Layer-Skipping Connections Improve the Effectiveness of Equilibrium Propagation on Layered Networks.
Ontology highlight
ABSTRACT:
SUBMITTER: Gammell J
PROVIDER: S-EPMC8165608 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Among the recent innovative technologies, memristor (memory-resistor) has attracted researchers attention as a fundamental computation element. It has been experimentally shown that memristive elements can emulate synaptic dynamics and are even capable of supporting spike timing dependent plasticity (STDP), an important adaptation rule that is gaining particular interest because of its simplicity and biological plausibility. The overall goal of this work is to provide a novel (theoretical) analog computing platform based on memristor devices and recurrent neural networks that exploits the memristor device physics to implement two variations of the backpropagation algorithm: recurrent backpropagation and equilibrium propagation. In the first learning technique, the use of memristor-based synaptic weights permits to propagate the error signals in the network by means of the nonlinear dynamics via an analog side network. This makes the processing non-digital and different from the current procedures. However, the necessity of a side analog network for the propagation of error derivatives makes this technique still highly biologically implausible. In order to solve this limitation, it is therefore proposed an alternative solution to the use of a side network by introducing a learning technique used for energy-based models: equilibrium propagation. Experimental results show that both approaches significantly outperform conventional architectures used for pattern reconstruction. Furthermore, due to the high suitability for VLSI implementation of the equilibrium propagation learning rule, additional results on the classification of the MNIST dataset are here reported.
| S-EPMC7105894 | biostudies-literature
Project description:Backpropagation (BP) has been used to train neural networks for many years, allowing them to solve a wide variety of tasks like image classification, speech recognition, and reinforcement learning tasks. But the biological plausibility of BP as a mechanism of neural learning has been questioned. Equilibrium Propagation (EP) has been proposed as a more biologically plausible alternative and achieves comparable accuracy on the CIFAR-10 image classification task. This study proposes the first EP-based reinforcement learning architecture: an Actor-Critic architecture with the actor network trained by EP. We show that this model can solve the basic control tasks often used as benchmarks for BP-based models. Interestingly, our trained model demonstrates more consistent high-reward behavior than a comparable model trained exclusively by BP.
| S-EPMC9446087 | biostudies-literature
Project description:Traditionally, astrocytes have been considered to couple via gap-junctions into a syncytium with only rudimentary spatial organization. However, this view is challenged by growing experimental evidence that astrocytes organize as a proper gap-junction mediated network with more complex region-dependent properties. On the other hand, the propagation range of intercellular calcium waves (ICW) within astrocyte populations is as well highly variable, depending on the brain region considered. This suggests that the variability of the topology of gap-junction couplings could play a role in the variability of the ICW propagation range. Since this hypothesis is very difficult to investigate with current experimental approaches, we explore it here using a biophysically realistic model of three-dimensional astrocyte networks in which we varied the topology of the astrocyte network, while keeping intracellular properties and spatial cell distribution and density constant. Computer simulations of the model suggest that changing the topology of the network is indeed sufficient to reproduce the distinct ranges of ICW propagation reported experimentally. Unexpectedly, our simulations also predict that sparse connectivity and restriction of gap-junction couplings to short distances should favor propagation while long-distance or dense connectivity should impair it. Altogether, our results provide support to recent experimental findings that point toward a significant functional role of the organization of gap-junction couplings into proper astroglial networks. Dynamic control of this topology by neurons and signaling molecules could thus constitute a new type of regulation of neuron-glia and glia-glia interactions.
| S-EPMC3997029 | biostudies-literature
Project description:In order to improve the long-term effectiveness of essential oil, a double-layered microcapsule was prepared using the inclusion encapsulation method in this study, with β-cyclodextrin as its inner layer and chitosan and sodium alginate as its outer layer. The optimized preparation process was obtained through the response surface method. The morphology, particle size, encapsulation efficiency, thermal stability and sustained release effect of the double-layered microcapsules were characterized. The microcapsules were spherical, with a particle size distribution between 2-6 μm, and had good thermal stability within 250 °C. Their encapsulation efficiency can be up to 80%, and it can continuously release the active ingredients of the essential oil under normal temperature and high temperature conditions for a long time. In order to further examine the application effect of the double-layered microcapsule, it was loaded onto the cotton fabric by the soak-roll method. The finished cotton fabric showed excellent washability and rubbing fastness. They can still maintain a light fragrance naturally for two months. The microcapsules prepared in this study can be potentially applied in sleep aid, antibacterial, mosquito prevention, food science and other related products.
| S-EPMC7465075 | biostudies-literature
Project description:Impulsive light excitation presents a powerful tool for investigating the interdependent structural and electronic responses in layered two-dimensional (2D) halide perovskites. However, detailed understanding of the nonlinear lattice dynamics in these soft hybrid materials remains limited. Here, we explicate the intrinsic strain propagation mechanisms in 2D perovskite single crystals using transient reflection spectroscopy. Ultrafast photoexcitation leads to the generation of strain pulses via thermoelastic (TE) stress and deformation potential (DP) interaction whence their detection proceed via Brillouin scattering. Using a two-temperature model together with strain wave propagation, we discern the TE and DP contributions in strain generation. Hot carrier cooling plays a dominant role in effecting the weak modulation amplitude. Out-of-plane lattice stiffness is reduced by the weak van der Waals bond between organic layers, resulting in a slow strain propagation velocity. Our findings inject fresh insights into the basic strain properties of layered perovskites critical for manipulating their functional properties for new applications.
