Recent advancements in artificial intelligence hold ample potential for monitoring applications using surveillance cameras. However, concerns about privacy and model bias have made it challenging to utilize them in public. Although de-identification approaches have been proposed in the literature, aiming to achieve a certain level of anonymization (AN), most of them employ deep learning models that are computationally demanding for real-time edge deployment. This study revisits conventional AN solutions for privacy protection and real-time video anomaly detection (VAD) applications. We propose a lightweight adaptive AN for VAD (LA3D) that employs dynamic adjustment to enhance full-body privacy protection. We have evaluated privacy protection and VAD utility retention efficacy using several publicly available datasets to examine the strengths and weaknesses of different AN methods and highlight the promising leverage of our approach. Our experiment demonstrates that the LA3D enables substantial improvement in privacy AN without severely degrading VAD efficacy, outperforming conventional and deep learning approaches. Code: https://github.com/muleina/LA3D.

Accurate blood glucose prediction is vital for effective diabetes management. However, traditional models focus primarily on improving accuracy, often overlooking the importance of quantifying uncertainty, which is essential for informed clinical decision-making. This study presents an uncertainty-aware blood glucose prediction model using a Bayesian neural network, which not only focuses on predictive accuracy but also emphasizes uncertainty analysis. The model is trained on blood glucose data and evaluated using Root Mean Squared Error (RMSE) for accuracy and confidence interval coverage to assess uncertainty calibration. The results show that the model achieves an RMSE of 22.21, indicating strong predictive performance. Moreover, the observed glucose values consistently fall within the predicted confidence intervals$(\pm 2 \sigma)$, demonstrating well-calibrated uncertainty estimates. The low correlation (0.06) between glucose rate of change and uncertainty further suggests that the model performs robustly across both stable and fluctuating glucose levels.

Retrieval of scenes from traffic videos is an important task in intelligent transportation systems (ITS) for efficient traffic management in AI smart cities. This work proposes natural language-based vehicle retrieval from traffic monitoring videos, emphasizing the significance of temporal information and context. We present contrastive learning as a technique to optimize joint representations of vision and language modalities within a shared latent representation space. The approach involves training contrastive losses to keep similar encodings closer in joint feature representation space by minimizing the distance between positive visual-text pairs and maximizing the distance between negative visual-text pairs. Our study employs state-of-the-art vision models for visual encoding and transformer-based language models for text encoding. We analyze the impact of feature selection from visuals and text on retrieval performance. We evaluate the efficacy of our proposed method on the AI City Challenge 2022 dataset for natural language-based vehicle retrieval, achieving performance accuracy of 49.84% Mean Reciprocal Rank (MRR) on the test dataset, securing second position on the leader board. Our approach highlights the effectiveness of feature selection and contrastive learning for enhancing multimodal retrieval tasks.

Abstract Extensive research has been conducted on the automatic classification of sleep stages utilizing deep neural networks and other neurophysiological markers. However, for sleep specialists to employ models as an assistive solution, it is necessary to comprehend how the models arrive at a particular outcome, necessitating the explainability of these models. This work proposes an explainable unified CNN-CRF approach (SleepXAI) for multi-class sleep stage classification designed explicitly for univariate time-series signals using modified gradient-weighted class activation mapping (Grad-CAM). The proposed approach significantly increases the overall accuracy of sleep stage classification while demonstrating the explainability of the multi-class labeling of univariate EEG signals, highlighting the parts of the signals emphasized most in predicting sleep stages. We extensively evaluated our approach to the sleep-EDF dataset, and it demonstrates the highest overall accuracy of 86.8% in identifying five sleep stage classes. More importantly, we achieved the highest accuracy when classifying the crucial sleep stage N1 with the lowest number of instances, outperforming the state-of-the-art machine learning approaches by 16.3%. These results motivate us to adopt the proposed approach in clinical practice as an aid to sleep experts.

open access article

Stock portfolio optimization is the process of continuous reallocation of funds to a selection of stocks. This is a particularly well-suited problem for reinforcement learning, as daily rewards are compounding and objective functions may include more than just profit, e.g., risk and sustainability. We developed a novel utility function with the Sharpe ratio representing risk and the environmental, social, and governance score (ESG) representing sustainability. We show that a state- of-the-art policy gradient method – multi-agent deep deterministic policy gradients (MADDPG) – fails to find the optimum policy due to flat policy gradients and we therefore replaced gradient descent with a genetic algorithm for parameter optimization. We show that our system outperforms MADDPG while improving on deep Q-learning approaches by allowing for continuous action spaces. Crucially, by incorporating risk and sustainability criteria in the utility function, we improve on the state-of-the-art in reinforcement learning for portfolio optimization; risk and sustainability are essential in any modern trading strategy, and we propose a system that does not merely report these metrics, but that actively optimizes the portfolio to improve on them.

