Deep Reinforcement Learning With Explicit Context Representation.

Though reinforcement learning (RL) has shown an outstanding capability for solving complex computational problems, most RL algorithms lack an explicit method that would allow learning from contextual information. On the other hand, humans often use context to identify patterns and relations among elements in the environment, along with how to avoid making wrong actions. However, what may seem like an obviously wrong decision from a human perspective could take hundreds of steps for an RL agent to learn to avoid. This article proposes a framework for discrete environments called Iota explicit context representation (IECR). The framework involves representing each state using contextual key frames (CKFs), which can then be used to extract a function that represents the affordances of the state; in addition, two loss functions are introduced with respect to the affordances of the state. The novelty of the IECR framework lies in its capacity to extract contextual information from the environment and learn from the CKFs' representation. We validate the framework by developing four new algorithms that learn using context: Iota deep Q-network (IDQN), Iota double deep Q-network (IDDQN), Iota dueling deep Q-network (IDuDQN), and Iota dueling double deep Q-network (IDDDQN). Furthermore, we evaluate the framework and the new algorithms in five discrete environments. We show that all the algorithms, which use contextual information, converge in around 40 000 training steps of the neural networks, significantly outperforming their state-of-the-art equivalents.

Value-Based Subgoal Discovery and Path Planning for Reaching Long-Horizon Goals.

Learning to reach long-horizon goals in spatial traversal tasks is a significant challenge for autonomous agents. Recent subgoal graph-based planning methods address this challenge by decomposing a goal into a sequence of shorter-horizon subgoals. These methods, however, use arbitrary heuristics for sampling or discovering subgoals, which may not conform to the cumulative reward distribution. Moreover, they are prone to learning erroneous connections (edges) between subgoals, especially those lying across obstacles. To address these issues, this article proposes a novel subgoal graph-based planning method called learning subgoal graph using value-based subgoal discovery and automatic pruning (LSGVP). The proposed method uses a subgoal discovery heuristic that is based on a cumulative reward (value) measure and yields sparse subgoals, including those lying on the higher cumulative reward paths. Moreover, LSGVP guides the agent to automatically prune the learned subgoal graph to remove the erroneous edges. The combination of these novel features helps the LSGVP agent to achieve higher cumulative positive rewards than other subgoal sampling or discovery heuristics, as well as higher goal-reaching success rates than other state-of-the-art subgoal graph-based planning methods.

Who are the 'silent spreaders'?: contact tracing in spatio-temporal memory models.

The COVID-19 epidemic has swept the world for over two years. However, a large number of infectious asymptomatic COVID-19 cases (ACCs) are still making the breaking up of the transmission chains very difficult. Efforts by epidemiological researchers in many countries have thrown light on the clinical features of ACCs, but there is still a lack of practical approaches to detect ACCs so as to help contain the pandemic. To address the issue of ACCs, this paper presents a neural network model called Spatio-Temporal Episodic Memory for COVID-19 (STEM-COVID) to identify ACCs from contact tracing data. Based on the fusion Adaptive Resonance Theory (ART), the model encodes a collective spatio-temporal episodic memory of individuals and incorporates an effective mechanism of parallel searches for ACCs. Specifically, the episodic traces of the identified positive cases are used to map out the episodic traces of suspected ACCs using a weighted evidence pooling method. To evaluate the efficacy of STEM-COVID, a realistic agent-based simulation model for COVID-19 spreading is implemented based on the recent epidemiological findings on ACCs. The experiments based on rigorous simulation scenarios, manifesting the current situation of COVID-19 spread, show that the STEM-COVID model with weighted evidence pooling has a higher level of accuracy and efficiency for identifying ACCs when compared with several baselines. Moreover, the model displays strong robustness against noisy data and different ACC proportions, which partially reflects the effect of breakthrough infections after vaccination on the virus transmission.

End-to-End Hierarchical Reinforcement Learning With Integrated Subgoal Discovery.

Hierarchical reinforcement learning (HRL) is a promising approach to perform long-horizon goal-reaching tasks by decomposing the goals into subgoals. In a holistic HRL paradigm, an agent must autonomously discover such subgoals and also learn a hierarchy of policies that uses them to reach the goals. Recently introduced end-to-end HRL methods accomplish this by using the higher-level policy in the hierarchy to directly search the useful subgoals in a continuous subgoal space. However, learning such a policy may be challenging when the subgoal space is large. We propose integrated discovery of salient subgoals (LIDOSS), an end-to-end HRL method with an integrated subgoal discovery heuristic that reduces the search space of the higher-level policy, by explicitly focusing on the subgoals that have a greater probability of occurrence on various state-transition trajectories leading to the goal. We evaluate LIDOSS on a set of continuous control tasks in the MuJoCo domain against hierarchical actor critic (HAC), a state-of-the-art end-to-end HRL method. The results show that LIDOSS attains better goal achievement rates than HAC in most of the tasks.

