Hidden Markov Model Research Articles

"Multimodal Sleep Signal Tensor Decomposition and Hidden Markov Modeling for Temazepam-Induced Anomalies Across Age Groups".

Recent advances in multimodal signal analysis enable the identification of subtle drug-induced anomalies in sleep that traditional methods often miss. We develop and introduce the Dynamic Representation of Multimodal Activity and Markov States (DREAMS) framework, which embeds explainable artificial intelligence (XAI) techniques to model hidden state transitions during sleep using tensorized EEG, EMG, and EOG signals from 22 subjects across three age groups (18-29, 30-49, and 50-66 years). By combining Tucker decomposition with probabilistic Hidden Markov Modeling, we quantified age-specific, temazepam-induced hidden states and significant differences in transition probabilities. Jensen-Shannon Divergence (JSD) was employed to assess variability in hidden state transitions, with older subjects (50-66 years) under temazepam displaying heightened transition variability and network instability as indicated by a 48.57% increase in JSD (from 0.35 to 0.52) and reductions in network density by 12.5% (from 0.48 to 0.42) and modularity by 21.88% (from 0.32 to 0.25). These changes reflect temazepam's disruptive impact on sleep architecture in older adults, aligning with known age-related declines in sleep stability and pharmacological sensitivity. In contrast, younger subjects exhibited lower divergence and retained relatively stable, cyclical transition patterns. Anomaly scores further quantified deviations in state transitions, with older subjects showing increased transition uncertainty and marked deviations in REM-like to NREM state transitions. This XAI-driven framework provides transparent, age-specific insights into temazepam's impact on sleep dynamics, going beyond traditional methods by identifying subtle, pharmacologically induced changes in sleep stage transitions that would otherwise be missed. DREAMS supports the development of personalized interventions based on sleep transition variability across age groups, offering a powerful tool to understand temazepam's age-dependent effects on sleep architecture.

Similar trajectories of psychological states predict marital satisfaction

Abstract Modern mate selection theories suggest that people are more likely to marry someone similar to themselves in terms of numerous attributes. Recent research has demonstrated a positive relationship between marital satisfaction and inter-subject correlation (ISC) of neural responses while viewing movies in married couples. Nevertheless, conventional ISC methods solely capture information about similarity in the temporal evolution of region-averaged neural responses, disregarding nuanced spatially distributed response topographies. Here, we integrated ISC and multi-voxel pattern (MVP) analysis to capitalize inter-subject trajectory similarity (ISTS) of MVP. We demonstrated that married couples showed significantly higher ISTS than randomly selected pairs, during movie viewing and resting-state. The ISTS was particularly positively associated with marital satisfaction in married couples while viewing movies. In order to investigate latent “psychological states” characterized by relatively stable patterns of MVP, a hidden Markov model was used to segment the neural events in married couples during viewing movies. We found the ISTS within manually defined events was a strong predictor of marital satisfaction. These results suggest that married couples with high-level marital satisfaction may experience similar trajectories of mental states when exposed to a common marital-related stimulus, and extend our understanding of the neurobiological signatures of intimate relationship.

Noise reduction of low-dose electron holograms using the wavelet hidden Markov model.

The precision in electron holography studies on electrostatic and magnetic fields depends on the image quality of an electron hologram. Enhancing the image quality of electron holograms is essential for the comprehensive analysis of weak electromagnetic fields; however, extended electron beam irradiation can lead to undesirable radiation damage and contamination. Recent studies have demonstrated that noise reduction using the wavelet hidden Markov model (WHMM) can improve the precision of phase analysis for limited thin-foiled crystals. In this study, we examine the effects of WHMM-based denoising on the electron holography data of weakly charged nanoparticles collected under low-electron-dose conditions. The results indicate that effective noise reduction with the WHMM allows for a reduction in the magnitude of the electron dose by approximately half relative to data collection without WHMM denoising, while maintaining the same level of the charge determination precision: less than one elementary charge. Notably, at a low electron dose of 0.40 e-/pixel, WHMM denoising enables the clear visualization of a weak stray electric field outside a charged latex sphere. This method offers significant advantages for electron holography studies of electron-beam-sensitive materials requiring minimal time for electron exposure.

