Unsupervised Hebbian Learning Research Articles

Performance of biologically grounded models of the early visual system on standard object recognition tasks

Computational neuroscience models of vision and neural network models for object recognition are often framed by different research agendas. Computational neuroscience mainly aims at replicating experimental data, while (artificial) neural networks target high performance on classification tasks. However, we propose that models of vision should be validated on object recognition tasks. At some point, mechanisms of realistic neuro-computational models of the visual cortex have to convince in object recognition as well.In order to foster this idea, we report the recognition accuracy for two different neuro-computational models of the visual cortex on several object recognition datasets. The models were trained using unsupervised Hebbian learning rules on natural scene inputs for the emergence of receptive fields comparable to their biological counterpart. We assume that the emerged receptive fields result in a general codebook of features, which should be applicable to a variety of visual scenes.We report the performances on datasets with different levels of difficulty, ranging from the simple MNIST to the more complex CIFAR-10 or ETH-80. We found that both networks show good results on simple digit recognition, comparable with previously published biologically plausible models. We also observed that our deeper layer neurons provide for naturalistic datasets a better recognition codebook. As for most datasets, recognition results of biologically grounded models are not available yet, our results provide a broad basis of performance values to compare methodologically similar models.

Memristor-based circuit implementation of Competitive Neural Network based on online unsupervised Hebbian learning rule for pattern recognition

Memristor-based circuit implementation of Competitive Neural Network based on online unsupervised Hebbian learning rule for pattern recognition

Nonlinear Dendritic Coincidence Detection for Supervised Learning.

Cortical pyramidal neurons have a complex dendritic anatomy, whose function is an active research field. In particular, the segregation between its soma and the apical dendritic tree is believed to play an active role in processing feed-forward sensory information and top-down or feedback signals. In this work, we use a simple two-compartment model accounting for the nonlinear interactions between basal and apical input streams and show that standard unsupervised Hebbian learning rules in the basal compartment allow the neuron to align the feed-forward basal input with the top-down target signal received by the apical compartment. We show that this learning process, termed coincidence detection, is robust against strong distractions in the basal input space and demonstrate its effectiveness in a linear classification task.

Erratum to “Bidirectional Associative Memories: Unsupervised Hebbian Learning to Bidirectional Backpropagation”

The energy velocity <xref ref-type="disp-formula" rid="deqn1-deqn2" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">equation (41)</xref> in <xref ref-type="bibr" rid="ref1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">[1]</xref> has two missing prime symbols “’” that indicate the derivative of the primed activation function with respect to its argument. The correct equation for the time derivative of the ABAM energy function <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$E$ </tex-math></inline-formula> should read as follows:

Bidirectional Associative Memories: Unsupervised Hebbian Learning to Bidirectional Backpropagation

Bidirectional associative memories (BAMs) pass neural signals forward and backward through the same web of synapses. Earlier BAMs had no hidden neurons and did not use supervised learning. They tuned their synaptic weights with unsupervised Hebbian or competitive learning. Two-layer feedback BAMs always converge to fixed-point equilibria for threshold or threshold-like neurons. Every rectangular connection matrix is bidirectionally stable. These simpler BAMs extend to arbitrary hidden layers with supervised learning if the resulting bidirectional backpropagation algorithm uses the proper layer likelihood in the forward and backward directions. Bidirectional backpropagation lets users run deep classifiers and regressors in reverse as well as forward. Bidirectional training exploits pattern and synaptic information that forward-only running ignores.

Detecting Generic Music Features with Single Layer Feedforward Network using Unsupervised Hebbian Computation

In this work, the authors extract information on distinct baseline features from a popular open-source music corpus and explore new recognition techniques by applying unsupervised Hebbian learning techniques on our single-layer neural network using the same dataset. They show the detailed empirical findings to simulate how such an algorithm can help a single layer feedforward network in training for music feature learning as patterns. The unsupervised training algorithm enhances the proposed neural network to achieve an accuracy of 90.36% for successful music feature detection. For comparative analysis against similar tasks, they put their results with the likes of several previous benchmark works. They further discuss the limitations and thorough error analysis of the work. They hope to discover and gather new information about this particular classification technique and performance, also further understand future potential directions that could improve the art of computational music feature recognition.