| S-EPMC9481117 | biostudies-literature
Project description:Elastic network (EN) models have been widely used in recent years for describing protein dynamics, based on the premise that the motions naturally accessible to native structures are relevant to biological function. We posit that equilibrium motions also determine communication mechanisms inherent to the network architecture. To this end, we explore the stochastics of a discrete-time, discrete-state Markov process of information transfer across the network of residues. We measure the communication abilities of residue pairs in terms of hit and commute times, i.e., the number of steps it takes on an average to send and receive signals. Functionally active residues are found to possess enhanced communication propensities, evidenced by their short hit times. Furthermore, secondary structural elements emerge as efficient mediators of communication. The present findings provide us with insights on the topological basis of communication in proteins and design principles for efficient signal transduction. While hit/commute times are information-theoretic concepts, a central contribution of this work is to rigorously show that they have physical origins directly relevant to the equilibrium fluctuations of residues predicted by EN models.
| S-EPMC1988854 | biostudies-literature
Project description:The use of cell-laden hydrogels to engineer soft tissue has been emerging within the past years. Despite, several newly developed and sophisticated techniques to encapsulate different cell types the importance of vascularization of the engineered constructs is often underestimated. As a result, cell death within a construct leads to impaired function and inclusion of the implant. Here, we discuss the fabrication of hollow channels within hydrogels as a promising strategy to facilitate vascularization. Furthermore, we present an overview on the feasible use of removable spacers, 3D laser-, and planar processing strategies to create channels within hydrogels. The implementation of these structures promotes control over cell distribution and increases oxygen transport and nutrient supply in vitro. However, many studies lack the use of endothelial cells in their approaches leaving out an important factor to enhance vessel ingrowth and anastomosis formation upon implantation. In addition, the adequate endothelial cell type needs to be considered to make these approaches bridge the gap to in vivo applications.
| S-EPMC4231943 | biostudies-literature
Project description:This work investigates the mechanism of the Steady-State Visual Evoked Potential (SSVEP). One theory suggests that different SSVEP neural networks exist whose strongest response are located in different frequency bands. This theory is based on the fact that there are similar SSVEP frequency-amplitude response curves in these bands. Previous studies that employed simultaneous stimuli of different frequencies illustrated that the distribution of these networks were similar, but did not discuss the physical connection between them. By comparing the SSVEP power and distribution under a single-eye stimulus and a simultaneous, dual-eye stimulus, this work demonstrates that the distributions of different SSVEP neural networks are similar to each other and that there should be physical overlapping between them. According to the band-pass filter theory in a signal transferring channel, which we propose in this work for the first time, there are different amounts of neurons that are involved under repetitive stimuli of different frequencies and that the response intensity of each neuron is similar to each other so that the total response (i.e., the SSVEP) that is observed from the scalp is different.
| S-EPMC4782133 | biostudies-other
Project description:We introduce Equilibrium Propagation, a learning framework for energy-based models. It involves only one kind of neural computation, performed in both the first phase (when the prediction is made) and the second phase of training (after the target or prediction error is revealed). Although this algorithm computes the gradient of an objective function just like Backpropagation, it does not need a special computation or circuit for the second phase, where errors are implicitly propagated. Equilibrium Propagation shares similarities with Contrastive Hebbian Learning and Contrastive Divergence while solving the theoretical issues of both algorithms: our algorithm computes the gradient of a well-defined objective function. Because the objective function is defined in terms of local perturbations, the second phase of Equilibrium Propagation corresponds to only nudging the prediction (fixed point or stationary distribution) toward a configuration that reduces prediction error. In the case of a recurrent multi-layer supervised network, the output units are slightly nudged toward their target in the second phase, and the perturbation introduced at the output layer propagates backward in the hidden layers. We show that the signal "back-propagated" during this second phase corresponds to the propagation of error derivatives and encodes the gradient of the objective function, when the synaptic update corresponds to a standard form of spike-timing dependent plasticity. This work makes it more plausible that a mechanism similar to Backpropagation could be implemented by brains, since leaky integrator neural computation performs both inference and error back-propagation in our model. The only local difference between the two phases is whether synaptic changes are allowed or not. We also show experimentally that multi-layer recurrently connected networks with 1, 2, and 3 hidden layers can be trained by Equilibrium Propagation on the permutation-invariant MNIST task.
| S-EPMC5415673 | biostudies-literature
Project description:Nash equilibrium is widely present in various social disputes. As of now, in structured static populations, such as social networks, regular, and random graphs, the discussions on Nash equilibrium are quite limited. In a relatively stable static gaming network, a rational individual has to comprehensively consider all his/her opponents' strategies before they adopt a unified strategy. In this scenario, a new strategy equilibrium emerges in the system. We define this equilibrium as a local Nash equilibrium. In this paper, we present an explicit definition of the local Nash equilibrium for the two-strategy games in structured populations. Based on the definition, we investigate the condition that a system reaches the evolutionary stable state when the individuals play the Prisoner's dilemma and snow-drift game. The local Nash equilibrium provides a way to judge whether a gaming structured population reaches the evolutionary stable state on one hand. On the other hand, it can be used to predict whether cooperators can survive in a system long before the system reaches its evolutionary stable state for the Prisoner's dilemma game. Our work therefore provides a theoretical framework for understanding the evolutionary stable state in the gaming populations with static structures.
| S-EPMC4148661 | biostudies-literature