NMDA receptors are not required for pattern completion during associative memory recall

Pattern completion, the ability to retrieve complete memories initiated by subsets of external cues, has been a major focus of many computation models. A previously study reports that such pattern completion requires NMDA receptors in the hippocampus. However, such a claim was derived from a non-inducible gene knockout experiment in which the NMDA receptors were absent throughout all stages of memory processes as well as animal's adult life. This raises the critical question regarding whether the previously described results were truly resulting from the requirement of the NMDA receptors in retrieval. Here, we have examined the role of the NMDA receptors in pattern completion via inducible knockout of NMDA receptors limited to the memory retrieval stage. By using two independent mouse lines, we found that inducible knockout mice, lacking NMDA receptor in either forebrain or hippocampus CA1 region at the time of memory retrieval, exhibited normal recall of associative spatial reference memory regardless of whether retrievals took place under full-cue or partial-cue conditions. Moreover, systemic antagonism of NMDA receptor during retention tests also had no effect on full-cue or partial-cue recall of spatial water maze memories. Thus, both genetic and pharmacological experiments collectively demonstrate that pattern completion during spatial associative memory recall does not require the NMDA receptor in the hippocampus or forebrain.     

Associative memory design using overlapping decomposition and generalized brain-state-in-a-box neural networks

This paper is concerned with large scale associative memory design. A serious problem with neural associative memories is the quadratic growth of the number of interconnections with the problem size. An overlapping decomposition algorithm is proposed to attack this problem. Specifically, a pattern to be processed is decomposed into overlapping sub-patterns. Then, neural sub-networks are constructed that process the sub-patterns. An error correction algorithm operates on the outputs of each sub-network in order to correct the mismatches between sub-patterns that are obtained from the independent recall processes of individual sub-networks. The performance of the proposed large scale associative memory is illustrated using two-dimensional images. It is shown that the proposed method reduces the computing cost of the design of the associative memories compared with non-interconnected associative memories.     

DNA methylation mediates the discriminatory power of associative long-term memory in honeybees

Memory is created by several interlinked processes in the brain, some of which require long-term gene regulation. Epigenetic mechanisms are likely candidates for regulating memory-related genes. Among these, DNA methylation is known to be a long lasting genomic mark and may be involved in the establishment of long-term memory. Here we demonstrate that DNA methyltransferases, which induce and maintain DNA methylation, are involved in a particular aspect of associative long-term memory formation in honeybees, but are not required for short-term memory formation. While long-term memory strength itself was not affected by blocking DNA methyltransferases, odor specificity of the memory (memory discriminatory power) was. Conversely, perceptual discriminatory power was normal. These results suggest that different genetic pathways are involved in mediating the strength and discriminatory power of associative odor memories and provide, to our knowledge, the first indication that DNA methyltransferases are involved in stimulus-specific associative long-term memory formation.     

On associative memory

The information storing capacity of certain associative and auto-associative memories is calculated. For example, in a 100×100 matrix of 1 bit storage elements more than 6,500 bits can be stored associatively, and more than 688,000 bits in a 1,000×1,000 matrix. Asymptotically, the storage capacity of an associative memory increases proportionally to the number of storage elements. The usefulness of associative memories, as opposed to conventional listing memories, is discussed — especially in connection with brain modelling.     

Memories in context

Context-dependent associative memories are models that allow the retrieval of different vectorial responses given a same vectorial stimulus, depending on the context presented to the memory. The contextualization is obtained by doing the Kronecker product between two vectorial entries to the associative memory: the key stimulus and the context. These memories are able to display a wide variety of behaviors that range from all the basic operations of the logical calculus (including fuzzy logics) to the selective extraction of features from complex vectorial patterns. In the present contribution, we show that a context-dependent memory matrix stores a large amount of possible virtual associative memories, that awaken in the presence of a context. We show how the vectorial context allows a memory matrix to be representable in terms of its singular-value decomposition. We describe a neural interpretation of the model in which the Kronecker product is performed on the same neurons that sustain the memory. We explored, with numerical experiments, the reliability of chains of contextualized associations. In some cases, random disconnection produces the emergence of oscillatory behaviors of the system. Our results show that associative chains retain their performances for relatively large dimensions. Finally, we analyze the properties of some modules of context-dependent autoassociative memories inserted in recursive nets: the perceptual autoorganization in the presence of ambiguous inputs (e.g. the disambiguation of the Necker’s cube figure), the construction of intersection filters, and the feature extraction capabilities.     

