not fully understood. Moreover, although acute stress is known to modulate memory, if and how stress changes mnemonic integration across time and space is unknown. To elucidate these issues, we exposed participants to a stressor or a control manipulation shortly before they completed, while electroencephalography was recorded, an encoding task that systematically varied the demands for spatial and temporal integration. Associative memory was tested 24 h later. While early event-related potentials, including the P300 and Late Positive Component, distinguished different levels of spatiotemporal discontinuity, only later Slow Waves were linked to subsequent remembering. Furthermore, theta oscillations were specifically associated with successful mnemonic binding. Although acute stress per se left mnemonic integration largely unaffected, autonomic activity facilitated object memory and glucocorticoids enhanced detail memory, indicative for mnemonic integration. At the neural level, stress amplified the effects of spatiotemporal discontinuity on early information processing. Together, our results indicate that temporal and spatial gaps recruit early neural processes, providing attentional resources. The actual binding success, however, appears to depend on later processes as well as theta power and may be shaped by major stress response systems.

There is a lack of research investigating associative memory errors in sign language. Thus, the Deese/Roediger-McDermott (DRM) paradigm was used to test recall and recognition errors in deaf signers following four types of encoding: written words, signs, written words presented along with signs, or written words combined with self-produced signs. The results revealed poor memory for studied material along with a relatively low level of false memory for critical items. However, dual coding appeared to enhance veridical memory. Interestingly, a high rate of noncritical intrusions was noticed across all four encoding conditions. Similar results were obtained for recognition, however, there were no differences between encoding conditions in the false alarms rate toward critical lures. The results point to difficulties in accessing the meaning of words in deaf participants. To test this hypothesis a follow-up study was conducted where false recall and recognition of words or signs were tested in hearing participants fluent in sign language. The second experiment revealed semantic memory errors comparable to previous studies with hearing participants. The results from both experiments are discussed in light of previous studies on vocabulary knowledge in deaf individuals.

A generalization of active neural associative knowledge graphs (ANAKGs) to their minicolumn form is presented in this paper. Each minicolumn represents a single symbol, and the activation of an individual neuron in a minicolumn depends on the context of the activation of the presynaptic neuron. The implemented memory model combines the ANAKG associative spiking neuron idea with the idea of the hierarchical temporal memory. This new associative memory organization preserves all properties of ANAKG memories, such as storage of knowledge based on the association of spatiotemporal input sequences, self-organization, quick learning, and recall of the sequential memories, while increasing the recall quality and the memory capacity. The recall quality advantage of the new approach over ANAKG increases with the length of the recalled episodes and the number of neurons used in each minicolumn. We introduced a new distance measure to compare the recalled sequences and defined a recall quality to determine the memory capacity. Performed tests confirmed our claims. Additional tests were performed to illustrate the computational complexity and the efficiency of the developed approach.

Memory being one of the essential credential in today’s computer world seeks forward newer research interests in its types. Hopfield neural networks of artificial neural networks are one of its classes that can be modelled to form an associative memory. In this paper, we have shown the Hopfield neural network constructed with spintronic memristor bridges accounting to act as an associative memory unit. The memristors are nanoscaled, in terms of size, which possess synaptic behaviour in the artificial neuromorphic system. The associative behaviour is realised by the updation of synaptic weights of memristive Hopfield with single- and multiple-bit associativity which is simulated in MATLAB. The application of the hardware in the field of cryptography is also proposed.

Background Humans constantly take in vast amounts of information, which must be filtered, flexibly manipulated, and integrated into cohesive relational memories in order to choose relevant behaviors. Relational memory is impaired in chronic schizophrenia, which has been linked to hippocampal dysfunction. It is unclear whether relational memory is impaired in the early stage of psychosis. Methods We studied eye movements during a face-scene pairs task as an indirect measure of relational memory in 89 patients in the early stage of psychosis and 84 healthy control participants. During testing, scenes were overlaid with three equally-familiar faces and participants were asked to recall the matching (i.e. previously-paired) face. During Match trials, one face had been previously paired with the scene. During Non-Match trials, no faces matched the scene. Forced-choice explicit recognition was recorded as a direct measure of relational memory. Results Healthy control subjects rapidly (within 250–500 ms) showed preferential viewing of the matching face during Match trials. In contrast, preferential viewing was delayed in patients in the early stage of psychosis. Explicit recognition of the matching face was also impaired in the patient group. Conclusions This study provides novel evidence for a relational memory deficit in the early stage of psychosis. Patients showed deficits in both explicit recognition as well as abnormal eye-movement patterns during memory recall. Eye movements provide a promising avenue for the study of relational memory in psychosis, as they allow for the assessment of rapid, nonverbal memory processes.