Micro-segmentation of customers in the finance sector is a nontrivial task and has been an atypical omission from recent scientific literature. Where traditional segmentation classifies customers based on coarse features such as demographics, micro-segmentation depicts more nuanced differences between individuals, bringing forth several advantages including the potential for improved personalization in financial services. AI and representation learning offer a unique opportunity to solve the problem of micro-segmentation. Although ubiquitous in many industries, the proliferation of AI in sensitive industries such as finance has become contingent on the explainability of deep models. We had previously solved the micro-segmentation problem by extracting temporal features from the state space of a recurrent neural network (RNN). However, due to the inherent opacity of RNNs, our solution lacked an explanation. In this study, we address this issue by extracting a symbolic explanation for our model and providing an interpretation of our temporal features. For the explanation, we use a linear regression model to reconstruct the features in the state space with high fidelity. We show that our linear regression coefficients have not only learned the rules used to recreate the features, but have also learned the relationships that were not directly evident in the raw data. Finally, we propose a novel method to interpret the dynamics of the state space by using the principles of inverse regression and dynamical systems to locate and label a set of attractors.

Automatic sleep stage classification can play a vital role when measuring sleep quality and diagnosing different sleep-related ailments. Several automated sleep stage identification algorithms have been proposed using various physiological signals. However, most of these methods use hand-crafted features or multiple Electroencephalography (EEG) signals. This work proposes a one-dimensional convolutional autoencoder (1D-CAE) based on a single-channel EEG signal for sleep stage identification. A total of five 1-D CAEs models are implemented, and each model is trained to reconstructs a specific sleep stage with the lowest reconstruction error, thus enabling the sleep stage identification based on this error. Furthermore, the proposed approach is evaluated on the Sleep EDF expanded datasets and achieved an overall classification accuracy of 87.2% using a single-channel EEG FPz-Cz signal. Also, our approach demonstrated the highest sleep stage identification accuracy compared with the recent algorithms, especially for sleep stage N1, a short period that transitions between sleep stages during a sleep cycle.

Polysomnography (PSG), the gold standard for sleep stage classification, requires a sleep expert for scoring and is both resource-intensive and expensive. Many researchers currently focus on the real-time classification of the sleep stages based on biomedical signals, such as Electroencephalograph (EEG) and electrooculography (EOG). However, most of the research work is based on machine learning models with multiple signal inputs or hand-engineered features requiring prior knowledge of the sleep domain. We propose a novel encoded Time-Distributed Convolutional Neural Network (TDConvNet) to automatically classify sleep stages based on a single raw PSG signal. The TDConvNet can infer sleep stages in just 30-second epochs for the 5-stage sleep classification using a single EEG or EOG signal from the open Sleep-EDF dataset. We evaluated the TDConvNet performance on EEGs and EOG signals. The evaluation results show that the proposed method achieved the best performance with the EEG Fpz-Cz signal (0.85) compared to current literature, followed by EEG Pz-Oz (0.84) and EOG horizontal (0.82). The source code is available at https://github.com/michealdutt/TDConvNet.

Cyclone track prediction is a two dimensional time series prediction problem that involves latitudes and longitudes which define the position of a cyclone. Recurrent neural networks have been suitable for time series prediction due to their architectural properties in modeling temporal sequences. Coevolutionary recurrent neural networks have been used for time series prediction and also applied to cyclone track prediction. In this paper, we present an architecture for encoding two dimensional time series problem into Elman recurrent neural networks composed of a single input neuron. We use cooperative coevolution and back-propagation through-time algorithms for training. Our experiments show an improvement in the accuracy when compared to previous results using a different recurrent network architecture.

There has been quite a lot of research done in the field of Facial Expression Recognition, yet there has not been so much development in Facial Action Coding System Action Unit intensity detection. In Automated Facial Expression Recognition, intensity recognition of the Facial Action Coding System Action Units is a crucial part for it would give much broad information about the facial expression of an individual. In this research, a computationally efficient yet effective logistic regression based method that operates on a novel feature vector extracted from geometric relations between facial feature points is presented. Said method uses angles between facial feature points which are rotation invariant. The method was trained and tested on DISFA database and gave state of the art results.