Vision-Based Topological Mapping and Navigation With Self-Organizing Neural Networks.

Spatial mapping and navigation are critical cognitive functions of autonomous agents, enabling one to learn an internal representation of an environment and move through space with real-time sensory inputs, such as visual observations. Existing models for vision-based mapping and navigation, however, suffer from memory requirements that increase linearly with exploration duration and indirect path following behaviors. This article presents e -TM, a self-organizing neural network-based framework for incremental topological mapping and navigation. e -TM models the exploration trajectories explicitly as episodic memory, wherein salient landmarks are sequentially extracted as "events" from streaming observations. A memory consolidation procedure then performs a playback mechanism and transfers the embedded knowledge of the environmental layout into spatial memory, encoding topological relations between landmarks. Fusion adaptive resonance theory (ART) networks, as the building block of the two memory modules, can generalize multiple input patterns into memory templates and, therefore, provide a compact spatial representation and support the discovery of novel shortcuts through inferences. For navigation, e -TM applies a transfer learning paradigm to integrate human demonstrations into a pretrained locomotion network for smoother movements. Experimental results based on VizDoom, a simulated 3-D environment, have shown that, compared to semiparametric topological memory (SPTM), a state-of-the-art model, e -TM reduces the time costs of navigation significantly while learning much sparser topological graphs.

EEG-based Emotion Recognition Using Spatial-Temporal Representation via Bi-GRU.

Many prior studies on EEG-based emotion recognition did not consider the spatial-temporal relationships among brain regions and across time. In this paper, we propose a Regionally-Operated Domain Adversarial Network (RODAN), to learn spatial-temporal relationships that correlate between brain regions and time. Moreover, we incorporate the attention mechanism to enable cross-domain learning to capture both spatial-temporal relationships among the EEG electrodes and an adversarial mechanism to reduce the domain shift in EEG signals. To evaluate the performance of RODAN, we conduct subject-dependent, subject-independent, and subject-biased experiments on both DEAP and SEED-IV data sets, which yield encouraging results. In addition, we also discuss the biased sampling issue often observed in EEG-based emotion recognition and present an unbiased benchmark for both DEAP and SEED-IV.

CoRRN: Cooperative Reflection Removal Network.

Removing the undesired reflections from images taken through the glass is of broad application to various computer vision tasks. Non-learning based methods utilize different handcrafted priors such as the separable sparse gradients caused by different levels of blurs, which often fail due to their limited description capability to the properties of real-world reflections. In this paper, we propose a network with the feature-sharing strategy to tackle this problem in a cooperative and unified framework, by integrating image context information and the multi-scale gradient information. To remove the strong reflections existed in some local regions, we propose a statistic loss by considering the gradient level statistics between the background and reflections. Our network is trained on a new dataset with 3250 reflection images taken under diverse real-world scenes. Experiments on a public benchmark dataset show that the proposed method performs favorably against state-of-the-art methods.

Salience-aware adaptive resonance theory for large-scale sparse data clustering.

Sparse data is known to pose challenges to cluster analysis, as the similarity between data tends to be ill-posed in the high-dimensional Hilbert space. Solutions in the literature typically extend either k-means or spectral clustering with additional steps on representation learning and/or feature weighting. However, adding these usually introduces new parameters and increases computational cost, thus inevitably lowering the robustness of these algorithms when handling massive ill-represented data. To alleviate these issues, this paper presents a class of self-organizing neural networks, called the salience-aware adaptive resonance theory (SA-ART) model. SA-ART extends Fuzzy ART with measures for cluster-wise salient feature modeling. Specifically, two strategies, i.e. cluster space matching and salience feature weighting, are incorporated to alleviate the side-effect of noisy features incurred by high dimensionality. Additionally, cluster weights are bounded by the statistical means and minimums of the samples therein, making the learning rate also self-adaptable. Notably, SA-ART allows clusters to have their own sets of self-adaptable parameters. It has the same time complexity of Fuzzy ART and does not introduce additional hyperparameters that profile cluster properties. Comparative experiments have been conducted on the ImageNet and BlogCatalog datasets, which are large-scale and include sparsely-represented data. The results show that, SA-ART achieves 51.8% and 18.2% improvement over Fuzzy ART, respectively. While both have a similar time cost, SA-ART converges faster and can reach a better local minimum. In addition, SA-ART consistently outperforms six other state-of-the-art algorithms in terms of precision and F1 score. More importantly, it is much faster and exhibits stronger robustness to large and complex data.

Self-organizing neural networks for universal learning and multimodal memory encoding.