Sparse coding with a somato-dendritic rule

Cortical neurons are silent most of the time: sparse activity enables low-energy computation in the brain, and promises to do the same in neuromorphic hardware. Beyond power efficiency, sparse codes have favourable properties for associative learning, as they can store more information than local codes but are easier to read out than dense codes. Auto-encoders with a sparse constraint can learn sparse codes, and so can single-layer networks that combine recurrent inhibition with unsupervised Hebbian learning. But the latter usually require fast homeostatic plasticity, which could lead to catastrophic forgetting in embodied agents that learn continuously. Here we set out to explore whether plasticity at recurrent inhibitory synapses could take up that role instead, regulating both the population sparseness and the firing rates of individual neurons. We put the idea to the test in a network that employs compartmentalised inputs to solve the task: rate-based dendritic compartments integrate the feedforward input, while spiking integrate-and-fire somas compete through recurrent inhibition. A somato-dendritic learning rule allows somatic inhibition to modulate nonlinear Hebbian learning in the dendrites. Trained on MNIST digits and natural images, the network discovers independent components that form a sparse encoding of the input and support linear decoding. These findings confirm that intrinsic homeostatic plasticity is not strictly required for regulating sparseness: inhibitory synaptic plasticity can have the same effect. Our work illustrates the usefulness of compartmentalised inputs, and makes the case for moving beyond point neuron models in artificial spiking neural networks.

Supervised and unsupervised learning in processing myographic patterns

Synaptic connections in neocortex are assumed to be formed by a self-organizing process leading to emergence of the so-called self-organized maps (SOMs). Formation of SOMs is based on the unsupervised Hebbian learning rule, according to which the weight of a synaptic connection depends on the co-activation of pre- and postsynaptic neurons. On the other hand, a variety of human-machine interfaces employ supervised learning based on the correction of synaptic weights in proportion to the network error. But this learning rule has not been verified as biologically relevant. In our work, we study artificial neural networks (ANN) that classify hand gestures through electromyographic recordings (EMG). We use eight-channel electromyographic (EMG) signals acquired by a Thalmic labs Myo device as an input to a multilayer perceptron (MLP) and Kohonen’s SOM. We compare supervised (MLP) and unsupervised (SOM) learning in the task of EMG-classification. The median value of recognition fidelity (F-measure) for SOM-based recognition is F = 0.87 and for MLP classification F = 0.96. Also we reveal strong correlation between F-measures for classification EMG patterns of 37 subjects by MLP and SOM. For estimation of clustering quality of SOM we introduce two indexes: the intra-cluster index describing the cluster “compactness” and the inter-cluster index measuring the degree of cluster overlapping. There are strong correlations between F-measures for SOM classification and introduced indexes. Differences in the significance level of the correlations suggest that the classification with SOM is more negatively affected by overlapping clusters than their large size.

Hebbian learning for online prediction, neural recall and classical conditioning of anthropomimetic robot arm motions

Classical conditioning plays a vital role in learning in every mammal. It is based on unsupervised neural learning embodied in a physical body that is in continuous interaction with the environment. Embedding the hierarchical temporal memory (HTM) in the closed loop of the sensorimotor space of a Myorobotics tendon-driven robotic arm we demonstrate learning, prediction and control of biomimetic body motions. Experiments finally lead to conditioned reactions in natural interaction with a human partner.The HTM is able to learn arm movements generated by interaction with a human partner in a short time. It predicts future positions in different time scales up to seconds in advance. Closing the loop we utilize HTM predictions for motor control. Hereby learned motions are recalled from synaptic connections proactively continuing motion execution. Association, prediction and control are required by the HTM for conditioning according to Pavlov: neutral stimuli get associated with motions, and after learning sensor impulses can trigger single arm lifting motions. Hereby, both the motions and the stimuli are learned from the environment and get associated efficiently. We can demonstrate high biological plausibility as for example even input variations result in similar variations in the action output.The robotic system consisting of biologically-derived hardware and software components utilizes only unsupervised Hebbian learning to act autonomously. Learning is executed in real time, can handle natural variations of human motions and takes morphologically plausible sensor input into account. The setup is fully scalable due to its modularity. Hereby, novel applications for the HTM are opened: it can be used in musculoskeletal robot control scenarios and robots being able to interactively learn from human partners and the environment.