Uncovering dissipation from coarse observables: A case study of a random walk with unobserved internal states.

Inferring underlying microscopic dynamics from low-dimensional experimental signals is a central problem in physics, chemistry, and biology. As a trade-off between molecular complexity and the low-dimensional nature of experimental data, mesoscopic descriptions such as the Markovian master equation are commonly used. The states in such descriptions usually include multiple microscopic states, and the ensuing coarse-grained dynamics are generally non-Markovian. It is frequently assumed that such dynamics can nevertheless be described as a Markov process because of the timescale separation between slow transitions from one observed coarse state to another and the fast interconversion within such states. Here, we use a simple model of a molecular motor with unobserved internal states to highlight that (1) dissipation estimated from the observed coarse dynamics may significantly underestimate microscopic dissipation even in the presence of timescale separation and even when mesoscopic states do not contain dissipative cycles and (2) timescale separation is not necessarily required for the Markov approximation to give the exact entropy production, provided that certain constraints on the microscopic rates are satisfied. When the Markov approximation is inadequate, we discuss whether including memory effects can improve the estimate. Surprisingly, when we do so in a "model-free" way by computing the Kullback-Leibler divergence between the observed probability distributions of forward trajectories and their time reverses, this leads to poorer estimates of entropy production. Finally, we argue that alternative approaches, such as hidden Markov models, may uncover the dissipative nature of the microscopic dynamics even when the observed coarse trajectories are completely time-reversible.

Tripartite organization of brain state dynamics underlying spoken narrative comprehension.

Speech comprehension involves the dynamic interplay of multiple cognitive processes, from basic sound perception, to linguistic encoding, and finally to complex semantic-conceptual interpretations. How the brain handles the diverse streams of information processing remains poorly understood. Applying Hidden Markov Modeling to fMRI data obtained during spoken narrative comprehension, we reveal that the whole brain networks predominantly oscillate within a tripartite latent state space. These states are, respectively, characterized by high activities in the sensory-motor (State #1), bilateral temporal (State #2), and default mode networks (DMN; State #3) regions, with State #2 acting as a transitional hub. The three states are selectively modulated by the acoustic, word-level semantic, and clause-level semantic properties of the narrative. Moreover, the alignment with both the best performer and the group-mean in brain state expression can predict participants' narrative comprehension scores measured from the post-scan recall. These results are reproducible with different brain network atlas and generalizable to two datasets consisting of young and older adults. Our study suggests that the brain underlies narrative comprehension by switching through a tripartite state space, with each state probably dedicated to a specific component of language faculty, and effective narrative comprehension relies on engaging those states in a timely manner.

Polariton lattices as binarized neuromorphic networks

We introduce a novel neuromorphic network architecture based on a lattice of exciton-polariton condensates, intricately interconnected and energized through nonresonant optical pumping. The network employs a binary framework, where each neuron, facilitated by the spatial coherence of pairwise coupled condensates, performs binary operations. This coherence, emerging from the ballistic propagation of polaritons, ensures efficient, network-wide communication. The binary neuron switching mechanism, driven by the nonlinear repulsion through the excitonic component of polaritons, offers computational efficiency and scalability advantages over continuous weight neural networks. Our network enables parallel processing, enhancing computational speed compared to sequential or pulse-coded binary systems. The system’s performance was evaluated using diverse datasets, including the MNIST dataset for image recognition and the Speech Commands dataset for voice recognition tasks. In both scenarios, the proposed system demonstrates the potential to outperform existing polaritonic neuromorphic systems. For image recognition, this is evidenced by an impressive predicted classification accuracy of up to 97.5%. In voice recognition, the system achieved a classification accuracy of about 68% for the ten-class subset, surpassing the performance of conventional benchmark, the Hidden Markov Model with Gaussian Mixture Model.