Learning and memory are two intertwined cognitive functions of the human brain. This paper shows how a family of biologically-inspired self-organizing neural networks, known as fusion Adaptive Resonance Theory (fusion ART), may provide a viable approach to realizing the learning and memory functions. Fusion ART extends the single-channel Adaptive Resonance Theory (ART) model to learn multimodal pattern associative mappings. As a natural extension of ART, various forms of fusion ART have been developed for a myriad of learning paradigms, ranging from unsupervised learning to supervised learning, semi-supervised learning, multimodal learning, reinforcement learning, and sequence learning. In addition, fusion ART models may be used for representing various types of memories, notably episodic memory, semantic memory and procedural memory. In accordance with the notion of embodied intelligence, such neural models thus provide a computational account of how an autonomous agent may learn and adapt in a real-world environment. The efficacy of fusion ART in learning and memory shall be discussed through various examples and illustrative case studies.

Perception Coordination Network: A Neuro Framework for Multimodal Concept Acquisition and Binding.

To simulate the concept acquisition and binding of different senses in the brain, a biologically inspired neural network model named perception coordination network (PCN) is proposed. It is a hierarchical structure, which is functionally divided into the primary sensory area (PSA), the primary sensory association area (SAA), and the higher order association area (HAA). The PSA contains feature neurons which respond to many elementary features, e.g., colors, shapes, syllables, and basic flavors. The SAA contains primary concept neurons which combine the elementary features in the PSA to represent unimodal concept of objects, e.g., the image of an apple, the Chinese word "[píng guǒ]" which names the apple, and the taste of the apple. The HAA contains associated neurons which connect the primary concept neurons of several PSA, e.g., connects the image, the taste, and the name of an apple. It means that the associated neurons have a multimodal response mode. Therefore, this area executes multisensory integration. PCN is an online incremental learning system, it is able to continuously acquire and bind multimodality concepts in an online way. The experimental results suggest that PCN is able to handle the multimodal concept acquisition and binding effectively.

Region-Aware Reflection Removal with Unified Content and Gradient Priors.

Removing the undesired reflections in images taken through the glass is of broad application to various image processing and computer vision tasks. Existing single image based solutions heavily rely on scene priors such as separable sparse gradients caused by different levels of blur, and they are fragile when such priors are not observed. In this paper, we notice that strong reflections usually dominant a limited region in the whole image, and propose a Region-aware Reflection Removal (R3) approach by automatically detecting and heterogeneously processing regions with and without reflections. We integrate content and gradient priors to jointly achieve missing contents restoration as well as background and reflection separation in a unified optimization framework. Extensive validation using 50 sets of real data shows that the proposed method outperforms state-of-the-art on both quantitative metrics and visual qualities.

Adaptive Scaling of Cluster Boundaries for Large-Scale Social Media Data Clustering.

The large scale and complex nature of social media data raises the need to scale clustering techniques to big data and make them capable of automatically identifying data clusters with few empirical settings. In this paper, we present our investigation and three algorithms based on the fuzzy adaptive resonance theory (Fuzzy ART) that have linear computational complexity, use a single parameter, i.e., the vigilance parameter to identify data clusters, and are robust to modest parameter settings. The contribution of this paper lies in two aspects. First, we theoretically demonstrate how complement coding, commonly known as a normalization method, changes the clustering mechanism of Fuzzy ART, and discover the vigilance region (VR) that essentially determines how a cluster in the Fuzzy ART system recognizes similar patterns in the feature space. The VR gives an intrinsic interpretation of the clustering mechanism and limitations of Fuzzy ART. Second, we introduce the idea of allowing different clusters in the Fuzzy ART system to have different vigilance levels in order to meet the diverse nature of the pattern distribution of social media data. To this end, we propose three vigilance adaptation methods, namely, the activation maximization (AM) rule, the confliction minimization (CM) rule, and the hybrid integration (HI) rule. With an initial vigilance value, the resulting clustering algorithms, namely, the AM-ART, CM-ART, and HI-ART, can automatically adapt the vigilance values of all clusters during the learning epochs in order to produce better cluster boundaries. Experiments on four social media data sets show that AM-ART, CM-ART, and HI-ART are more robust than Fuzzy ART to the initial vigilance value, and they usually achieve better or comparable performance and much faster speed than the state-of-the-art clustering algorithms that also do not require a predefined number of clusters.

Self-organizing neural networks integrating domain knowledge and reinforcement learning.

