Sparse Distributed Representations Research Articles

Performance Comparison of Different HTM-Spatial Pooler Algorithms Based on Information-Theoretic Measures

Hierarchical temporal memory (HTM) is a promising unsupervised machine-learning algorithm that models key principles of neocortical computation. One of the main components of HTM is the spatial pooler (SP), which encodes binary input streams into sparse distributed representations (SDRs). In this paper, we propose an information-theoretic framework for the performance comparison of HTM-spatial pooler (SP) algorithms, specifically, for quantifying the similarities and differences between sparse distributed representations in SP algorithms. We evaluate SP's standalone performance, as well as HTM's overall performance. Our comparison of various SP algorithms using Renyi mutual information, Renyi divergence, and Henze–Penrose divergence measures reveals that the SP algorithm with learning and a logarithmic boosting function yields the most effective and useful data representation. Moreover, the most effective SP algorithm leads to superior HTM results. In addition, we utilize our proposed framework to compare HTM with other state-of-the-art sequential learning algorithms. We illustrate that HTM exhibits superior adaptability to pattern changes over time than long short term memory (LSTM), gated recurrent unit (GRU) and online sequential extreme learning machine (OS-ELM) algorithms. This superiority is evident from the lower Renyi divergence of HTM (0.23) compared to LSTM6000 (0.33), LSTM3000 (0.38), GRU (0.41), and OS-ELM (0.49). HTM also achieved the highest Renyi mutual information value of 0.79, outperforming LSTM6000 (0.73), LSTM3000 (0.71), GRU (0.68), and OS-ELM (0.62). These findings not only confirm the numerous advantages of HTM over other sequential learning algorithm, but also demonstrate the effectiveness of our proposed information-theoretic approach as a powerful framework for comparing and evaluating various learning algorithms.

Information-theoretic analysis of Hierarchical Temporal Memory-Spatial Pooler algorithm with a new upper bound for the standard information bottleneck method.

Hierarchical Temporal Memory (HTM) is an unsupervised algorithm in machine learning. It models several fundamental neocortical computational principles. Spatial Pooler (SP) is one of the main components of the HTM, which continuously encodes streams of binary input from various layers and regions into sparse distributed representations. In this paper, the goal is to evaluate the sparsification in the SP algorithm from the perspective of information theory by the information bottleneck (IB), Cramer-Rao lower bound, and Fisher information matrix. This paper makes two main contributions. First, we introduce a new upper bound for the standard information bottleneck relation, which we refer to as modified-IB in this paper. This measure is used to evaluate the performance of the SP algorithm in different sparsity levels and various amounts of noise. The MNIST, Fashion-MNIST and NYC-Taxi datasets were fed to the SP algorithm separately. The SP algorithm with learning was found to be resistant to noise. Adding up to 40% noise to the input resulted in no discernible change in the output. Using the probabilistic mapping method and Hidden Markov Model, the sparse SP output representation was reconstructed in the input space. In the modified-IB relation, it is numerically calculated that a lower noise level and a higher sparsity level in the SP algorithm lead to a more effective reconstruction and SP with 2% sparsity produces the best results. Our second contribution is to prove mathematically that more sparsity leads to better performance of the SP algorithm. The data distribution was considered the Cauchy distribution, and the Cramer-Rao lower bound was analyzed to estimate SP's output at different sparsity levels.

Variable Binding for Sparse Distributed Representations: Theory and Applications.

Variable binding is a cornerstone of symbolic reasoning and cognition. But how binding can be implemented in connectionist models has puzzled neuroscientists, cognitive psychologists, and neural network researchers for many decades. One type of connectionist model that naturally includes a binding operation is vector symbolic architectures (VSAs). In contrast to other proposals for variable binding, the binding operation in VSAs is dimensionality-preserving, which enables representing complex hierarchical data structures, such as trees, while avoiding a combinatoric expansion of dimensionality. Classical VSAs encode symbols by dense randomized vectors, in which information is distributed throughout the entire neuron population. By contrast, in the brain, features are encoded more locally, by the activity of single neurons or small groups of neurons, often forming sparse vectors of neural activation. Following Laiho et al. (2015), we explore symbolic reasoning with a special case of sparse distributed representations. Using techniques from compressed sensing, we first show that variable binding in classical VSAs is mathematically equivalent to tensor product binding between sparse feature vectors, another well-known binding operation which increases dimensionality. This theoretical result motivates us to study two dimensionality-preserving binding methods that include a reduction of the tensor matrix into a single sparse vector. One binding method for general sparse vectors uses random projections, the other, block-local circular convolution, is defined for sparse vectors with block structure, sparse block-codes. Our experiments reveal that block-local circular convolution binding has ideal properties, whereas random projection based binding also works, but is lossy. We demonstrate in example applications that a VSA with block-local circular convolution and sparse block-codes reaches similar performance as classical VSAs. Finally, we discuss our results in the context of neuroscience and neural networks.