Prior research has demonstrated that stimulus emotionality can affect associative memory, but the patterns are varied: negative valence can either impair or improve it, while positive valence often facilitates it. In addition, it remains unclear whether the influence of stimulus emotionality on associative memory is sensitive to the manipulation of directed forgetting (DF). Thus, this experiment further explored the modulation of stimulus emotionality on associative memory in the case of DF. Besides the intact, rearranged, and new pairs usually adopted, we also included a fourth type of “old+new” pair. The results showed that the correct responses were not exactly the same among pair types; it tended to be better for both neutral and positive pairs vs negative pairs; significant DF effect was confirmed, but its susceptibility to stimulus emotionality only held true for “old+new” case. Further analyses of old proportions and response times provided further evidence for the above patterns. These findings demonstrate that the involvements of the recollection-driven process differ as the function of pair type; the modulation of stimulus emotionality on associative memory reinforces the dual representation account; reliable DF effect suggests both the pivotal roles of selective rehearsal to TBR (to-be-remembered) pairs and of inhibitory control to TBF (to-be-forgotten) pairs; furthermore, the susceptibility of the magnitude of DF effect to stimulus emotionality depends upon the involvement of recollection-driven process.

This letter proposes a novel approach to design a ternary content-addressable memory (TCAM) on a field-programmable gate array (FPGA). The proposed architecture uses slice registers (SRs) for storage instead of lookup tables and static random-access memory, and a control signal in place of a memory element to accommodate a don’t care state. The proposed architecture is implemented on Xilinx Virtex-6 FPGA. Compared with the latest prior work, the proposed TCAM reduces the number of SRs by 50.04% while not compromising on throughput.

In this paper, the fixed-time synchronization control problem of memristive multidirectional associative memory neural networks (MMAMNNs) is considered. Based on the nonlinear and chaos characteristics of memristor, a chaotic model is constructed. And then, utilizing the Lyapunov stability theory, two appropriate controllers are constructed and different activation functions are used. This control method ensures that drive system and response system can achieve synchronization within a fixed time. So, compared with previous studies, it has more practical value. In addition, we present a fixed-time synchronization chaotic encryption method, the chaos characteristic of the model is used to encrypt plaintext, and the decryption of ciphertext is realized based on the synchronization control theories. Finally, several numerical simulations are given to demonstrate the validity of the theories and the chaotic secure communication scheme.

Strong evidence suggests that sleep plays a role in memory consolidation, which involves both stabilizing memory into long-term storage as well as integrating new information into existing stores. The current study investigated consolidation, across a day of wakefulness or night of sleep, of emotional and neutral directly learned visual paired associates (A-B/B-C pairs) as well as formation of memory for relational pairs formed via overlapping learned components (A-C pairs). Participants learned 40 negative and 40 neutral face-object pairs followed by a baseline test in session 1 either in the morning or evening. They then spent a 12-hour retention period during which participants either went about their normal day or spent the night in the sleep lab. During session 2, participants completed a surprise test to assess their memory for relational pairs (A-C) as well as memory for direct associates (A-B/B-C). As hypothesized, the results demonstrated that a 12-hour retention period predominantly spent asleep, compared to awake, benefited memory for both relational and direct associative memory. However, contrary to the hypothesis that emotional salience would promote preferential consolidation, sleep appeared to benefit both negative and neutral information similarly for direct associative and relational memories, suggesting that sleep may interact with other factors affecting encoding (e.g., depth of encoding) to benefit direct and relational associative memory. As one of the few studies examining the role of nocturnal sleep and emotion on both direct and relational associative memory, our findings suggest key insights into how overnight sleep consolidates these different forms of memory.

