Sparse Associative Memory.

It is still unknown how associative biological memories operate. Hopfield networks are popular models of associative memory, but they suffer from spurious memories and low efficiency. Here, we present a new model of an associative memory that overcomes these deficiencies. We call this model sparse associative memory (SAM) because it is based on sparse projections from neural patterns to pattern-specific neurons. These sparse projections have been shown to be sufficient to uniquely encode a neural pattern. Based on this principle, we investigate theoretically and in simulation our SAM model, which turns out to have high memory efficiency and a vanishingly small probability of spurious memories. This model may serve as a basic building block of brain functions involving associative memory.

Effects of Cigarette Smoking on Plasma Concentration of the Appetite-Regulating Peptide Ghrelin.

BACKGROUND: Weight gain is a common but only a partially understood consequence of smoking cessation. Existing data suggest modulating effects of the orexigenic peptide ghrelin on food intake. The aim of the present study was to investigate the effect of tobacco withdrawal on plasma concentration of acetylated and total ghrelin. METHODS: Fifty four normal-weighted smokers and 30 non-smoking healthy controls were enrolled in our study. Concentrations of acetylated and total ghrelin were measured in blood plasma drawn two hours after a standardized meal and three hours after the smokers smoked their last cigarette. The severity of tobacco addiction was assessed based on cotinine plasma concentration, the Fagerström Test for Nicotine Dependence (FTND) and the number of cigarettes smoked per day. RESULTS: The plasma concentration of acetylated ghrelin, but not total ghrelin, was significantly higher in smokers than in non-smokers. Moreover, we found significant negative correlations between acetylated ghrelin and all measures of the severity of nicotine dependence. CONCLUSIONS: Early abstinence from tobacco smoking seems to be associated with increased plasma concentration of the orexigenic peptide acetylated ghrelin. This could be one reason for increased food craving during nicotine withdrawal and subsequent weight gain. Smokers might compensate these effects by increasing tobacco intake.

Dynamical movement primitives: learning attractor models for motor behaviors.

Nonlinear dynamical systems have been used in many disciplines to model complex behaviors, including biological motor control, robotics, perception, economics, traffic prediction, and neuroscience. While often the unexpected emergent behavior of nonlinear systems is the focus of investigations, it is of equal importance to create goal-directed behavior (e.g., stable locomotion from a system of coupled oscillators under perceptual guidance). Modeling goal-directed behavior with nonlinear systems is, however, rather difficult due to the parameter sensitivity of these systems, their complex phase transitions in response to subtle parameter changes, and the difficulty of analyzing and predicting their long-term behavior; intuition and time-consuming parameter tuning play a major role. This letter presents and reviews dynamical movement primitives, a line of research for modeling attractor behaviors of autonomous nonlinear dynamical systems with the help of statistical learning techniques. The essence of our approach is to start with a simple dynamical system, such as a set of linear differential equations, and transform those into a weakly nonlinear system with prescribed attractor dynamics by means of a learnable autonomous forcing term. Both point attractors and limit cycle attractors of almost arbitrary complexity can be generated. We explain the design principle of our approach and evaluate its properties in several example applications in motor control and robotics.

Target switching in curved human arm movements is predicted by changing a single control parameter.

Straight-line movements have been studied extensively in the human motor-control literature, but little is known about how to generate curved movements and how to adjust them in a dynamic environment. The present work studied, for the first time to my knowledge, how humans adjust curved hand movements to a target that switches location. Subjects (n = 8) sat in front of a drawing tablet and looked at a screen. They moved a cursor on a curved trajectory (spiral or oval shaped) toward a goal point. In half of the trials, this goal switched 200 ms after movement onset to either one of two alternative positions, and subjects smoothly adjusted their movements to the new goal. To explain this adjustment, we compared three computational models: a superposition of curved and minimum-jerk movements (Flash and Henis in J Cogn Neurosci 3(3):220-230, 1991), Vector Planning (Gordon et al. in Exp Brain Res 99(1):97-111, 1994) adapted to curved movements (Rescale), and a nonlinear dynamical system, which could generate arbitrarily curved smooth movements and had a point attractor at the goal. For each model, we predicted the trajectory adjustment to the target switch by changing only the goal position in the model. As result, the dynamical model could explain the observed switch behavior significantly better than the two alternative models (spiral: P = 0.0002 vs. Flash, P = 0.002 vs. Rescale; oval: P = 0.04 vs. Flash; P values obtained from Wilcoxon test on R (2) values). We conclude that generalizing arbitrary hand trajectories to new targets may be explained by switching a single control command, without the need to re-plan or re-optimize the whole movement or superimpose movements.

Impact of network topology on self-organized criticality.

The general mechanisms behind self-organized criticality (SOC) are still unknown. Several microscopic and mean-field theory approaches have been suggested, but they do not explain the dependence of the exponents on the underlying network topology of the SOC system. Here, we first report the phenomena that in the Bak-Tang-Wiesenfeld (BTW) model, sites inside an avalanche area largely return to their original state after the passing of an avalanche, forming, effectively, critically arranged clusters of sites. Then, we hypothesize that SOC relies on the formation process of these clusters, and present a model of such formation. For low-dimensional networks, we show theoretically and in simulation that the exponent of the cluster-size distribution is proportional to the ratio of the fractal dimension of the cluster boundary and the dimensionality of the network. For the BTW model, in our simulations, the exponent of the avalanche-area distribution matched approximately our prediction based on this ratio for two-dimensional networks, but deviated for higher dimensions. We hypothesize a transition from cluster formation to the mean-field theory process with increasing dimensionality. This work sheds light onto the mechanisms behind SOC, particularly, the impact of the network topology.