The Gaussian-Linear Hidden Markov model: a Python package

Abstract We propose the Gaussian-Linear Hidden Markov model (GLHMM), a generalisation of different types of HMMs commonly used in neuroscience. In short, the GLHMM is a general framework where linear regression is used to flexibly parameterise the Gaussian state distribution, thereby accommodating a wide range of uses —including unsupervised, encoding and decoding models. GLHMM is available as a Python toolbox with an emphasis on statistical testing and out-of-sample prediction —i.e. aimed at finding and characterising brain-behaviour associations. The toolbox uses a stochastic variational inference approach, enabling it to handle large data sets at reasonable computational time. The GLHMM can work with various types of data, including animal recordings or non-brain data, and is suitable for a broad range of experimental paradigms. For demonstration, we show examples with fMRI, local field potential, electrocorticography, magnetoencephalography and pupillometry.

Asynchronous passive control under event‐triggered scheme for discrete‐time Markov jump systems with complex probabilities

AbstractThis paper deals with the asynchronous passive control based on event‐triggered scheme for Markov jump systems with complex probabilities. A hidden Markov model with uncertain and unknown conditional probabilities is presented to describe the asynchronization. The complex probabilities consider that both transition probability matrix and conditional probability matrix in hidden Markov model contain uncertain and unknown elements simultaneously, with the former governing the jump of system mode and the latter representing the asynchronous relation between the system mode and the controller mode. Asynchronous control strategies based on hidden Markov model with complex probabilities are proposed under event‐triggered scheme. Furthermore, sufficient conditions are established to guarantee the stochastic passivity, under which the mode‐dependent controller gains and weighting matrices in the event‐triggered scheme are co‐designed. Theoretical results are finally verified via an example.

Pretrained Deep Neural Network Kin-SiM for Single-Molecule FRET Trace Idealization.

Single-molecule fluorescence resonance energy transfer (smFRET) has emerged as a pivotal technique for probing biomolecular dynamics over time at nanometer scales. Quantitative analyses of smFRET time traces remain challenging due to confounding factors such as low signal-to-noise ratios, photophysical effects such as bleaching and blinking, and the complexity of modeling the underlying biomolecular states and kinetics. The dynamic distance information shaping the smFRET trace powerfully uncovers even transient conformational changes in single biomolecules both at or far from equilibrium, relying on trace idealization to identify specific interconverting states. Conventional trace idealization methods based on hidden Markov models (HMMs) require substantial a priori knowledge of the system under study, manual intervention, and assumptions about the number of states and transition probabilities. Here, we present a deep learning framework using long short-term memory (LSTM) to automate the trace idealization, termed Kin-SiM. Our approach employs neural networks pretrained on simulated data to learn high-order correlations in the multidimensional FRET trajectories. Without user input of Markovian assumptions, the trained LSTM networks directly idealize the FRET traces to extract the number of underlying biomolecular states, their interstate dynamics, and associated kinetic parameters. On benchmark smFRET data sets, Kin-SiM achieves a performance similar to conventional HMM-based methods but with less hands-on time and lower risk of bias. We further systematically evaluate the key training factors that affect network performance to define the correct hyperparameter tuning for applying deep neural networks to smFRET data analyses.

Capturing crisis dynamics: a novel personalized approach using multilevel hidden Markov modeling.