Significance This work makes 2 contributions. First, we present a neural network model of associative memory that stores and retrieves sparse patterns of complex variables. This network can store analog information as fixed-point attractors in the complex domain; it is governed by an energy function and has increased memory capacity compared to early models. Second, we translate complex attractor networks into spiking networks, where the timing of the spike indicates the phase of a complex number. We show that complex fixed points correspond to stable periodic spike patterns. It is demonstrated that such networks can be constructed with resonate-and-fire or integrate-and-fire neurons with biologically plausible mechanisms and be used for robust computations, such as image retrieval., Information coding by precise timing of spikes can be faster and more energy efficient than traditional rate coding. However, spike-timing codes are often brittle, which has limited their use in theoretical neuroscience and computing applications. Here, we propose a type of attractor neural network in complex state space and show how it can be leveraged to construct spiking neural networks with robust computational properties through a phase-to-timing mapping. Building on Hebbian neural associative memories, like Hopfield networks, we first propose threshold phasor associative memory (TPAM) networks. Complex phasor patterns whose components can assume continuous-valued phase angles and binary magnitudes can be stored and retrieved as stable fixed points in the network dynamics. TPAM achieves high memory capacity when storing sparse phasor patterns, and we derive the energy function that governs its fixed-point attractor dynamics. Second, we construct 2 spiking neural networks to approximate the complex algebraic computations in TPAM, a reductionist model with resonate-and-fire neurons and a biologically plausible network of integrate-and-fire neurons with synaptic delays and recurrently connected inhibitory interneurons. The fixed points of TPAM correspond to stable periodic states of precisely timed spiking activity that are robust to perturbation. The link established between rhythmic firing patterns and complex attractor dynamics has implications for the interpretation of spike patterns seen in neuroscience and can serve as a framework for computation in emerging neuromorphic devices.

There are documented deficits in older adults’ abilities to bind numerical information to other types of information, perhaps due to the arbitrariness and specificity of numbers. Although some studies have found that memory for associative details is more accurate for emotionally salient information than for emotionally neutral information, other research has failed to find this benefit. We investigated whether older adults’ associative memory deficit for numerical information may be reduced when information is encountered in an emotionally salient context. We presented younger and older adults with numerical information in a sentence that was emotionally positive, negative, or neutral and later asked them to recall the numbers when given their corresponding context. Although younger adults recalled more information than older adults, both groups of participants recalled more numbers in emotionally valenced as compared with emotionally neutral contexts, with the most accurate memory for numbers in the highly arousing negative context. Both groups of participants also rated the negative information as more important and easier to remember. These results provide evidence that emotion-enhanced binding is consistent between younger and older adults in some contexts and that memory for specific and arbitrary numerical information may be more accurate in an emotionally salient as compared with emotionally neutral context.

In this paper, we propose an adaptive-bias-training circuit for enhancing the read reliability of ternary content-addressable memory (TCAM), which is implemented with resistive nonvolatile memories such as spin–torque transfer magnetic random access memory (STT-MRAM) and PRAM. The 3T2R nonvolatile-based TCAM (nvTCAM) is area-efficient, yet vulnerable to the process variation with fixed bias condition that affects the sensing margin. With adaptive bias training, the 3T2R nvTCAM cell can maintain a sufficient sensing margin when combined with a high-performance STT-MRAM with a resistance ratio (R-ratio) as low as 3.0. The foreground train followed by the on-the-fly track compensates for variations in the resistances of memory cells, threshold voltage, and substrate temperature. We evaluated the proposed bias technique using the 180-nm CMOS process with 1.8 V of VDD. The search speed is enhanced by 50% at a 64-bit word length, and the sensing margin is improved by 20%. Using our proposed bias technique, the read error rate induced by the variability is estimated to be contained below 1 ppb at an R-ratio of 2, indicating the decent productivity of gigabit density nvTCAM with STT-MRAM.

