Flexible Memory Networks

Networks of neurons in some brain areas are flexible enough to encode new memories quickly. Using a standard firing rate model of recurrent networks, we develop a theory of flexible memory networks. Our main results characterize networks having the maximal number of flexible memory patterns, given a constraint graph on the network’s connectivity matrix. Modulo a mild topological condition, we find a close connection between maximally flexible networks and rank 1 matrices. The topological condition is H ₁(X;ℤ)=0, where X is the clique complex associated to the network’s constraint graph; this condition is generically satisfied for large random networks that are not overly sparse. In order to prove our main results, we develop some matrix-theoretic tools and present them in a self-contained section independent of the neuroscience context.     

Modeling Reconsolidation in Kernel Associative Memory

Memory reconsolidation is a central process enabling adaptive memory and the perception of a constantly changing reality. It causes memories to be strengthened, weakened or changed following their recall. A computational model of memory reconsolidation is presented. Unlike Hopfield-type memory models, our model introduces an unbounded number of attractors that are updatable and can process real-valued, large, realistic stimuli. Our model replicates three characteristic effects of the reconsolidation process on human memory: increased association, extinction of fear memories, and the ability to track and follow gradually changing objects. In addition to this behavioral validation, a continuous time version of the reconsolidation model is introduced. This version extends average rate dynamic models of brain circuits exhibiting persistent activity to include adaptivity and an unbounded number of attractors.     

A Novel Implementation of a Flexible Robotic Fin Actuated by Shape Memory Alloy

In this paper,study of a novel flexible robotic-fin actuated by Shape Memory Alloy (SMA) is presented.The developed robotic fin is capable of implementing various 3-Dimensional (3D) motions,which plays an important role in robot propulsion and maneuverability.Firstly,the morphological and mechanics parameters of a real pectoral fin from a carp are investigated.Secondly,a detailed design of the flexible pectoral fin driven by SMA is presented according to the previous morphological and mechanics analyses.Thirdly,a simplified theoretical model on the SMA fin plate is derived.The thermodynamics of the SMA plate and the relationship between curvature and phase transformation are analyzed.Finally,several simulations and model experiments are conducted according to the previous analyses.The results of the experiments are useful for the control of the robotic fin.The experimental results reveal that the SMA actuated fin ray has a good actuating performance.     

Kernel stick-breaking processes

We propose a class of kernel stick-breaking processes for uncountable collections of dependent random probability measures. The process is constructed by first introducing an infinite sequence of random locations. Independent random probability measures and beta-distributed random weights are assigned to each location. Predictor-dependent random probability measures are then constructed by mixing over the locations, with stick-breaking probabilities expressed as a kernel multiplied by the beta weights. Some theoretical properties of the process are described, including a covariate-dependent prediction rule. A retrospective Markov chain Monte Carlo algorithm is developed for posterior computation, and the methods are illustrated using a simulated example and an epidemiological application.     

A Study on the Triglyceride Composition in Apricot Kernel Oil,Peach Kernel Oil and Plum Kernel Oil

The triglyceride composition of kernel oils from apricot, peach and plum was separated and analysed by reverse phase high performance liquid chromatography (RP-HPLC). The results showed that their oils predominantly contained OOO and OLO among 11 triglycerides.     

Kernel Abortion in Maize : II. Distribution of C among Kernel Carbohydrates

This study was designed to compare the uptake and distribution of ¹⁴C among fructose, glucose, sucrose, and starch in the cob, pedicel, and endosperm tissues of maize (Zea mays L.) kernels induced to abort by high temperature with those that develop normally. Kernels cultured in vitro at 30 and 35°C were transferred to [¹⁴C]sucrose media 10 days after pollination. Kernels cultured at 35°C aborted prior to the onset of linear dry matter accumulation. Significant uptake into the cob, pedicel, and endosperm of radioactivity associated with the soluble and starch fractions of the tissues was detected after 24 hours in culture on labeled media. After 8 days in culture on [¹⁴C]sucrose media, 48 and 40% of the radioactivity associated with the cob carbohydrates was found in the reducing sugars at 30 and 35°C, respectively. This indicates that some of the sucrose taken up by the cob tissue was cleaved to fructose and glucose in the cob. Of the total carbohydrates, a higher percentage of label was associated with sucrose and a lower percentage with fructose and glucose in pedicel tissue of kernels cultured at 35°C compared to kernels cultured at 30°C. These results indicate that sucrose was not cleaved to fructose and glucose as rapidly during the unloading process in the pedicel of kernels induced to abort by high temperature. Kernels cultured at 35°C had a much lower proportion of label associated with endosperm starch (29%) than did kernels cultured at 30°C (89%). Kernels cultured at 35°C had a correspondingly higher proportion of ¹⁴C in endosperm fructose, glucose, and sucrose. These results indicate that starch synthesis in the endosperm is strongly inhibited in kernels induced to abort by high temperature even though there is an adequate supply of sugar.     