Attractor Dynamics in Networks with Learning Rules Inferred from In Vivo Data

The attractor neural network scenario is a popular scenario for memory storage in the association cortex, but there is still a large gap between models based on this scenario and experimental data. We study a recurrent network model in which both learning rules and distribution of stored patterns are inferred from distributions of visual responses for novel and familiar images in the inferior temporal cortex (ITC). Unlike classical attractor neural network models, our model exhibits graded activity in retrieval states, with distributions of firing rates that are close to lognormal. Inferred learning rules are close to maximizing the number of stored patterns within a family of unsupervised Hebbian learning rules, suggesting that learning rules in ITC are optimized to store a large number of attractor states. Finally, we show that there exist two types of retrieval states: one in which firing rates are constant in time and another in which firing rates fluctuate chaotically.

Unsupervised Hebbian learning experimentally realized with analogue memristive crossbar arrays

Conventional transistor electronics are reaching their limits in terms of scalability, power dissipation, and the underlying Boolean system architecture. To overcome this obstacle neuromorphic analogue systems are recently highly investigated. Particularly, the use of memristive devices in VLSI analogue concepts provides a promising pathway to realize novel bio-inspired computing architectures, which are able to unravel the foreseen difficulties of traditional electronics. Currently, a variety of materials and device structures are being studied along with novel computing schemes to make use of the attractive features of memristive devices for neuromorphic computing. However, a number of obstacles still have to be overcome to cast memristive devices into hardware systems. Most important is a physical implementation of memristive devices, which can cope with the high complexity of neural networks. This includes the integration of analogue and electroforming-free memristive devices into crossbar structures with no additional electronic components, such as selector devices. Here, an unsupervised, bio-motivated Hebbian based learning platform for visual pattern recognition is presented. The heart of the system is a crossbar array (16 × 16) which consists of selector-free and forming-free (non-filamentary) memristive devices, which exhibit analogue I-V characteristics.

Multidimensional Hebbian Learning With Temporal Coding in Neocognitron Visual Recognition

Previous work on Fuskushima’s neocognitron neural network model has mostly focused on binary character recognition, but grayscale images are preferable to binary images in real applications. Image classes of grayscale objects can be formed with an unsupervised learning process using the neocognitron neural network model. Since Hebbian learning as one that uses a time dependent and strongly interactive mechanism to increase synaptic efficiency as a correlation function between presynaptic and postsynaptic activity, and principal component analysis (PCA) is used in neuroscience extensively. An analogy is shown between unsupervised Hebbian learning and PCA when applied to the neocognitron model. The Hebbian learning theory is extended taking into consideration of temporal coding. Successful computer simulation models for grayscale object recognition are discussed. This paper is the first to carry out quantitative analysis of the neuron responses for gray scale image classification.

An Adaptive Signal Processing Framework for PV Power Maximization

This paper discusses the possibility of using adaptive signal processing techniques for maximum power point tracking controllers, in order to extract peak power from individual photovoltaic modules. A new technique grounded on unsupervised Hebbian learning theory (maximum eigenvector of the output power) is presented, which works on-online and is capable of operating without a desired response. Important modifications were made to the generic Hebbian adaptation to accommodate the intrinsic feedback loop between the controller and the plant. Analytic derivation of the new update rule is presented, as well as stability analysis by means of Lyapunov theory. Simulation results showing its effectiveness are presented, as well as experimental results.

Event signature extraction and traffic modeling in WSNs

Motivated by earlier work on adaptive event forecasting, this paper proposes an iterative event signature extraction method for wireless sensor networks, and a probabilistic approach to model non i.i.d. (independent and identically distributed) aperiodic traffic. The model is validated on well described (collection included) real world measurements of various types, and it is then used to demonstrate the effectiveness of the proposed signature extraction method to support reliable event forecasting. Our scheme can continuously keep the event signature database low on artifacts, dynamically estimate the number of sequences and extract the signatures from noisy, overlapped events. The proposed solution is inspired by unsupervised competitive Hebbian learning used in self-organizing Kohonen maps. We evaluate the proposed solution analytically, but also empirically, by means of simulations.