Modern field-programmable gate arrays (FPGAs) provide a vast amount of logic resources that can be used to implement complex systems while providing the flexibility to modify the design once deployed. This makes them attractive for software-defined networks (SDNs) applications, and, in fact, most vendors provide the building blocks needed for those applications, which include basic packet classification functions such as exact match, longest prefix match, and match with wildcards. Those are needed for different functions such as routing, security filtering, monitoring or quality of service. The match with wildcards can be done using ternary content addressable memories (TCAMs). TCAMs can be implemented as independent standalone devices or as Internet Protocol (IP) blocks that are used inside networking application-specific integrated circuits (ASICs) such as switching ICs. In both cases, the cells of a TCAM are more complex than that of a normal memory and also than that of a binary content addressable memory (CAMs). This is due to the more complex matching that they need to implement. As FPGAs are used in many different applications, it does not make sense to include TCAM blocks inside them as they would be used only in a small fraction of the systems. Therefore, TCAMs are emulated using the logic resources available inside the FPGA. In recent years, a number of schemes to emulate TCAMs on FPGAs have been proposed, some of them based on the use of the logic resources and others on the use of the embedded memory blocks available on the FPGA. In this brief, a technique to efficiently emulate TCAMs on Xilinx FPGAs is presented. The proposed scheme is based on the use of lookup tables (LUTs) and partial reconfiguration to achieve a more effective use of the FPGA resources while supporting the addition and removal of rules. The proposed scheme has been compared to existing implementations and the results show that it can achieve significant savings in resource usage. In addition, it enables the use of all the LUTs in the device for TCAM implementation, something that is not supported by existing approaches that use LUTRAMs.

Memory deficits in aging are characterized by impaired hippocampus-mediated relational binding-the formation of links between items in memory. By reducing reliance on relational binding, unitization of two items into one concept enhances associative recognition among older adults. Can a similar enhancement be obtained when probing memory with recall? This question has yet to be examined, because recall has been assumed to rely predominantly on relational binding. Inspired by recent evidence challenging this assumption, we investigated individual differences in older adults' recall of unitized and nonunitized associations. Compared with successfully aging individuals, older adults with mild memory deficits, typically mediated by the hippocampus, were impaired in recall of paired-associates in a task which relies on relational binding (study: "PLAY-TUNNEL"; test: PLAY-T?). In stark contrast, the two groups showed similar performance when items were unitized into a novel compound word (study: "LOVEGIGGLE"; test: LOVEG?). Thus, boosting nonrelational aspects of recall enhances associative memory among aging individuals with subtle memory impairments to comparable levels as successfully aging older adults.

We present a quantum probabilistic associative memory using the inverse of quantum Fourier transform and Grover’s algorithm to recover existing or similar patterns in the memory. The content of the memory is created using a generator of a superposition state representing a given set of patterns. We discuss the architecture of the proposed memory including the storing, recovering and processing of similarity tolerance of the input query. The associative memory can extrapolate and recover similar stored patterns. The system is unitary and runs in O(n) steps, where n is the number of qubits of the patterns.

A data bank is using during designing technological processes of machining, in which it is necessary to find the required information and put it together depending on the task. This process raises the need to build a multi-level structure of data processing. It is also necessary to provide a quick search for the required information in the data bank. This problem can be solved with the help of an associative memory, which can be applied as during searching for information and while further saving the obtained technological process. The aim of the article is the development of neural networks of associative memory for the design and saving of technological processes for high-precision and unique parts. Results. A technological process for the production of a specific part with the using of the proposed neural networks of associative memory has been developed. The algorithm for training individual modules of a multilayer network is the process of determining the training set of images and constructing the weight matrices of the links between the input and output layers of the neurons. When using associative memory, the speed of data work is increased due to the parallel processing of information. Mathematical modeling of the production process details confirmed the correctness of the theoretical principles. Conclusions. The neural networks for the design and saving of technological processes for the production of high-precision parts have been developed.