The use of domain knowledge in learning systems is expected to improve learning efficiency and reduce model complexity. However, due to the incompatibility with knowledge structure of the learning systems and real-time exploratory nature of reinforcement learning (RL), domain knowledge cannot be inserted directly. In this paper, we show how self-organizing neural networks designed for online and incremental adaptation can integrate domain knowledge and RL. Specifically, symbol-based domain knowledge is translated into numeric patterns before inserting into the self-organizing neural networks. To ensure effective use of domain knowledge, we present an analysis of how the inserted knowledge is used by the self-organizing neural networks during RL. To this end, we propose a vigilance adaptation and greedy exploitation strategy to maximize exploitation of the inserted domain knowledge while retaining the plasticity of learning and using new knowledge. Our experimental results based on the pursuit-evasion and minefield navigation problem domains show that such self-organizing neural network can make effective use of domain knowledge to improve learning efficiency and reduce model complexity.

Neural modeling of episodic memory: encoding, retrieval, and forgetting.

This paper presents a neural model that learns episodic traces in response to a continuous stream of sensory input and feedback received from the environment. The proposed model, based on fusion adaptive resonance theory (ART) network, extracts key events and encodes spatio-temporal relations between events by creating cognitive nodes dynamically. The model further incorporates a novel memory search procedure, which performs a continuous parallel search of stored episodic traces. Combined with a mechanism of gradual forgetting, the model is able to achieve a high level of memory performance and robustness, while controlling memory consumption over time. We present experimental studies, where the proposed episodic memory model is evaluated based on the memory consumption for encoding events and episodes as well as recall accuracy using partial and erroneous cues. Our experimental results show that: 1) the model produces highly robust performance in encoding and recalling events and episodes even with incomplete and noisy cues; 2) the model provides enhanced performance in a noisy environment due to the process of forgetting; and 3) compared with prior models of spatio-temporal memory, our model shows a higher tolerance toward noise and errors in the retrieval cues.

Creating an immersive game world with evolutionary fuzzy cognitive maps.

The Evolutionary Fuzzy Cognitive Map improves on serious games by modeling both fuzzy and probabilistic causal relationships among the game's variables. It permits asynchronous updates of the variables so that they can evolve dynamically and stochastically. These improvements give players a more engaging, immersive experience.

Self-organizing neural architectures and cooperative learning in a multiagent environment.

Temporal-Difference-Fusion Architecture for Learning, Cognition, and Navigation (TD-FALCON) is a generalization of adaptive resonance theory (a class of self-organizing neural networks) that incorporates TD methods for real-time reinforcement learning. In this paper, we investigate how a team of TD-FALCON networks may cooperate to learn and function in a dynamic multiagent environment based on minefield navigation and a predator/prey pursuit tasks. Experiments on the navigation task demonstrate that TD-FALCON agent teams are able to adapt and function well in a multiagent environment without an explicit mechanism of collaboration. In comparison, traditional Q-learning agents using gradient-descent-based feedforward neural networks, trained with the standard backpropagation and the resilient-propagation (RPROP) algorithms, produce a significantly poorer level of performance. For the predator/prey pursuit task, we experiment with various cooperative strategies and find that a combination of a high-level compressed state representation and a hybrid reward function produces the best results. Using the same cooperative strategy, the TD-FALCON team also outperforms the RPROP-based reinforcement learners in terms of both task completion rate and learning efficiency.

Predictive neural networks for gene expression data analysis.

Gene expression data generated by DNA microarray experiments have provided a vast resource for medical diagnosis and disease understanding. Most prior work in analyzing gene expression data, however, focuses on predictive performance but not so much on deriving human understandable knowledge. This paper presents a systematic approach for learning and extracting rule-based knowledge from gene expression data. A class of predictive self-organizing networks known as Adaptive Resonance Associative Map (ARAM) is used for modelling gene expression data, whose learned knowledge can be transformed into a set of symbolic IF-THEN rules for interpretation. For dimensionality reduction, we illustrate how the system can work with a variety of feature selection methods. Benchmark experiments conducted on two gene expression data sets from acute leukemia and colon tumor patients show that the proposed system consistently produces predictive performance comparable, if not superior, to all previously published results. More importantly, very simple rules can be discovered that have extremely high diagnostic power. The proposed methodology, consisting of dimensionality reduction, predictive modelling, and rule extraction, provides a promising approach to gene expression analysis and disease understanding.

Modified ART 2A growing network capable of generating a fixed number of nodes.

This paper introduces the Adaptive Resonance Theory under Constraint (ART-C 2A) learning paradigm based on ART 2A, which is capable of generating a user-defined number of recognition nodes through online estimation of an appropriate vigilance threshold. Empirical experiments compare the cluster validity and the learning efficiency of ART-C 2A with those of ART 2A, as well as three closely related clustering methods, namely online K-Means, batch K-Means, and SOM, in a quantitative manner. Besides retaining the online cluster creation capability of ART 2A, ART-C 2A gives the alternative clustering solution, which allows a direct control on the number of output clusters generated by the self-organizing process.