Functional model of biological neural networks

A functional model of biological neural networks, called temporal hierarchical probabilistic associative memory (THPAM), is proposed in this paper. THPAM comprises functional models of dendritic trees for encoding inputs to neurons, a first type of neuron for generating spike trains, a second type of neuron for generating graded signals to modulate neurons of the first type, supervised and unsupervised Hebbian learning mechanisms for easy learning and retrieving, an arrangement of dendritic trees for maximizing generalization, hardwiring for rotation-translation-scaling invariance, and feedback connections with different delay durations for neurons to make full use of present and past informations generated by neurons in the same and higher layers. These functional models and their processing operations have many functions of biological neural networks that have not been achieved by other models in the open literature and provide logically coherent answers to many long-standing neuroscientific questions. However, biological justifications of these functional models and their processing operations are required for THPAM to qualify as a macroscopic model (or low-order approximate) of biological neural networks.

Adaptive-Wave Alternative for the Black-Scholes Option Pricing Model

A nonlinear wave alternative for the standard Black-Scholes option-pricing model is presented. The adaptive-wave model, representing controlled Brownian behavior of financial markets, is formally defined by adaptive nonlinear Schrödinger (NLS) equations, defining the option-pricing wave function in terms of the stock price and time. The model includes two parameters: volatility (playing the role of dispersion frequency coefficient), which can be either fixed or stochastic, and adaptive market potential that depends on the interest rate. The wave function represents quantum probability amplitude, whose absolute square is probability density function. Four types of analytical solutions of the NLS equation are provided in terms of Jacobi elliptic functions, all starting from de Broglie’s plane-wave packet associated with the free quantum-mechanical particle. The best agreement with the Black-Scholes model shows the adaptive shock-wave NLS-solution, which can be efficiently combined with adaptive solitary-wave NLS-solution. Adjustable ‘weights’ of the adaptive market-heat potential are estimated using either unsupervised Hebbian learning or supervised Levenberg–Marquardt algorithm. In the case of stochastic volatility, it is itself represented by the wave function, so we come to the so-called Manakov system of two coupled NLS equations (that admits closed-form solutions), with the common adaptive market potential, which defines a bidirectional spatio-temporal associative memory.

Learning with incomplete information in the committee machine

We study the problem of learning with incomplete information in a student-teacher setup for the committee machine. The learning algorithm combines unsupervised Hebbian learning of a series of associations with a delayed reinforcement step, in which the set of previously learnt associations is partly and indiscriminately unlearnt, to an extent that depends on the success rate of the student on these previously learnt associations. The relevant learning parameter lambda represents the strength of Hebbian learning. A coarse-grained analysis of the system yields a set of differential equations for overlaps of student and teacher weight vectors, whose solutions provide a complete description of the learning behavior. It reveals complicated dynamics showing that perfect generalization can be obtained if the learning parameter exceeds a threshold lambda ( c ), and if the initial value of the overlap between student and teacher weights is non-zero. In case of convergence, the generalization error exhibits a power law decay as a function of the number of examples used in training, with an exponent that depends on the parameter lambda. An investigation of the system flow in a subspace with broken permutation symmetry between hidden units reveals a bifurcation point lambda* above which perfect generalization does not depend on initial conditions. Finally, we demonstrate that cases of a complexity mismatch between student and teacher are optimally resolved in the sense that an over-complex student can emulate a less complex teacher rule, while an under-complex student reaches a state which realizes the minimal generalization error compatible with the complexity mismatch.

Discrete combinatorial circuits emerging in neural networks: A mechanism for rules of grammar in the human brain?