Neural networks do not readily provide an explanation of the knowledge stored in their weights as part of their information processing. Until recently, neural networks were considered to be black boxes, with the knowledge stored in their weights not readily accessible. Since then, research has resulted in a number of algorithms for extracting knowledge in symbolic form from trained neural networks. This article addresses the extraction of knowledge in symbolic form from recurrent neural networks trained to behave like deterministic finite-state automata (DFAs). To date, methods used to extract knowledge from such networks have relied on the hypothesis that networks' states tend to cluster and that clusters of network states correspond to DFA states. The computational complexity of such a cluster analysis has led to heuristics that either limit the number of clusters that may form during training or limit the exploration of the space of hidden recurrent state neurons. These limitations, while necessary, may lead to decreased fidelity, in which the extracted knowledge may not model the true behavior of a trained network, perhaps not even for the training set. The method proposed here uses a polynomial time, symbolic learning algorithm to infer DFAs solely from the observation of a trained network's input-output behavior. Thus, this method has the potential to increase the fidelity of the extracted knowledge.

We p r e s e n t DEADLINER, a search engine that catalogs conference and workshop announcements, and ultimately will monitor and extract a wide range of academic convocation material from the web.The system currently extracts speakers, locations, dates, paper submission (and other) deadlines, topics, program committees, abstracts, and aÆliations.A user or user agent can perform detailed searches on these elds.DEADLINER was constructed using a methodology for rapid implementation of specialized search engines.This methodology avoids complex hand-tuned text extraction solutions, or natural language processing, by B a yesian integration of simple extractors that exploit loose formatting and keyw ord con ventions.The Bayesian framework further produces a search engine where each user can control the false alarm rate on a eld in an intuitive y et rigorous fashion.

There has been an increased interest in the use of prior knowledge for training neural networks. Given a set of training examples and an initial domain theory, a neural network is constructed that fits the training examples by preprogramming some of the weights. The initialized neural network is then trained using backpropagation to refine the knowledge. This paper proposes a heuristic for determining the strength of the inductive bias by making use of gradient information in weight space in the direction of the programmed weights. The network starts its search in weight space where the gradient is maximal thus speeding-up convergence. Tests on a benchmark problem from molecular biology demonstrate that our heuristic, on average, reduces the training time by 60% compared to a random choice of the strength of the inductive bias; this performance is within 20% of the training time that can be achieved with the optimal inductive bias. The difference in generalization performance is not statistically significant.

Both hidden Markov models (HMMs) and recurrent neural networks (RNNs) have been applied to sequence recognition problems. While HMMs are easy to train, they generally do not perform satisfactorily on difficult recognition problems. On the other hand, RNNs are excellent recognizers but are very hard to train. Hybrid HMM/NN approaches aim at taking advantage of the strengths of both paradigms while avoiding their respective weaknesses. The paper proposes an approach of combining HMMs with RNNs. We discuss an algorithm for directly mapping a trained HMM into a RNN architecture and derive a gradient-descent learning algorithm for knowledge refinement.

Biometric authentication has become a popular research topic due to its wide applicability, including the prevention of fraud in financial transactions. Handwritten signature verification, in contrast with other biometric based authentication methods such as fingerprint and retinal scanning, has the advantage that it is already widely used to endorse financial transactions. However, very little verification on these signatures is done today in practical scenarios. The paper reports on our ongoing research on automatic, online, handwritten signature verification. The hybrid system consists of a Kohonen self-organizing map which finds cluster centers in the training data and hidden Markov models which are trained to model the dynamics of signatures. Our initial results are very promising: the system achieves a 0% false rejection rate and a 13% false acceptance rate.

Neurofuzzy systems-the combination of artificial neural networks with fuzzy logic-have become useful in many application domains. However, conventional neurofuzzy models usually need enhanced representation power for applications that require context and state (e.g., speech, time series prediction, control). Some of these applications can be readily modeled as finite state automata. Previously, it was proved that deterministic finite state automata (DFA) can be synthesized by or mapped into recurrent neural networks by directly programming the DFA structure into the weights of the neural network. Based on those results, a synthesis method is proposed for mapping fuzzy finite state automata (FFA) into recurrent neural networks. Furthermore, this mapping is suitable for direct implementation in very large scale integration (VLSI), i.e., the encoding of FFA as a generalization of the encoding of DFA in VLSI systems. The synthesis method requires FFA to undergo a transformation prior to being mapped into recurrent networks. The neurons are provided with an enriched functionality in order to accommodate a fuzzy representation of FFA states. This enriched neuron functionality also permits fuzzy parameters of FFA to be directly represented as parameters of the neural network. We also prove the stability of fuzzy finite state dynamics of the constructed neural networks for finite values of network weight and, through simulations, give empirical validation of the proofs. Hence, we prove various knowledge equivalence representations between neural and fuzzy systems and models of automata.