Ternary content-addressable memory (TCAM) is a type of associative memory used in many applications for high-speed data searching. We present herein a gate-all-around (GAA) carbon nanotube field-effect transistor (CNTFET)-based self-controlled TCAM cell design with a precharge-free match line. We compare the power–delay product (PDP) and static noise margin between the GAA-CNTFET-based traditional and proposed TCAM cell designs at the 11-nm technology node with a supply voltage of 0.8 V. The simulations are performed using the Virtuoso tool for different parameter values with the Stanford University GAA-CNTFET model. The simulation results show that, compared with the traditional design, the proposed design exhibits a significant reduction in power by 51.30%, delay by 17.16%, and PDP by 59.66% for a chiral vector of (20, 16, 0) with two channels. It is observed that the best chirality for the proposed design is (14, 20, 0) for a single channel, but (16, 16, 0) and (20, 16, 0) for a dual channel in terms of power, delay, stability, and PDP.

Packet Classification is the enabling function performed in commodity switches for providing various services such as access control, intrusion detection, load balancing, and so on. Ternary Content Addressable Memories (TCAMs) are the de facto standard for performing packet classification at high speeds. However, TCAMs are highly costlier both in terms of cost and power consumption, forcing the switch vendors towards placing lots of effort for power management. Hence, power-efficient solutions for TCAM-based packet classification are highly relevant even today. In this article, we propose a novel rule placement algorithm based on the unique field values’ presence within the rule databases. We evaluate the total search that is needed to be inspected with respect to the traditional placement approach and the proposed placement approach based on the information content within the fields. Simulation results showed an average reduction of 30.55% in the search space by the proposed placement approach, thereby resulting in an average reduction of 18.85% per search energy over TCAM. With typical TCAM clock speeds ranging between 200--400MHz, this reduction in the per-search energy maps to a huge reduction in the total energy consumed by the TCAM-based network switches. The proposed solution is plug-and-play type requiring only minimal pre-processing within the Network Processing Unit (NPU) of the switches and edge routers.

Learning to predict rewards based on environmental cues is essential for survival. The orbitofrontal cortex (OFC) contributes to such learning by conveying reward-related information to brain areas such as the ventral tegmental area (VTA). Despite this, how cue-reward memory representations form in individual OFC neurons and are modified based on new information is unknown. To address this, using in vivo two-photon calcium imaging in mice, we tracked the response evolution of thousands of OFC output neurons, including those projecting to VTA, through multiple days and stages of cue-reward learning. Collectively, we show that OFC contains several functional clusters of neurons distinctly encoding cue-reward memory representations, with only select responses routed downstream to VTA. Unexpectedly, these representations were stably maintained by the same neurons even after extinction of the cue-reward pairing, and supported behavioral learning and memory. Thus, OFC neuronal activity represents a long-term cue-reward associative memory to support behavioral adaptation.

Associative memory (AM) deficits are common in neurodegenerative disease and novel therapies aimed at improving these faculties are needed. Theta band oscillations within AM networks have been shown to be important for successful memory encoding and modulating these rhythms represents a promising strategy for cognitive enhancement. Transcranial alternating current stimulation (TACS) has been hypothesized to entrain and increase power of endogenous brain rhythms. For this reason, we hypothesized that focal delivery of theta band electrical current, using high-definition TACS, would result in improved AM performance compared to sham stimulation or transcranial direct current stimulation (TDCS). In this pilot study, 60 healthy subjects were randomized to receive high definition TACS, high definition TDCS, or sham stimulation delivered to the right fusiform cortex during encoding of visual associations. Consistent with our hypothesis, improved AM performance was observed in the TACS group, while TDCS had no effect. However, TACS also resulted in improved correct rejection of never seen items, reduced false memory, and reduced forgetting, suggesting the effect may not be specific for AM processes. Overall, this work informs strategies for improving associative memory and suggests alternating current is more effective than direct current stimulation in some contexts.