Thai Word Segmentation with a Brain-Inspired Sparse Distributed Representations Learning Memory.

Word segmentation is necessary for many natural language processing, especially Thai language, that is, unsegmented words. However, wrong segmentation causes terrible performance in the final result. In this study, we propose two new brain-inspired methods based on Hawkins' approach to address Thai word segmentation. Sparse Distributed Representations (SDRs) are used to model the neocortex structure of the brain to store and transfer information. The first proposed method, THDICTSDR, improves the dictionary-based approach by utilizing SDRs to learn the surrounding context and combine with n-gram to select the correct word. The second method uses SDRs instead of a dictionary and is called THSDR. The evaluation uses the BEST2010 and LST20 standard datasets for segmentation words by comparing them with the longest matching, newmm, and Deepcut, which is state-of-the-art in the deep learning approach. The result shows that the first method provides the accuracy, and performances are significantly better than other dictionary bases. The first new method can achieve F1-Score at 95.60%, comparable to the state-of-the-art and Deepcut F1-Score at 96.34%. However, it provides a better performance F1-Score at 96.78% in learning all vocabularies. In addition, it can achieve 99.48% F1-Score beyond Deepcut 97.65% in case of all sentences being learnt. The second method has fault tolerance to noise and provides overall result over deep learning in all cases.

Parallel Implementation of Sparse Distributed Memory for Semantic Storage

Introduction. Sparse Distributed Memory (SDM) and Binary Sparse Distributed Representations (Binary Sparse Distributed Representations, BSDR), as two phenomenological approaches to biological memory modelling, have many similarities. The idea of ??their integration into a hybrid semantic storage model with SDM as a low-level cleaning memory (brain cells) for BSDR, which is used as an encoder of high-level symbolic information, is natural. A hybrid semantic store should be able to store holistic data (for example, structures of interconnected and sequential key-value pairs) in a neural network. A similar design has been proposed several times since the 1990s. However, the previously proposed models are impractical due to insufficient scalability and/or low storage density. The gap between SDM and BSDR can be bridged by the results of a third theory related to sparse signals: Compressive Sensing or Sampling (CS). In this article, we focus on the highly efficient parallel implementation of the CS-SDM hybrid memory model for graphics processing units on the NVIDIA CUDA platform, analyze the computational complexity of CS-SDM operations for the case of parallel implementation, and offer optimization techniques for conducting experiments with big sequential batches of vectors. The purpose of the paper is to propose an efficient software implementation of sparse-distributed memory for preserving semantics on modern graphics processing units. Results. Parallel algorithms for CS-SDM operations are proposed, their computational complexity is estimated, and a parallel implementation of the CS-SDM hybrid semantic store is given. Optimization of vector reconstruction for experiments with sequential data batches is proposed. Conclusions. The obtained results show that the design of CS-SDM is naturally parallel and that its algorithms are by design compatible with the architecture of systems with massive parallelism. The conducted experiments showed high performance of the developed implementation of the SDM memory block. Keywords: GPU, CUDA, neural network, Sparse Distributed Memory, associative memory, Compressive Sensing.

Memory Model for Morphological Semantics of Visual Stimuli Using Sparse Distributed Representation

Recent achievements on CNN (convolutional neural networks) and DNN (deep neural networks) researches provide a lot of practical applications on computer vision area. However, these approaches require construction of huge size of training data for learning process. This paper tries to find a way for continual learning which does not require prior high-cost training data construction by imitating a biological memory model. We employ SDR (sparse distributed representation) for information processing and semantic memory model, which is known as a representation model of firing patterns on neurons in neocortex area. This paper proposes a novel memory model to reflect remembrance of morphological semantics of visual input stimuli. The proposed memory model considers both memory process and recall process separately. First, memory process converts input visual stimuli to sparse distributed representation, and in this process, morphological semantic of input visual stimuli can be preserved. Next, recall process can be considered by comparing sparse distributed representation of new input visual stimulus and remembered sparse distributed representations. Superposition of sparse distributed representation is used to measure similarities. Experimental results using 10,000 images in MNIST (Modified National Institute of Standards and Technology) and Fashion-MNIST data sets show that the sparse distributed representation of the proposed model efficiently keeps morphological semantic of the input visual stimuli.

