Jensen-Shannon Divergence Research Articles

Entropic uncertainty relations and quantum coherence in the two-dimensional XXZ spin model with Dzyaloshinskii–Moriya interaction

Quantum renormalization group (QRG) is a tractable method for studying the criticalities of one-dimensional (1D) and two-dimensional (2D) many-body systems. By employing the QRG method, we first derive the effective Hamiltonian and QRG equations of a 2D XXZ model with Dzyaloshinskii–Moriya (DM) interaction analytically. The linear-entropy-based uncertainty, the quantum discord (QD), and the multipartite quantum coherence based on the square root of the quantum Jensen–Shannon divergence of the 2D XXZ model are then studied as the indicators of quantum phase transitions (QPTs). The nonanalytic and scaling behaviors of the uncertainty, QD and quantum coherence are also analyzed through numerical calculations. Moreover, we investigate the effect of the easy-axis anisotropy parameter and DM interaction on the QPT. We find that the uncertainty, QD, and quantum coherence can all be utilized to detect QPTs. Our findings could shed new light on the observable of the QPT of the many-body system with the uncertainty and quantum coherence, and enrich the application of QRG method to Heisenberg spin models.

Incipient Near Surface Cracks Characterization and Crack Size Estimation based on Jensen–Shannon Divergence and Wasserstein Distance

Incipient Near Surface Cracks Characterization and Crack Size Estimation based on Jensen–Shannon Divergence and Wasserstein Distance

A Random Focusing Method with Jensen–Shannon Divergence for Improving Deep Neural Network Performance Ensuring Architecture Consistency

Multiple hidden layers in deep neural networks perform non-linear transformations, enabling the extraction of meaningful features and the identification of relationships between input and output data. However, the gap between the training and real-world data can result in network overfitting, prompting the exploration of various preventive methods. The regularization technique called ’dropout’ is widely used for deep learning models to improve the training of robust and generalized features. During the training phase with dropout, neurons in a particular layer are randomly selected to be ignored for each input. This random exclusion of neurons encourages the network to depend on different subsets of neurons at different times, fostering robustness and reducing sensitivity to specific neurons. This study introduces a novel approach called random focusing, departing from complete neuron exclusion in dropout. The proposed random focusing selectively highlights random neurons during training, aiming for a smoother transition between training and inference phases while keeping network architecture consistent. This study also incorporates Jensen–Shannon Divergence to enhance the stability and efficacy of the random focusing method. Experimental validation across tasks like image classification and semantic segmentation demonstrates the adaptability of the proposed methods across different network architectures, including convolutional neural networks and transformers.

Mapping morphological cortical networks with joint probability distributions from multiple morphological features

Morphological features sourced from structural magnetic resonance imaging can be used to infer human brain connectivity. Although integrating different morphological features may theoretically be beneficial for obtaining more precise morphological connectivity networks (MCNs), the empirical evidence to support this supposition is scarce. Moreover, the incorporation of different morphological features remains an open question. In this study, we proposed a method to construct cortical MCNs based on multiple morphological features. Specifically, we adopted a multi-dimensional kernel density estimation algorithm to fit regional joint probability distributions (PDs) from different combinations of four morphological features, and estimated inter-regional similarity in the joint PDs via Jensen-Shannon divergence. We evaluated the method by comparing the resultant MCNs with those built based on different single morphological features in terms of topological organization, test-retest reliability, biological plausibility, and behavioral and cognitive relevance. We found that, compared to MCNs built based on different single morphological features, MCNs derived from multiple morphological features displayed less segregated, but more integrated network architecture and different hubs, had higher test-retest reliability, encompassed larger proportions of inter-hemispheric edges and edges between brain regions within the same cytoarchitectonic class, and explained more inter-individual variance in behavior and cognition. These findings were largely reproducible when different brain atlases were used for cortical parcellation. Further analysis of macaque MCNs revealed weak, but significant correlations with axonal connectivity from tract-tracing, independent of the number of morphological features. Altogether, this paper proposes a new method for integrating different morphological features, which will be beneficial for constructing MCNs.

Distribution-informed and wavelength-flexible data-driven photoacoustic oximetry.