Optimization by Self-Organized Criticality.

Self-organized criticality (SOC) is a phenomenon observed in certain complex systems of multiple interacting components, e.g., neural networks, forest fires, and power grids, that produce power-law distributed avalanche sizes. Here, we report the surprising result that the avalanches from an SOC process can be used to solve non-convex optimization problems. To generate avalanches, we use the Abelian sandpile model on a graph that mirrors the graph of the optimization problem. For optimization, we map the avalanche areas onto search patterns for optimization, while the SOC process receives no feedback from the optimization itself. The resulting method can be applied without parameter tuning to a wide range of optimization problems, as demonstrated on three problems: finding the ground-state of an Ising spin glass, graph coloring, and image segmentation. We find that SOC search is more efficient compared to other random search methods, including simulated annealing, and unlike annealing, it is parameter free, thereby eliminating the time-consuming requirement to tune an annealing temperature schedule.

Perception through visuomotor anticipation in a mobile robot.

Several scientists suggested that certain perceptual qualities are based on sensorimotor anticipation: for example, the softness of a sponge is perceived by anticipating the sensations resulting from a grasping movement. For the perception of spatial arrangements, this article demonstrates that this concept can be realized in a mobile robot. The robot first learned to predict how its visual input changes under movement commands. With this ability, two perceptual tasks could be solved: judging the distance to an obstacle in front by 'mentally' simulating a movement toward the obstacle, and recognizing a dead end by simulating either an obstacle-avoidance algorithm or a recursive search for an exit. A simulated movement contained a series of prediction steps. In each step, a multilayer perceptron anticipated the next image, which, however, became increasingly noisy. To denoise an image, it was split into patches, and each patch was projected onto a manifold obtained by modelling the density of the distribution of training patches with a mixture of Gaussian functions.

B-Myb promotes S-phase independently of its sequence-specific DNA binding activity and interacts with polymerase delta-interacting protein 1 (Pdip1).

B-Myb is a highly conserved member of the Myb transcription factor family, which plays an essential role in cell cycle progression by regulating the transcription of genes at the G 2/M-phase boundary. The role of B-Myb in other parts of the cell cycle is less well-understood. By employing siRNA-mediated silencing of B-Myb expression, we found that B-Myb is required for efficient entry into S-phase. Surprisingly, a B-Myb mutant that lacks sequence-specific DNA-binding activity and is unable to activate transcription of B-Myb target genes is able to rescue the S-phase defect observed after B-Myb knockdown. Moreover, we have identified polymerase delta-interacting protein 1 (Pdip1), a BTB domain protein known to bind to the DNA replication and repair factor PCNA as a novel B-Myb interaction partner. We have shown that Pdip1 is able to interact with B-Myb and PCNA simultaneously. In addition, we found that a fraction of endogenous B-Myb can be co-precipitated via PCNA, suggesting that B-Myb might be involved in processes related to DNA replication or repair. Taken together, our work suggests a novel role for B-Myb in S-phase that appears to be independent of its sequence-specific DNA-binding activity and its ability to stimulate the expression of bona fide B-Myb target genes.

Computational Models for Neuromuscular Function.

Computational models of the neuromuscular system hold the potential to allow us to reach a deeper understanding of neuromuscular function and clinical rehabilitation by complementing experimentation. By serving as a means to distill and explore specific hypotheses, computational models emerge from prior experimental data and motivate future experimental work. Here we review computational tools used to understand neuromuscular function including musculoskeletal modeling, machine learning, control theory, and statistical model analysis. We conclude that these tools, when used in combination, have the potential to further our understanding of neuromuscular function by serving as a rigorous means to test scientific hypotheses in ways that complement and leverage experimental data.

Learning visuomotor transformations for gaze-control and grasping.

For reaching to and grasping of an object, visual information about the object must be transformed into motor or postural commands for the arm and hand. In this paper, we present a robot model for visually guided reaching and grasping. The model mimics two alternative processing pathways for grasping, which are also likely to coexist in the human brain. The first pathway directly uses the retinal activation to encode the target position. In the second pathway, a saccade controller makes the eyes (cameras) focus on the target, and the gaze direction is used instead as positional input. For both pathways, an arm controller transforms information on the target's position and orientation into an arm posture suitable for grasping. For the training of the saccade controller, we suggest a novel staged learning method which does not require a teacher that provides the necessary motor commands. The arm controller uses unsupervised learning: it is based on a density model of the sensor and the motor data. Using this density, a mapping is achieved by completing a partially given sensorimotor pattern. The controller can cope with the ambiguity in having a set of redundant arm postures for a given target. The combined model of saccade and arm controller was able to fixate and grasp an elongated object with arbitrary orientation and at arbitrary position on a table in 94% of trials.