Memristor-Based HTM Spatial Pooler With On-Device Learning for Pattern Recognition

This article investigates hardware implementation of hierarchical temporal memory (HTM), a brain-inspired machine learning algorithm that mimics the key functions of the neocortex and is applicable to many machine learning tasks. Spatial pooler (SP) is one of the main parts of HTM, designed to learn the spatial information and obtain the sparse distributed representations (SDRs) of input patterns. The other part is temporal memory (TM) which aims to learn the temporal information of inputs. The memristor, which is an appropriate synapse emulator for neuromorphic systems, can be used as the synapse in SP and TM circuits. In this article, a memristor-based SP (MSP) circuit structure is designed to accelerate the execution of the SP algorithm. The presented MSP has properties of modeling both the synaptic permanence and the synaptic connection state within a single synapse, and on-device and parallel learning. Simulation results of statistic metrics and classification tasks on several real-world datasets substantiate the validity of MSP.

FOUR-DIMENSIONAL ENCODING OF CHARACTER SEQUENCES AND EVALUATION OF THEIR SIMILARITIES AND DIFFERENCES

This paper describes a string encoding algorithm, which produces sparse dis-tributed representations (SDR) of text data. In essence, this is a modified version of a prior algorithm and the modifications have the following benefits: - the ability to decode data, without loss of information; - greatly increased capacity of the encoding space; - the possibility of performing more detailed comparisons of encoded strings. The main disadvantage compared to the prior algorithm is the increased complexity of the procedure for encoded string comparison. This is due to the use of a four-dimensional encoding space, instead of a two-dimensional space.

A Computational Theory for the Emergence of Grammatical Categories in Cortical Dynamics.

A general agreement in psycholinguistics claims that syntax and meaning are unified precisely and very quickly during online sentence processing. Although several theories have advanced arguments regarding the neurocomputational bases of this phenomenon, we argue that these theories could potentially benefit by including neurophysiological data concerning cortical dynamics constraints in brain tissue. In addition, some theories promote the integration of complex optimization methods in neural tissue. In this paper we attempt to fill these gaps introducing a computational model inspired in the dynamics of cortical tissue. In our modeling approach, proximal afferent dendrites produce stochastic cellular activations, while distal dendritic branches–on the other hand–contribute independently to somatic depolarization by means of dendritic spikes, and finally, prediction failures produce massive firing events preventing formation of sparse distributed representations. The model presented in this paper combines semantic and coarse-grained syntactic constraints for each word in a sentence context until grammatically related word function discrimination emerges spontaneously by the sole correlation of lexical information from different sources without applying complex optimization methods. By means of support vector machine techniques, we show that the sparse activation features returned by our approach are well suited—bootstrapping from the features returned by Word Embedding mechanisms—to accomplish grammatical function classification of individual words in a sentence. In this way we develop a biologically guided computational explanation for linguistically relevant unification processes in cortex which connects psycholinguistics to neurobiological accounts of language. We also claim that the computational hypotheses established in this research could foster future work on biologically-inspired learning algorithms for natural language processing applications.

A Neurologically Inspired Sequence Processing Model for Mobile Robot Place Recognition

Visual place recognition is the problem of camera-based localization of a robot given a database of images of known places, potentially under severe appearance changes (e.g., different weather or illumination). Typically, the current query images and the database images are sequences of consecutive images. In previous work, we proposed to adapt hierarchical temporal memory, a biologically plausible model of sequence processing in the human neocortex to address this task. The previous work described the algorithmic steps and showed synthetic experimental results from simulation. This letter extends the approach to application on real-world data based on a novel encoder for state-of-the-art image processing front ends. The neurologically inspired approach is compared with several state-of-the-art algorithms on a variety of datasets and shows preferable performance. Beyond the place recognition performance, the neurological roots of the algorithm result in appealing properties like potentially very energy efficient implementation due to the usage of sparse distributed representations and natural extendability like the integration of motion estimates similar to entorhinal grid cells. Finally, we underline its practical applicability by online, soft real-time application on a mobile robot.

Phonetic acquisition in cortical dynamics, a computational approach.