In neural network research on language, the existence of discrete combinatorial rule representations is commonly denied. Combinatorial capacity of networks and brains is rather attributed to probability mapping and pattern overlay. Here, we demonstrate that networks incorporating relevant features of neuroanatomical connectivity and neuronal function give rise to discrete neuronal circuits that store combinatorial information and exhibit a function similar to elementary rules of grammar. Key properties of these networks are rich auto- and hetero-associative connectivity, availability of sequence detectors similar to those found in a range of animals, and unsupervised Hebbian learning. Input of specific word sequences establishes sequence detectors in the network, and substitutions of words and larger string segments from one syntactic category, occurring in the context of elements of a second syntactic class, lead to binding between them into neuronal assemblies. Critically, these newly formed aggregates of sequence detectors now respond in a discrete generalizing fashion when members of specific substitution classes of string elements are combined with each other. The discrete combinatorial neuronal assemblies (DCNAs) even respond in the same way to learned strings and to word sequences that never appeared in the input but conform to a rule. We also show how combinatorial information interacts with information about functional and anatomical properties of the brain in the emergence of discrete neuronal circuits that may implement rules and discuss the model in the wider context of brain mechanism for syntax and grammar. Implications for the evolution of human language are discussed in closing.

Learning binocular disparity encoding simple cells in a model of primary visual cortex

Event Abstract Back to Event Learning binocular disparity encoding simple cells in a model of primary visual cortex Mark Voss1*, Jan Wiltschut1 and Fred H. Hamker2 1 Westfalischen Wilhelms-Universitat, Psychologisches Institut II, Germany 2 Technichal University Chemnitz, Computer Science Department, Germany The neural process of stereoscopic depth discrimination is thought to be initiated in the primary visual cortex. So far, most models incorporating binocular disparity in primary visual cortex build upon constructed, disparity encoding neurons (e.g. Read, Progress in Molecular Biology and Biophysics, 2004), but see (Hoyer & Hyvarinen, Network, 2000) for applying ICA to stereo images. Although these artificially constructed neurons can take into account different types of binocular disparity encoding, namely by position or phase, and can cover a defined range of disparities, they give no insight into the development of structural and functional patterns in primary visual cortex and depict a very limited subset of neurons that might contribute to disparity encoding. Here, we have extended our monocular model of primary visual cortex with nonlinear dynamics and Hebbian learning (Wiltschut & Hamker, Vis. Neurosci., 2009) to binocular vision. After presenting natural stereo scenes to our model, the learned neurons show disparity tuning in diverse degrees and with complex structure. We observe different types of near- and far- tuned, oriented receptive fields similar as has been observed in V1. As compared to ICA, our model appears to provide a better fit to physiological data. We conclude that unsupervised Hebbian learning provides a useful model to explain the development of receptive fields, not only in the orientation and spatial frequency domain but also with respect to disparity. Conference: Bernstein Conference on Computational Neuroscience, Frankfurt am Main, Germany, 30 Sep - 2 Oct, 2009. Presentation Type: Poster Presentation Topic: Neural encoding and decoding Citation: Voss M, Wiltschut J and Hamker FH (2009). Learning binocular disparity encoding simple cells in a model of primary visual cortex. Front. Comput. Neurosci. Conference Abstract: Bernstein Conference on Computational Neuroscience. doi: 10.3389/conf.neuro.10.2009.14.104 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 27 Aug 2009; Published Online: 27 Aug 2009. * Correspondence: Mark Voss, Westfalischen Wilhelms-Universitat, Psychologisches Institut II, Munster, Germany, mark.voss@uni-muenster.de Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Mark Voss Jan Wiltschut Fred H Hamker Google Mark Voss Jan Wiltschut Fred H Hamker Google Scholar Mark Voss Jan Wiltschut Fred H Hamker PubMed Mark Voss Jan Wiltschut Fred H Hamker Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

Enactive Robot Vision

Enactivism claims that sensory-motor activity and embodiment are crucial in perceiving the environment and that machine vision could be a much simpler business if considered in this context. However, computational models of enactive vision are very rare and often rely on handcrafted control systems. In this article, we argue that the apparent complexity of the environment and of the robot brain can be significantly simplified if perception, behavior, and learning are allowed to co-develop on the same timescale. In doing so, robots become sensitive to, and actively exploit, characteristics of the environment that they can tackle within their own computational and physical constraints. We describe the application of this methodology in three sets of experiments: shape discrimination, car driving, and wheeled robot navigation. A further set of experiments, where the visual system can develop the receptive fields by means of unsupervised Hebbian learning, demonstrates that the receptive fields are consistently and significantly affected by the behavior of the system and differ from those predicted by most computational models of the visual cortex. Finally, we show that our robots can also replicate the performance deficiencies observed in experiments of motor deprivation with kittens.