Site-specific DNA-affinity chromatography of the lac repressor.

To test the feasibility of site-specific DNA-affinity chromatography, E. coli lac repressor was bound to an operator-containing DNA column, and in parallel to a non-operator DNA column. Salt gradient elution shows: 1) elution from non-operator DNA was near 250mM KCl or NaCl; interpretation of this result suggests the usefulness of the procedure for studying salt-dependence of DNA-protein affinities; 2) elution from operator-containing DNA was delayed (average elution = 1000mM salt), demonstrating a feasibility of site-specific DNA-affinity chromatography, if one provides a sufficiently favorable ratio of specific to non-specific DNA binding sites; 3) repressor eluted from operator-containing DNA over a very broad salt range, which may represent chromatography-generated repressor heterogeneity.     

On the interpretation of the observed linear free energy relationship in phosphate hydrolysis: a thorough computational study of phosphate diester hydrolysis in solution

The hydrolysis of phosphate esters is crucially important to biological systems, being involved in, among other things, signaling, energy transduction, biosynthesis, and the regulation of protein function. Despite this, there are many questions that remain unanswered in this important field, particularly with regard to the preferred mechanism of hydrolysis of phosphate esters, which can proceed through any of multiple pathways that are either associative or dissociative in nature. Previous comparisons of calculated and observed linear free energy relationships (LFERs) for phosphate monoester dianions with different leaving groups showed that the TS character gradually changes from associative to dissociative with the increasing acidity of the leaving group, while reproducing the experimental LFER. Here, we have generated ab initio potential energy surfaces for the hydrolysis of phosphate diesters in solution, with a variety of leaving groups. Once again, the reaction changes from a compact concerted pathway to one that is more expansive in character when the acidity of the leaving group increases. When such systems are examined in solution, it is essential to take into consideration the contribution of solute to the overall activation entropy, which remains a major computational challenge. The popular method of calculating the entropy using a quasi-harmonic approximation appears to markedly overestimate the configurational entropy for systems with multiple occupied energy wells. We introduce an improved restraint release approach for evaluating configurational entropies and apply this approach to our systems. We demonstrate that when this factor is taken into account, it is possible to reproduce the experimental LFER for this system with reasonable accuracy.     

Pseudo-relaxation learning algorithm for complex-valued associative memory

HAM (Hopfield Associative Memory) and BAM (Bidirectinal Associative Memory) are representative associative memories by neural networks. The storage capacity by the Hebb rule, which is often used, is extremely low. In order to improve it, some learning methods, for example, pseudo-inverse matrix learning and gradient descent learning, have been introduced. Oh introduced pseudo-relaxation learning algorithm to HAM and BAM. In order to accelerate it, Hattori proposed quick learning. Noest proposed CAM (Complex-valued Associative Memory), which is complex-valued HAM. The storage capacity of CAM by the Hebb rule is also extremely low. Pseudo-inverse matrix learning and gradient descent learning have already been generalized to CAM. In this paper, we apply pseudo-relaxation learning algorithm to CAM in order to improve the capacity.     

Bayesian information criterion for censored survival models

We investigate the Bayesian Information Criterion (BIC) for variable selection in models for censored survival data. Kass and Wasserman (1995, Journal of the American Statistical Association 90, 928-934) showed that BIC provides a close approximation to the Bayes factor when a unit-information prior on the parameter space is used. We propose a revision of the penalty term in BIC so that it is defined in terms of the number of uncensored events instead of the number of observations. For a simple censored data model, this revision results in a better approximation to the exact Bayes factor based on a conjugate unit-information prior. In the Cox proportional hazards regression model, we propose defining BIC in terms of the maximized partial likelihood. Using the number of deaths rather than the number of individuals in the BIC penalty term corresponds to a more realistic prior on the parameter space and is shown to improve predictive performance for assessing stroke risk in the Cardiovascular Health Study.     