Many computational theories have been developed to improve artificial phonetic classification performance from linguistic auditory streams. However, less attention has been given to psycholinguistic data and neurophysiological features recently found in cortical tissue. We focus on a context in which basic linguistic units–such as phonemes–are extracted and robustly classified by humans and other animals from complex acoustic streams in speech data. We are especially motivated by the fact that 8-month-old human infants can accomplish segmentation of words from fluent audio streams based exclusively on the statistical relationships between neighboring speech sounds without any kind of supervision. In this paper, we introduce a biologically inspired and fully unsupervised neurocomputational approach that incorporates key neurophysiological and anatomical cortical properties, including columnar organization, spontaneous micro-columnar formation, adaptation to contextual activations and Sparse Distributed Representations (SDRs) produced by means of partial N-Methyl-D-aspartic acid (NMDA) depolarization. Its feature abstraction capabilities show promising phonetic invariance and generalization attributes. Our model improves the performance of a Support Vector Machine (SVM) classifier for monosyllabic, disyllabic and trisyllabic word classification tasks in the presence of environmental disturbances such as white noise, reverberation, and pitch and voice variations. Furthermore, our approach emphasizes potential self-organizing cortical principles achieving improvement without any kind of optimization guidance which could minimize hypothetical loss functions by means of–for example–backpropagation. Thus, our computational model outperforms multiresolution spectro-temporal auditory feature representations using only the statistical sequential structure immerse in the phonotactic rules of the input stream.

Using the Hierarchical Temporal Memory Spatial Pooler for short-term forecasting of electrical load time series

In this paper, an emerging state-of-the-art machine intelligence technique called the Hierarchical Temporal Memory (HTM) is applied to the task of short-term load forecasting (STLF). A HTM Spatial Pooler (HTM-SP) stage is used to continually form sparse distributed representations (SDRs) from a univariate load time series data, a temporal aggregator is used to transform the SDRs into a sequential bivariate representation space and an overlap classifier makes temporal classifications from the bivariate SDRs through time. The comparative performance of HTM on several daily electrical load time series data including the Eunite competition dataset and the Polish power system dataset from 2002 to 2004 are presented. The robustness performance of HTM is also further validated using hourly load data from three more recent electricity markets. The results obtained from experimenting with the Eunite and Polish dataset indicated that HTM will perform better than the existing techniques reported in the literature. In general, the robustness test also shows that the error distribution performance of the proposed HTM technique is positively skewed for most of the years considered and with kurtosis values mostly lower than a base value of 3 indicating a reasonable level of outlier rejections.

Increment Learning and Rapid Retrieval of Visual Information Based on Pattern Association Memory

In this paper, we present a simple yet powerful visual memory model called Increment Pattern Association Memory Model to achieve increment learning and rapid retrieval of visual information. We extend the basic pattern association network by learning patterns based on Leabra framework which combines error-driven learning and Hebbian learning together. In the proposed model, image features act as conditioned stimulus and class labels act as unconditioned stimulus. Increment Learning is achieved by assigning individual synaptic weight to different categories. Meanwhile, k-winners-take-all inhibitory competition function is employed to achieve sparse distributed representations. In addition, normalized dot product is used to realize rapid recall of visual memory. To evaluate the performance of the proposed model, we conduct a series of experiments on three benchmark datasets using three feature-extraction methods. Experimental results demonstrate that the proposed model has a compelling advantage over some prevalent neural networks such as SOINN, BPNN, Biased ART and Self-supervised ART.

The HTM Spatial Pooler-A Neocortical Algorithm for Online Sparse Distributed Coding.

Hierarchical temporal memory (HTM) provides a theoretical framework that models several key computational principles of the neocortex. In this paper, we analyze an important component of HTM, the HTM spatial pooler (SP). The SP models how neurons learn feedforward connections and form efficient representations of the input. It converts arbitrary binary input patterns into sparse distributed representations (SDRs) using a combination of competitive Hebbian learning rules and homeostatic excitability control. We describe a number of key properties of the SP, including fast adaptation to changing input statistics, improved noise robustness through learning, efficient use of cells, and robustness to cell death. In order to quantify these properties we develop a set of metrics that can be directly computed from the SP outputs. We show how the properties are met using these metrics and targeted artificial simulations. We then demonstrate the value of the SP in a complete end-to-end real-world HTM system. We discuss the relationship with neuroscience and previous studies of sparse coding. The HTM spatial pooler represents a neurally inspired algorithm for learning sparse representations from noisy data streams in an online fashion.