Accelerating the Original Profile Kernel

One of the most accurate multi-class protein classification systems continues to be the profile-based SVM kernel introduced by the Leslie group. Unfortunately, its CPU requirements render it too slow for practical applications of large-scale classification tasks. Here, we introduce several software improvements that enable significant acceleration. Using various non-redundant data sets, we demonstrate that our new implementation reaches a maximal speed-up as high as 14-fold for calculating the same kernel matrix. Some predictions are over 200 times faster and render the kernel as possibly the top contender in a low ratio of speed/performance. Additionally, we explain how to parallelize various computations and provide an integrative program that reduces creating a production-quality classifier to a single program call. The new implementation is available as a Debian package under a free academic license and does not depend on commercial software. For non-Debian based distributions, the source package ships with a traditional Makefile-based installer. Download and installation instructions can be found at https://rostlab.org/owiki/index.php/Fast_Profile_Kernel. Bugs and other issues may be reported at https://rostlab.org/bugzilla3/enter_bug.cgi?product=fastprofkernel.     

Kernel estimates of dose response

A nonparametric method for analyzing quantal response data from an indirect bioassay experiment is proposed. Kernel estimates of the dose-response curve are used to develop approximate confidence intervals for (i) the optimal combination dose of a drug with therapeutic effects at low doses and toxic effects at high doses, and (ii) the lethal dose levels of a toxic chemical. This nonparametric procedure was implemented on real and simulated data. The confidence interval for problem (i) has high coverage probabilities when the dose-response curve is symmetric about the optima. However, the coverage probabilities are adversely affected by asymmetry about the optima and consequently are not reliable unless the sample sizes are large. The use of kernel estimators with higher-order kernels may alleviate this sensitivity to asymmetry. The confidence interval for problem (ii) has high coverage probabilities robust with respect to the shape or symmetry of the underlying dose-response curve.     

Memory time

In this issue of Neuron, MacDonald et al. describe hippocampal "time cells" that fire during specific delay periods as rats performed a memory task. Converging results in monkeys suggest that the hippocampus encodes episodes by signaling events in time.     

Memory Consolidation

Conscious memory for a new experience is initially dependent on information stored in both the hippocampus and neocortex. Systems consolidation is the process by which the hippocampus guides the reorganization of the information stored in the neocortex such that it eventually becomes independent of the hippocampus. Early evidence for systems consolidation was provided by studies of retrograde amnesia, which found that damage to the hippocampus-impaired memories formed in the recent past, but typically spared memories formed in the more remote past. Systems consolidation has been found to occur for both episodic and semantic memories and for both spatial and nonspatial memories, although empirical inconsistencies and theoretical disagreements remain about these issues. Recent work has begun to characterize the neural mechanisms that underlie the dialogue between the hippocampus and neocortex (e.g., “neural replay,” which occurs during sharp wave ripple activity). New work has also identified variables, such as the amount of preexisting knowledge, that affect the rate of consolidation. The increasing use of molecular genetic tools (e.g., optogenetics) can be expected to further improve understanding of the neural mechanisms underlying consolidation.Memory consolidation refers to the process by which a temporary, labile memory is transformed into a more stable, long-lasting form. Memory consolidation was first proposed in 1900 (Müller and Pilzecker 1900; Lechner et al. 1999) to account for the phenomenon of retroactive interference in humans, that is, the finding that learned material remains vulnerable to interference for a period of time after learning. Support for consolidation was already available in the facts of retrograde amnesia, especially as outlined in the earlier writings of Ribot (1881). The key observation was that recent memories are more vulnerable to injury or disease than remote memories, and the significance of this finding for consolidation was immediately appreciated.