Memory use in insect visual navigation

The navigational strategies that are used by foraging ants and bees to reach a goal are similar to those of birds and mammals. Species from all these groups use path integration and memories of visual landmarks to navigate through familiar terrain. Insects have far fewer neural resources than vertebrates, so data from insects might be useful in revealing the essential components of efficient navigation. Recent work on ants and bees has uncovered a major role for associative links between long-term memories. We emphasize the roles of these associations in the reliable recognition of visual landmarks and the reliable performance of learnt routes. It is unknown whether such associations also provide insects with a map-like representation of familiar terrain. We suggest, however, that landmarks act primarily as signposts that tell insects what particular action they need to perform, rather than telling them where they are.     

Abstract and associatively based representations in human sequence learning

We give an analysis of performance in an artificial neural network for which the claim had been made that it could learn abstract representations. Our argument is that this network is associative in nature, and cannot develop abstract representations. The network thus converges to a solution that is solely based on the statistical regularities of the training set. Inspired by human experiments that have shown that humans can engage in both associative (statistical) and abstract learning, we present a new, hybrid computational model that combines associative and more abstract, cognitive processes. To cross-validate the model we attempted to predict human behaviour in further experiments. One of these experiments reveals some evidence for the use of abstract representations, whereas the others provide evidence for associatively based performance. The predictions of the hybrid model stand in line with our empirical data.     

Sparse coding of sensory inputs

Several theoretical, computational, and experimental studies suggest that neurons encode sensory information using a small number of active neurons at any given point in time. This strategy, referred to as 'sparse coding', could possibly confer several advantages. First, it allows for increased storage capacity in associative memories; second, it makes the structure in natural signals explicit; third, it represents complex data in a way that is easier to read out at subsequent levels of processing; and fourth, it saves energy. Recent physiological recordings from sensory neurons have indicated that sparse coding could be a ubiquitous strategy employed in several different modalities across different organisms.     

π-π stacking tackled with density functional theory

Through comparison with ab initio reference data, we have evaluated the performance of various density functionals for describing pi-pi interactions as a function of the geometry between two stacked benzenes or benzene analogs, between two stacked DNA bases, and between two stacked Watson-Crick pairs. Our main purpose is to find a robust and computationally efficient density functional to be used specifically and only for describing pi-pi stacking interactions in DNA and other biological molecules in the framework of our recently developed QM/QM approach "QUILD". In line with previous studies, most standard density functionals recover, at best, only part of the favorable stacking interactions. An exception is the new KT1 functional, which correctly yields bound pi-stacked structures. Surprisingly, a similarly good performance is achieved with the computationally very robust and efficient local density approximation (LDA). Furthermore, we show that classical electrostatic interactions determine the shape and depth of the pi-pi stacking potential energy surface.     

The dynamics of sparse random networks

Recurrent neural networks with full symmetric connectivity have been extensively studied as associative memories and pattern recognition devices. However, there is considerable evidence that sparse, asymmetrically connected, mainly excitatory networks with broadly directed inhibition are more consistent with biological reality. In this paper, we use the technique of return maps to study the dynamics of random networks with sparse, asymmetric connectivity and nonspecific inhibition. These networks show three qualitatively different kinds of behavior: fixed points, cycles of low period, and extremely long cycles verging on aperiodicity. Using statistical arguments, we relate these behaviors to network parameters and present empirical evidence for the accuracy of this statistical model. The model, in turn, leads to methods for controlling the level of activity in networks. Studying random, untrained networks provides an understanding of the intrinsic dynamics of these systems. Such dynamics could provide a substrate for the much more complex behavior shown when synaptic modification is allowed.     