Modification of Holographic Graph Neuron Using Sparse Distributed Representations

This article presents a modification of the recently proposed Holographic Graph Neuron approach for memorizing patterns of generic sensor stimuli. The original approach represents patterns as dense binary vectors, where zeros and ones are equiprobable. The presented modification employs sparse binary distributed representations where the number of ones is less than zeros. Sparse representations are more biologically plausible because activities of real neurons are sparse. Performance was studied comparing approaches for different sizes of dimensionality.

A cortical theory of super-efficient probabilistic inference based on sparse distributed representations

The remarkable structural homogeneity of isocortex strongly suggests a canonical cortical algorithm that performs the same essential function in all regions [1]. That function is widely construed/modeled as probabilistic inference, i.e., the ability, given an input, to retrieve the best-matching memory (or, most likely hypothesis) stored in memory. In [2], I described a cortical model for which both storage (learning) of new items into memory and probabilistic inference are constant time operations, which is a level of performance not present in any other published information processing system. This efficiency depends critically on: a) representing inputs with sparse distributed representations (SDRs), i.e., relatively small sets of binary units chosen from a large pool; and on b) choosing (learning) new SDRs so that more similar inputs are mapped to more highly

Building a world model with structure-sensitive sparse binary distributed representations

We present a new cognitive architecture named Associative-Projective Neural Networks (APNNs). APNNs have a multi-module, multi-level, and multi-modal design that works with an original scheme of sparse binary distributed representations to construct world models of varied complexity required for both task-specific and more general cognitive modeling. APNNs provide scalability and flexibility due to a number of design features. Internal representations of APNNs are sparse binary vectors of fixed dimensionality for items of various complexity and generality. Representations of input scalars, vectors, or compositional relational structures are constructed on-the-fly, so that similar items produce representations similar in terms of vector dot-products. Thus, for example, similarity of relational structures (taking into account similarity of their components, their grouping and order) can be estimated by dot-products of their representations, without the need to follow edges or to match vertices of underlying graphs. Decoding distributed representations through the input representations is also possible. Storage, retrieval, and decoding of distributed representations are implemented by efficient auto-associative memories; using distributed memories based on the idea of Hebb’s cell assemblies additionally provides a natural tool for emergence of generalization hierarchies. In addition, we consider how APNNs account for representation grounding, deal with recent challenges for distributed representations, and present some open problems.

Quantum Computation via Sparse Distributed Representation

Quantum superposition says that any physical system simultaneously exists in all of its possible states, the number of which is exponential in the number of entities composing the system. The strength of presence of each possible state in the superposition, i.e., its probability of being observed, is represented by its probability amplitude coefficient. The assumption that these coefficients must be represented physically disjointly from each other, i.e., localistically, is nearly universal in the quantum theory/computing literature. Alternatively, these coefficients can be represented using sparse distributed representations (SDR), wherein each coefficient is represented by small subset of an overall population of units, and the subsets can overlap. Specifically, I consider an SDR model in which the overall population consists of Q WTA clusters, each with K binary units. Each coefficient is represented by a set of Q units, one per cluster. Thus, K^Q coefficients can be represented with KQ units. Thus, the particular world state, X, whose coefficient's representation, R(X), is the set of Q units active at time t has the max probability and the probability of every other state, Y_i, at time t, is measured by R(Y_i)'s intersection with R(X). Thus, R(X) simultaneously represents both the particular state, X, and the probability distribution over all states. Thus, set intersection may be used to classically implement quantum superposition. If algorithms exist for which the time it takes to store (learn) new representations and to find the closest-matching stored representation (probabilistic inference) remains constant as additional representations are stored, this meets the criterion of quantum computing. Such an algorithm has already been described: it achieves this "quantum speed-up" without esoteric hardware, and in fact, on a single-processor, classical (Von Neumann) computer.

High resolution fMRI of the human ventral stream: Reproducible sparse-distributed representations of faces and limbs

High resolution fMRI of the human ventral stream: Reproducible sparse-distributed representations of faces and limbs

Some approaches to analogical mapping with structure-sensitive distributed representations

This paper presents some techniques for analogical mapping using associative-projective neural networks (APNNs). Sparse binary distributed representations of constant high dimensionality are constructed on-the-fly for hierarchical structures of various complexity. Such representations encode both surface and structural similarity of analogical episodes. The introduced mapping approaches are illustrated using test analogies.