Prevention of (suicidal) crisis starts with appreciating its dynamics. However, crisis is a complex multidimensional phenomenon and how it evolves over time is still poorly understood. This study aims to clarify crisis dynamics by clustering fluctuations in the interplay of cognitive, affective, and behavioral (CAB) crisis factors within persons over time into latent states. To allow for fine grained information on CAB factors over a prolonged period of time, ecological momentary assessment data comprised of self-report questionnaires (3 × daily) on five CAB symptoms (self-control, negative affect, contact avoidance, contact desire and suicidal ideation) was collected in twenty-six patients (60 measurements per patient). Empirically-derived crisis states and personalized state dynamics were isolated utilizing multilevel hidden Markov models. In this proof-of-concept study, four distinct and ascending CAB-based crisis states were derived. At the sample level, remaining within the current CAB crisis state from one five-hour interval to the next was most likely, with staying likeliness decreasing with ascending states. When residing in CAB crisis state 2 or higher, it was least likely to transition back to CAB crisis state 1. However, large patient heterogeneity was observed both in the tendency to remain within a certain CAB crisis state and transitioning between crisis states. The uncovered crisis states using multilevel HMM quantify and visualize the pattern of crisis trajectories at the patient individual level. The observed differences between patients underlines the need for future innovation in personalized crisis prevention, and statistical models that facilitate such a personalized approach.

Study of dynamic brain function in irritable bowel syndrome via Hidden Markov Modeling.

Irritable bowel syndrome (IBS) is a common bowel-brain interaction disorder whose pathogenesis is unclear. Many studies have investigated abnormal changes in brain function in IBS patients. In this study, we analyzed the dynamic changes in brain function in IBS patients using a Hidden Markov Model (HMM). Resting-state functional magnetic resonance imaging (rs-fMRI) data and the clinical characteristics of 35 patients with IBS and 31 healthy controls (HCs) were collected. The rs-fMRI data of all participants were analyzed using HMM to identify recurrent brain activity states that evolve over time during the resting state. Additionally, the temporal properties of these HMM states and their correlations with clinical scale scores were examined. This study utilized the Hidden Markov Model (HMM) method to identify six distinct HMM states. Significant differences in fractional occupancy (FO) and lifetime (LT) were observed in states 5 and 6 between the IBS and HCs. The state transition probabilities differed between IBS and HCs, with an increased probability of transitioning from state 2 to state 6 in IBS patients. The reconfiguration of HMM states over time scales in IBS patients was associated with abnormal activity in the default mode network (DMN), sensorimotor network (SMN), and cingulo-opercular network (CON). This study offers novel insights into the dynamic reorganization of brain activity patterns in IBS and elucidates potential links between these patterns and IBS-related emotional regulation and symptom experience, thereby contributing to a deeper understanding of the neural mechanisms underlying IBS.

Gravity-aided navigation using Viterbi map matching algorithm

Abstract In Global Navigation Satellite Systems (GNSS)-denied environments, aiding a vehicle's inertial navigation system (INS) is crucial to reducing the accumulated navigation drift caused by sensor errors (e.g. bias and noise). One potential solution is to use measurements of gravity as an aiding source. The measurements are matched to a geo-referenced map of Earth's gravity to estimate the vehicle's position. In this paper, we propose a novel formulation of the map matching problem using a hidden Markov model (HMM). Specifically, we treat the spatial cells of the map as the hidden states of the HMM and present a Viterbi style algorithm to estimate the most likely sequence of states, i.e. most likely sequence of vehicle positions, that results in the sequence of observed gravity measurements. Using a realistic gravity map, we demonstrate the accuracy of our Viterbi map matching algorithm in a navigation scenario and illustrate its robustness compared with existing methods.