Effect of juvenile hormone on short-term olfactory memory in young honeybees (Apis mellifera).

Reliable retention of olfactory learning following a 1-trial classical conditioning of the proboscis extension reflex (PER) is not achieved in honeybees until they are 6-7 days old. Here we show that treatment of newly emerged honeybees with juvenile hormone (JH) has a profound effect on the maturation of short-term olfactory memory. JH-treated individuals display excellent short-term (1 h) memory of associative learning at times as early as 3 days of age and perform consistently better than untreated bees for at least the first week of their lives. By contrast, the retention of long-term (24 h) memory following a 3-trial conditioning of the PER is not significantly improved in JH-treated bees. Our study also shows that experience and (or) chemosensory activation are not essential to improve learning performance in olfactory tasks. The lack of accelerated development of long-term retention of olfactory memories in JH-treated honeybees is discussed in the context of neural circuits suspected to mediate memory formation and retrieval in the honeybee brain.     

Sleep supports selective retention of associative memories based on relevance for future utilization

An outstanding question is whether memory consolidation occurs passively or involves active processes that selectively stabilize memories based on future utility. Here, we differentially modulated the expected future relevance of two sets of picture-location associations after learning. Participants first studied two sets of picture-location associations. After a baseline memory test, they were instructed that only one set of associations would be retested after a 14-hour delay. For half of the participants, this test-retest delay contained a night of sleep; for the other half the delay included a normal working day. At retest, participants were re-instructed and against their expectations tested on both sets of associations. Our results show that post-learning instruction about subsequent relevance selectively improves memory retention for specific associative memories. This effect was sleep-dependent; it was present only in the group of subjects for which the test-retest delay contained sleep. Moreover, time spent asleep for participants in this sleep group correlated with retention of relevant but not irrelevant associations; participants who slept longer forgot fewer associations from the relevant category. In contrast, participants that did not sleep forgot more relevant than irrelevant associations across the test-retest delay. In summary, our results indicate that it is possible to modulate the retention of selected memories after learning with simple verbal instructions on their future relevance. The finding that this effect depends on sleep demonstrates this state's active role in memory consolidation and may have utility for educational settings.     

A New Rapid Protocol for Eyeblink Conditioning to Assess Cerebellar Motor Learning

Mice with spontaneous and induced mutations causing cerebellar phenotypes have provided key insights into how motor-related memories are stored in cerebellar circuits. Delayed eyeblink conditioning is a form of associative motor learning that depends on the cerebellum. However, neurochemical investigation of the underlying mechanisms has been hampered by the long training period (usually several days) required to establish conditioning. Here, we report a new rapid-training protocol that reliably induced delayed eyeblink conditioning within a single day. The associative memory formation depended on the expression of the δ2 glutamate receptor (GluD2) in cerebellar Purkinje cells. It lasted for several weeks, but could be erased by extinction sessions in a single day. In addition, using the rapid protocol, we found that eyeblink conditioning could be induced in juvenile mice at postnatal day 21, and that the Sindbis-virus-mediated expression of GluD2 could rescue the impaired eyeblink conditioning in GluD2-null mice in vivo.     

Helix-destabilization of deoxyribonucleic acid and poly[d(A-T).d(A-T)] by bovine seminal ribonuclease

A ribonuclease isolated earlier from bovine seminal plasma by DNA-affinity chromatography (Ramakrishnamurti, T. and Pandit, M.W. (1983) J. Chromatogr. 260, 216-222) has now been shown by thermal denaturation studies to destabilize the double-helical structure of DNA and poly[d(A-T).d(A-T)]. Thermal denaturation profiles of DNA in the presence of the protein are much more complicated due to the denaturation of protein itself in the temperature range over which DNA predominantly melts. The protein shows relatively stronger affinity towards denatured DNA as compared to native DNA. The action of micrococcal nuclease on DNA and its complexes with ribonuclease A and bovine seminal ribonuclease indicates that both of these proteins destabilize the double-helical structure of native DNA and thereby render the DNA more sensitive to the micrococcal nuclease.