Resource-constrained computing devices such as those used in IoTs require low-power, high performance, and small size to be enabled to operate efficiently. Resistive random access memory (ReRAM) is a promising technology for building novel in-memory computing architectures, due to its ability to perform storage and computation using the same physical element with low energy and high density. ReRAM-based search engines and neural network (NN) accelerators have grown significantly especially for IoT devices. In this paper, we propose a memristor-based voltage-resistance xnor (VR-XNOR) cell. The advantages of this cell are demonstrated through building a reconfigurable content-addressable memory (CAM) architecture that can support binary-CAM (BCAM) and ternary-CAM (TCAM) and enable approximate search operations. Moreover, the memristor-based VR-XNOR cell is utilized for binarized convolutional neural networks (CNN) with focus on the convolution operation. This is achieved by replacing the convolution module with XNOR-based filter banks. Simulations of the proposed architectures for search engine and feature extraction were carried out using VTEAM model in Cadence Virtuoso Analog Design Environment. The proposed filter bank architecture achieves a 1-ns extraction cycle time over ${N}$ filters and produces multiple output feature maps in a single processing cycle. The filter uses two memristor devices to realize each XNOR gate and has shown a significant reduction in the number of multiply-add operations.

Researchers have recently proposed that educated adults solve the simplest addition problems (e.g., 3 + 2) by an automatic counting procedure, challenging the long-held view that educated adults solve small additions by associative memory retrieval. We tested predications of a sum-counting model that assumes a procedure in which the 2 quantities represented by the operands are encoded and counted sequentially. Here, we presented the 2 operands sequentially (e.g., "3 +" first and then "2") and manipulated the preview time for the first operand (O1) and operator across 2 experiments (both n = 36); the O1 preview times were 1000 ms and 500 ms in Experiments 1 and 2, respectively. We measured response time (RT) from the presentation of the second operand (O2) and compared it with RT when both operands appeared simultaneously. Contrary to the sum-counting model, with sequential presentation, problems with the same O2 sizes (e.g., 3 + 2, 4 + 2) demonstrated significant RT differences across levels of O1, and the sum of the operands was a better RT predictor than O2 with both sequential and simultaneous displays. These results challenge a sum-counting model of the present data but are consistent with a memory retrieval theory. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Ternary content-addressable memory (TCAM) designed using static random-access memory (SRAM)-based field-programmable gate arrays (FPGAs) offers a promising lookup performance. However, the update process in a TCAM table poses significant challenges for efficiently employing SRAM-based TCAM. SRAM-based TCAM for FPGAs is designed using block RAM or distributed RAM resources in FPGAs. Such designs suspend search operations during an already high-latency update operation, rendering them infeasible in applications that require high-frequency updates. This paper presents a dynamically updateable energy- and resource-efficient TCAM design (DURE) based on FPGAs. DURE exploits the distributed RAM resources in FPGAs. More specifically, the lookup table RAMs (LUTRAMs) available in SLICEM resources are configured as quad-port RAM, which constitutes the basic memory (BM) block in the implementation of DURE. The contents of the TCAM table are divided into chunks of equal size and mapped onto the LUTRAMs of the proposed BM blocks. DURE implements dynamic updates by reconfiguring the LUTRAMs of only those BM blocks that are associated with the word being updated, thereby allowing search and update operations to be performed simultaneously. This achieves a lookup rate of 335 million lookups per second, with an update rate of 5.15 million updates per second on a $512\times36$ size TCAM on a Virtex-6 FPGA. Compared with the existing SRAM-based TCAMs, DURE has a smaller single-cycle search latency and achieves at least 2.5 times more energy efficiency and a 67% higher performance per area.

Amount of information exchange in modern technology era is increasing rapidly for accommodating both correlated and non-correlated contents. This can be best handled by filling the data as associations (linked data in a pre-defined order), rather than just indexing. Through association, available data are divided into related subsets and stored in a linked fashion. Content addressable memory (CAM) can reduce the time of locating the data that eliminates the following two inevitable issues: 1) The path of address location limits the memory access time and address-line size accounts for the depth (number of entries) in memory; 2) The search content under process is unknown, therefore even for an unavailable data, the full memory is searched. An analog to digital converter (ADC) acts as a primary interfacing device to communicate between the digital processor with real world applications. This work is a continuous study and improvement in various factors involved in the development of CAM and ADC.