Artificial Intelligence Tools Addressing Challenges of Cancer Progression Due to Antimicrobial Resistance in Pathogenic Biofilm Systems

Infections, inflammation, and progression of multifactorial diseases are found to be integratively linked, including most Cancers. Dysfunctional microbiomes are also associated with several cancers in their tumor microenvironments. Antimicrobial peptides (AMPs) are short, positively charged peptides found in a diverse range of species, including bacteria and humans. As host defense peptides, they can destroy pathogenic infections, particularly those that are multidrug resistant. AMPs have raised hopes in the biomedical and pharmaceutical industries as fresh non-antibiotic strategies for combating infectious diseases. However, in vitro and in vivo verification of AMPs is problematic and may miss new antimicrobial drugs. Creating computational methods for quick and precise identification of AMPs and their functional forms is critical for developing new and more effective antimicrobial drugs. Machine learning techniques were recently discovered effective at mining, predicting, and producing efficient antimicrobial peptides from a large AMP database. We reviewed 76 articles, after following literature search rubrics to come to the following conclusions. Distance metric-constant K-based nearest neighbor algorithms (KNN), hidden Markov models (HMMs), support vector machine models (SVMs), random forest models (RFs), decision tree models, and deep neural network (DNN)-based models are some of the most popular AI tools for detecting antimicrobial activity in peptide sequence-derived structure and function. Knowledge graphs can further assist in identifying hub genes and antimicrobial peptides that target and block quorum sensing (QS) signals within the microbial networks. In conclusion, we state that currently no single AI method has been found appropriate for AMP discovery and accurately capable of predicting high-efficacy AMPs. Our current literature review and analysis identify cutting-edge algorithms or innovations that might be included in hybrid machine-learning approaches for the most effective AMP identification, creation, and prediction. Non-peptide, natural molecule-based approaches to AMR reduction are also being studied for development, with natural peptide scaffolds serving as the foundation.

State-Dependent Missingness in Hidden Markov Models, with an Application to Drop-Out in a Clinical Trial

Abstract Establishing the effectiveness of treatments for psychopathology requires accurate models of its progression over time and the factors that impact it. Longitudinal data is however fraught with missingness, hindering accurate modeling. We re-analyse data on schizophrenia severity in a clinical trial using hidden Markov models (HMMs). We consider missing data in HMMs with a focus on situations where data is missing not at random (MNAR) and missingness depends on the latent states, allowing symptom severity to indirectly impact probability of missingness. In simulations, we show that including a submodel for state-dependent missingness reduces bias when data is MNAR and state-dependent, whilst not reducing accuracy when data is missing-at-random (MAR). When missingness depends on time, a model that allows missingness to be both state- and time-dependent is unbiased. Overall, these results show that modelling missingness as state-dependent and including relevant covariates is a useful strategy in applications of HMMs to time-series with missing data. Applying the model to data from a clinical trial, we find that drop-out is more likely for patients with less severe symptoms, which may lead to a biased assessment of treatment effectiveness.

A sticky Poisson Hidden Markov Model for solving the problem of over-segmentation and rapid state switching in cortical datasets.

The application of hidden Markov models (HMMs) to neural data has uncovered hidden states and signatures of neural dynamics that are relevant for sensory and cognitive processes. However, training an HMM on cortical data requires a careful handling of model selection, since models with more numerous hidden states generally have a higher likelihood on new (unseen) data. A potentially related problem is the occurrence of very rapid state switching after decoding the data with an HMM. The first problem can lead to overfitting and over-segmentation of the data. The second problem is due to intermediate-to-low self-transition probabilities and is at odds with many reports that hidden states in cortex tend to last from hundred of milliseconds to seconds. Here, we show that we can alleviate both problems by regularizing a Poisson-HMM during training so as to enforce large self-transition probabilities. We call this algorithm the 'sticky Poisson-HMM' (sPHMM). When used together with the Bayesian Information Criterion for model selection, the sPHMM successfully eliminates rapid state switching, outperforming an alternative strategy based on an HMM with a large prior on the self-transition probabilities. The sPHMM also captures the ground truth in surrogate datasets built to resemble the statistical properties of the experimental data.

Tracking the Ghosts of the Himalayas: Snow Leopard Conservation Insights From Satellite Collar Data.