There has been an increased interest in combining fuzzy systems with neural networks because fuzzy neural systems merge the advantages of both paradigms. On the one hand, parameters in fuzzy systems have clear physical meanings and rule-based and linguistic information can be incorporated into adaptive fuzzy systems in a systematic way. On the other hand, there exist powerful algorithms for training various neural network models. However, most of the proposed combined architectures are only able to process static input-output relationships; they are not able to process temporal input sequences of arbitrary length. Fuzzy finite-state automats (FFAs) can model dynamical processes whose current state depends on the current input and previous states. Unlike in the case of deterministic finite-state automats (DFAs), FFAs are not in one particular state, rather each state is occupied to some degree defined by a membership function. Based on previous work on encoding DFAs in discrete-time second-order recurrent neural networks, we propose an algorithm that constructs an augmented recurrent neural network that encodes a FFA and recognizes a given fuzzy regular language with arbitrary accuracy. We then empirically verify the encoding methodology by correct string recognition of randomly generated FFAs. In particular, we examine how the networks' performance varies as a function of synaptic weight strengths.

Recurrent neural networks that are trained to behave like deterministic finite-state automata (DFAs) can show deteriorating performance when tested on long strings. This deteriorating performance can be attributed to the instability of the internal representation of the learned DFA states. The use of a sigmoidel discriminant function together with the recurrent structure contribute to this instability. We prove that a simple algorithm can construct second-order recurrent neural networks with a sparse interconnection topology and sigmoidal discriminant function such that the internal DFA state representations are stable, that is, the constructed network correctly classifies strings of arbitrary length . The algorithm is based on encoding strengths of weights directly into the neural network. We derive a relationship between the weight strength and the number of DFA states for robust string classification. For a DFA with n state and m input alphabet symbols, the constructive algorithm generates a “programmed” neural network with O ( n ) neurons and O ( mn ) weights. We compare our algorithm to other methods proposed in the literature.

We propose an algorithm for encoding deterministic finite-state automata (DFAs) in second-order recurrent neural networks with sigmoidal discriminant function and we prove that the languages accepted by the constructed network and the DFA are identical. The desired finite-state network dynamics is achieved by programming a small subset of all weights. A worst case analysis reveals a relationship between the weight strength and the maximum allowed network size, which guarantees finite-state behavior of the constructed network. We illustrate the method by encoding random DFAs with 10, 100, and 1000 states. While the theory predicts that the weight strength scales with the DFA size, we find empirically the weight strength to be almost constant for all the random DFAs. These results can be explained by noting that the generated DFAs represent average cases. We empirically demonstrate the existence of extreme DFAs for which the weight strength scales with DFA size.

Recurrent neural networks readily process, recognize and generate temporal sequences. By encoding grammatical strings as temporal sequences, recurrent neural networks can be trained to behave like deterministic sequential finite-state automata. Algorithms have been developed for extracting grammatical rules from trained networks. Using a simple method for inserting prior knowledge (or rules) into recurrent neural networks, we show that recurrent neural networks are able to perform rule revision. Rule revision is performed by comparing the inserted rules with the rules in the finite-state automata extracted from trained networks. The results from training a recurrent neural network to recognize a known non-trivial, randomly-generated regular grammar show that not only do the networks preserve correct rules but that they are able to correct through training inserted rules which were initially incorrect (i.e. the rules were not the ones in the randomly generated grammar).

Determining the architecture of a neural network is an important issue for any learning task. For recurrent neural networks no general methods exist that permit the estimation of the number of layers of hidden neurons, the size of layers or the number of weights. We present a simple pruning heuristic that significantly improves the generalization performance of trained recurrent networks. We illustrate this heuristic by training a fully recurrent neural network on positive and negative strings of a regular grammar. We also show that rules extracted from networks trained with this pruning heuristic are more consistent with the rules to be learned. This performance improvement is obtained by pruning and retraining the networks. Simulations are shown for training and pruning a recurrent neural net on strings generated by two regular grammars, a randomly-generated 10-state grammar and an 8-state, triple-parity grammar. Further simulations indicate that this pruning method can have generalization performance superior to that obtained by training with weight decay.