Objectives It is well known that age differentially impacts aspects of long-term episodic memory (EM): Whereas a binding deficit indicates that older adults are less capable than younger adults to encode or retrieve associations between information (e.g., the pairing between two memoranda, such as lock – race), item memory is relatively intact (e.g., recognizing lock without its original pairing). Method We tested whether this deficit could be corrected by facilitating establishment of the bindings in working memory (WM) through adapting the semantic relatedness of studied pairs according to participants’ ongoing performance (Experiments 1 and 2). We also examined whether this was evident for the long-term retention of pairs that were not tested in WM (Experiment 2). Results The results revealed matched binding and item memory in WM and EM between age groups. Most importantly, older adults required increased semantic strength between word pairs to achieve similar performance to that of younger adults, regardless of whether pairs were immediately tested during the WM task. Discussion These findings indicate that relying on their superior semantic memory can correct the commonly exhibited profound deficit in binding memory in older age.

This paper presents a deep associative neural network (DANN) based on unsupervised representation learning for associative memory. In brain, the knowledge is learnt by associating different types of sensory data, such as image and voice. The associative memory models which imitate such a learning process have been studied for decades but with simpler architectures they fail to deal with large scale complex data as compared with deep neural networks. Therefore, we define a deep architecture consisting of a perception layer and hierarchical propagation layers. To learn the network parameters, we define a probabilistic model for the whole network inspired from unsupervised representation learning models. The model is optimized by a modified contrastive divergence algorithm with a novel iterated sampling process. After training, given a new data or corrupted data, the correct label or corrupted part is associated by the network. The DANN is able to achieve many machine learning problems, including not only classification, but also depicting the data given a label and recovering corrupted images. Experiments on MNIST digits and CIFAR-10 datasets demonstrate the learning capability of the proposed DANN.

Aging is often accompanied by associative memory changes, although their precise nature remains unclear. This study examines how recognition of item position in the context of associative memory differs between younger and older adults. Participants studied word pairs (A-B, C-D) and were later tested with intact (A-B), reversed (D-C), recombined (A-D), and recombined and reversed (B-C) pairs. When participants were instructed to respond "Old" to both intact and reversed pairs, and "New" to recombined, and recombined and reversed pairs, older adults showed worse recognition for recombined and reversed pairs relative to younger adults (Experiment 1). This finding also emerged when flexible retrieval demands were increased by asking participants to respond "Old"

Visual memory in children with high-functioning autism (HFA) is an area of debate. According to the few studies that have examined visual memory in children with autism, the memory profile appears to vary according to the memory process and type of stimuli, and contrasting results may be found. This study aims to analyze the visual memory profile of children with HFA. Fifteen children with HFA (mean age 9.6) and 15 typically developing children (TD; mean age 9.2) matched by chronological age and Leiter-R Brief IQ score took part in the study. Associative and recognition memory as well as visuospatial working memory were assessed. Impairments in face recognition and forward memory were found, whereas associative memory and shape recognition were preserved. The memory profile in children with Autism Spectrum Disorder (ASD) showed relatively stronger abilities in associative memory than in the other visual memory domains. The results support the hypothesis that the level of stimulus processing may influence memory performance by having a large impact on tasks and stimuli that require access to a semantic or global level of processing. In contrast to the TD population, children with ASD may have difficulty extracting underlying regularities from experiences and generalizing that information. Highlights Children with high-functioning autism (HFA) show preserved ability in associative memory and shape recognition. Face recognition appears to be a specific deficit in children with HFA. Associative memory appeared to be the strongest ability in the memory profile of children with autism spectrum disorder (ASD) and typically developing (TD) children.

Today’s computing systems use a huge amount of energy and time to process basic queries in database. A large part of it is spent in data movement between the memory and processing cores, owing to the limited cache capacity and memory bandwidth of traditional computers. In this paper, we propose a nonvolatile memory-based query accelerator, called NVQuery, which performs several basic query functions in memory including aggregation, prediction, bit-wise operations, join operations, as well as exact and nearest distance search queries. NVQuery is implemented on a content addressable memory and exploits the analog characteristic of nonvolatile memory in order to enable in-memory processing. To implement nearest distance search in memory, we introduce a novel bitline driving scheme to give weights to the indices of the bits during the search operation. To further improve the energy efficiency, our design supports configurable approximation by adaptively putting memory blocks under voltage overscaling. Our experimental evaluation shows that a NVQuery can provide $49.3{\boldsymbol \times }$ performance speedup and $32.9{\boldsymbol \times }$ energy savings as compared to running the same query on traditional processor. Approximation improves the energy-delay product (EDP) of NVQuery by $7.3{\boldsymbol \times }$ , while providing acceptable accuracy. In addition, NVQuery can achieve $30.1{\boldsymbol \times }$ EDP improvement as compared to the state-of-the-art query accelerators.