This study presents the first movement analysis of snow leopards (Panthera uncia) using satellite telemetry data, focusing on the northeastern Himalayas of Nepal. By examining GPS-based satellite collar data between 2013 and 2017 from five collared snow leopards (effectively three individuals), the research uncovered distinct movement patterns, activity budgeting and home range utilisation from one adult male and two sub adult females. Hidden Markov models (HMMs) revealed three behavioural states based on the movement patterns-slow (indicative of resting), moderateand fast (associated with travelling) and demonstrated that the time of day influenced their behavioural state. While adult males exhibited behaviour focused on moderately active states, juvenile females presented behaviour focused on highly active states. Home ranges, estimated over a 5-21 month tracking period, were larger than those observed in previously studied snow leopards and included crossings of international boundaries from Nepal into China and India. These relatively large home ranges may be attributed to the rugged terrain and scarce resources within the study area. This research suggested that movement patterns and home range sizes might differ between male and female snow leopards, which may indicatedifferent ecological needs and resource-use techniques. Furthermore, this study provides reliable information on snow leopards from the telemetry data and links it to conservation implications in northeastern Nepal to ensure their long-term survival, promote coexistence and foster cross-border collaboration.

Application of Video Game Algorithm Based on Deep Q-Network Learning in Music Rhythm Teaching

The difficulty of music rhythm teaching makes its teaching effect limited. Some game algorithms show problems of stuttery and unclear picture when they are introduced into teaching, which greatly restricts the development of teaching content. Therefore, it is proposed to activate the deep Q network and improve the filter to improve the processing performance of game image data and the integrity of information retention. The model fuses the extracted features and processes them through the neural network. According to the performance characteristics and forms of music rhythm, constant Q transformation, dynamic programming algorithm and hidden Markov model are proposed to analyze and design music signal, beat point analysis and signal traveling speed, so as to realize the identification and matching of music rhythm. The performance test and application analysis of the proposed fusion improvement algorithm are carried out: the video game algorithm can effectively extract data feature information, and its accuracy rate is basically 90% or above. The improved Deep Q-network (DQN) and Artificial Fish Swarms Algorithm (AFSA) algorithms can maintain an accuracy of over 75%, followed by the ABC algorithm, which maintains an accuracy of over 70%, but exhibits significant fluctuations in the curve changes and nodes. The recognition accuracy of music signal feature extraction is over 70%, and it is less affected by signal-to-noise ratio. The improved activation function showed that the highest game score exceeded 10 points, while the scores of Rectified Linear Unit (ReLU) function and Hyperbolic tangent function (Tanh) function were basically hovering below 5 points, and the overall score curve fluctuated significantly. The recognition accuracy of music signal features extraction is above 70% and is less affected by Signal to Noise Ratio. The algorithm has a good effect of music rhythm teaching, the students' interest in classroom learning is more than 45%, and the learning weakness is less. The average score of students in music style judgment, rhythm division and index is 85 points and 90 points respectively, which is much higher than the 40 points and 63 points of traditional teaching. The proposed algorithm can better combine video games with music teaching and help students improve their learning effect under the gamification teaching method, and has good application value

Enhancing bridge performance assessment with a hybrid attention-based long short-term memory and hidden Markov model using sparse inspection data

Enhancing bridge performance assessment with a hybrid attention-based long short-term memory and hidden Markov model using sparse inspection data

Social bot detection using variational generative adversarial networks with hidden Markov models in Twitter network

Social bot detection using variational generative adversarial networks with hidden Markov models in Twitter network

Annotation of RxLR Effectors in Oomycete Genomes.

Pathogens have evolved effector proteins to suppress host immunity and facilitate plant infections. RxLR effectors are small, secreted effector proteins with conserved RxLR and dEER amino acid motifs at the N terminus and highly variable C termini and are commonly found in oomycete species. We provide computational approaches to annotate RxLR candidate effector genes in a genome assembly in FASTA format with an available GFF file. Hidden Markov Modeling (HHM) is used in combination with regular expressions to search for RxLR and EER amino acid patterns.