In normal memory a process of organization is continually going on—a physical process of organization and a psychological process of repetition and association. In order that ideas may become a part of permanent memory, time must elapse for these processes of organization to be completed. (Burnham 1903, p. 132)

It is useful to note that the term consolidation has different contemporary usages that derive from the same historical sources. For example, the term is commonly used to describe events at the synaptic/cellular level (e.g., protein synthesis), which stabilize synaptic plasticity within hours after learning. In contrast, systems consolidation, which is the primary focus of this review, refers to gradual reorganization of the brain systems that support memory, a process that occurs within long-term memory itself (Squire and Alvarez 1995; Dudai and Morris 2000; Dudai 2012).Systems consolidation is typically, and accurately, described as the process by which memories, initially dependent on the hippocampus, are reorganized as time passes. By this process, the hippocampus gradually becomes less important for storage and retrieval, and a more permanent memory develops in distributed regions of the neocortex. The idea is not that memory is literally transferred from the hippocampus to the neocortex, for information is encoded in the neocortex as well as in hippocampus at the time of learning. The idea is that gradual changes in the neocortex, beginning at the time of learning, establish stable long-term memory by increasing the complexity, distribution, and connectivity among multiple cortical regions. Recent findings have enriched this perspective by emphasizing the dynamic nature of long-term memory (Dudai and Morris 2013). Memory is reconstructive and vulnerable to error, as in false remembering (Schacter and Dodson 2001). Also, under some conditions, long-term memory can transiently return to a labile state (and then gradually stabilize), a phenomenon termed reconsolidation (Nader et al. 2000; Sara 2000; Alberini 2005). In addition, the rate of consolidation can be influenced by the amount of prior knowledge that is available about the material to be learned (Tse et al. 2007; van Kesteren et al. 2012).Neurocomputational models of consolidation (McClelland et al. 1995; McClelland 2013) describe how the acquisition of new knowledge might proceed and suggest a purpose for consolidation. As originally described, elements of information are first stored in a fast-learning hippocampal system. This information directs the training of a “slow learning” neocortex, whereby the hippocampus gradually guides the development of connections between the multiple cortical regions that are active at the time of learning and that represent the memory. Training of the neocortex by the hippocampus (termed “interleaved” training) allows new information to be assimilated into neocortical networks with a minimum of interference. In simulations (McClelland et al. 1995), rapid learning of new information, which was inconsistent with prior knowledge, was shown to cause interference and disrupt previously established representations (“catastrophic interference”). The gradual incorporation of information into the neocortex during consolidation avoids this problem. In a recent revision of this framework (McClelland 2013), neocortical learning is characterized, not so much as fast or slow, but as dependent on prior knowledge. If the information to be learned is consistent with prior knowledge, neocortical learning can be more rapid.This review considers several types of evidence that illuminate the nature of the consolidation process: studies of retrograde amnesia in memory-impaired patients, studies of healthy volunteers with neuroimaging, studies of sleep and memory, studies of experimental animals, both with lesions or other interventions, and studies that track neural activity as time passes after learning.     

Memory.

The key interrelated issues in the neurobiology of memory are to identify the neural circuitries essential for memory formation, localize sites of memory storage and analyze mechanisms of memory formation, storage and retrieval. Several circuits have now been identified in vertebrates and researchers are investigating their properties, in particular the role of glutamate receptors and long-term potentiation, in memory formation. Invertebrate preparations continue to be of value and recent studies suggest that changes in gene expression and protein synthesis may be important in long-term sensitization.     

In Memory

Abstract

Dr. Mathias P. Mertes, well-known medicinal chemist, member of the Editorial Advisory Board of Nucleosides & Nucleotides, and a good friend, died unexpectedly earlier this year. An obituary written by Drs. Gary Grunewald and Robert Hanzlik of the University of Kansas is presented below followed by a list of the publications of Dr. Mertes. Many of us have pleasant memories of stimulating interactions with Matt. We can hope that the spirit and enthusiasm with which Matt approached science has rubbed off on some of us. He will be missed.