We (people) are memory machines. Our decision processes, emotions, and interactions with the world around us are based on and driven by associations to our memories. This natural association paradigm will become critical in future memory systems, namely, the key question will not be “How do I store more information?” but rather “Do I have the relevant information? How do I retrieve it?” The focus of this paper is to make a first step in this direction. We define and solve a very basic problem in associative retrieval. Given a word $W$ , the words in the memory, which are $t$ -associated with $W$ , are the words in the ball of radius $t$ around $W$ . In general, given a set of words, say $W$ , $X$ , and $Y$ , the words that are $t$ -associated with $\{W,X,Y\}$ are those in the memory that are within distance $t$ from all the three words. Our main goal is to study the maximum size of the $t$ -associated set as a function of the number of input words and the minimum distance of the words in memory—we call this value the uncertainty of an associative memory . In this paper, we consider the Hamming distance and derive the uncertainty of the associative memory that consists of all the binary vectors with an arbitrary number of input words. In addition, we study the retrieval problem, namely, how do we get the $t$ -associated set given the inputs? We note that this paradigm is a generalization of the sequences reconstruction problem that was proposed by Levenshtein (2001). In this model, a word is transmitted over multiple channels. A decoder receives all the channel outputs and decodes the transmitted word. Levenshtein computed the minimum number of channels that guarantee a successful decoder—this value happens to be the uncertainty of an associative memory with two input words.

Neural associative memory (AM) is one of the critical building blocks for cognitive computing systems. It memorizes (learns) and retrieves input data by information content itself. One of the key challenges of designing AM for intelligent devices is to expand memory capacity while using a minimal amount of hardware and energy resources. However, prior arts show that memory capacity increases slowly, i.e., in square root with the total number of synaptic weights. To tackle this problem, we propose a synapse model called recursive synaptic bit reuse, which enables near-linear scaling of memory capacity with total synaptic bits. Our model can also handle input data that are correlated more robustly than the conventional model. We evaluated our model in the context of Hopfield neural networks (HNNs) that contain 5–327-KB data storage for synaptic weights. Our model can increase the memory capacity of HNNs as large as $30\times $ over the conventional ones. The very large scale integration implementation of HNNs in 65 nm confirms that our proposed model can save up to $19\times $ area and up to $232\times $ energy dissipation as compared to the conventional model. These savings are expected to grow with the network size.

This paper deals with a generalized multidirectional associative memory neural network with time-varying coefficients and mixed delays. Some criteria are established for the existence and the global exponential stability of weighted pseudo almost-periodic solutions for the studied model. To do so, the exponential dichotomy theory, the contraction mapping fixed point theorem and the differential inequality techniques are used. Moreover, an example and its numerical simulation are employed to illustrate the main results.

Content-addressable memory (CAM) is a prominent hardware for high-speed lookup search, but consumes larger power. Traditional NOR and NAND match-line (ML) architectures suffer from a short circuit current path sharing and charge sharing respectively during precharge. The recently proposed precharge-free CAM suffers from high search delay and the subsequently proposed self-controlled precharge-free CAM suffers from high power consumption. This paper presents a hybrid self-controlled precharge-free (HSCPF) CAM architecture, which uses a novel charge control circuitry to reduce search delay as well as power consumption. The proposed and existing CAM ML architectures were developed using CMOS 45nm technology node with a supply voltage of 1 V. Simulation results show that the proposed HSCPF CAM-type ML design reduces power consumption and search delay effectively when compared to recent precharge-free CAM-type ML architectural designs.