Photoacoustic imaging (PAI) promises to measure spatially resolved blood oxygen saturation but suffers from a lack of accurate and robust spectral unmixing methods to deliver on this promise. Accurate blood oxygenation estimation could have important clinical applications from cancer detection to quantifying inflammation. We address the inflexibility of existing data-driven methods for estimating blood oxygenation in PAI by introducing a recurrent neural network architecture. We created 25 simulated training dataset variations to assess neural network performance. We used a long short-term memory network to implement a wavelength-flexible network architecture and proposed the Jensen-Shannon divergence to predict the most suitable training dataset. The network architecture can flexibly handle the input wavelengths and outperforms linear unmixing and the previously proposed learned spectral decoloring method. Small changes in the training data significantly affect the accuracy of our method, but we find that the Jensen-Shannon divergence correlates with the estimation error and is thus suitable for predicting the most appropriate training datasets for any given application. A flexible data-driven network architecture combined with the Jensen-Shannon divergence to predict the best training data set provides a promising direction that might enable robust data-driven photoacoustic oximetry for clinical use cases.

Simulating appliance-level household electricity data: Accounting for residential behavior and usage patterns in China

Data scarcity presents a significant challenge for improving deep learning algorithms in Non-intrusive Load Disaggregation. These algorithms heavily rely on extensive appliance electricity consumption data, which is expensive to gather through on-site measurements. Therefore, generating simulated data has emerged as a viable alternative. However, existing approaches often randomly combine appliance power, failing to capture realistic household electricity usage patterns. To address this, this study investigates the dynamics of household electricity consumption throughout a 24-h day, considering factors such as household size, age, occupation, and education. Markov chains and Variational Auto-encoder are employed as a data generation model to simulate electrical power consumption data. Appliance power consumption data are aggregated to represent the total household load by incorporating resident electricity consumption behavior. Comparing the generated data with actual data from a self-measurement dataset, the Jensen-Shannon divergence, RMSE, and MAE are utilized to assess their similarity, and the Entropy and the Coefficient of Variation are utilized to evaluate randomness and volatility. The results indicate a high level of distribution similarity between the generated and actual data. Moreover, the observed variations show the model's capability to produce diverse and rich datasets. Overall, this research demonstrates the value of simulated data created as a low-cost method for Non-intrusive Load Disaggregation.

Multi-Scale Object Detection in Remote Sensing Images Based on Feature Interaction and Gaussian Distribution

Remote sensing images are usually obtained from high-altitude observation. The spatial resolution of the images varies greatly and there are scale differences both between and within object classes, resulting in a diversified distribution of object scales. In order to solve these problems, we propose a novel object detection algorithm that maintains adaptability to multi-scale object detection based on feature interaction and Gaussian distribution in remote sensing images. The proposed multi-scale feature interaction model constructs feature interaction modules in the feature layer and spatial domain and combines them to fully utilize the spatial and semantic information of multi-level features. The proposed regression loss algorithm based on Gaussian distribution takes the normalized generalized Jensen–Shannon divergence with Gaussian angle loss as the regression loss function to ensure the scale invariance of the model. The experimental results demonstrate that our method achieves 77.29% mAP on the DOTA-v1.0 dataset and 97.95% mAP on the HRSC2016 dataset, which are, respectively, 1.12% and 1.41% higher than that of the baseline. These experimental results indicate the effectiveness of our method for object detection in remote sensing images.

Nonparametric CD-CAT for multiple-choice items: Item selection method and Q-optimality.

Computerized adaptive testing for cognitive diagnosis (CD-CAT) achieves remarkable estimation efficiency and accuracy by adaptively selecting and then administering items tailored to each examinee. The process of item selection stands as a pivotal component of a CD-CAT algorithm, with various methods having been developed for binary responses. However, multiple-choice (MC) items, an important item type that allows for the extraction of richer diagnostic information from incorrect answers, have been underemphasized. Currently, the Jensen-Shannon divergence (JSD) index introduced by Yigit etal. (Applied Psychological Measurement, 2019, 43, 388) is the only item selection method exclusively designed for MC items. However, the JSD index requires a large sample to calibrate item parameters, which may be infeasible when there is only a small or no calibration sample. To bridge this gap, the study first proposes a nonparametric item selection method for MC items (MC-NPS) by implementing novel discrimination power that measures an item's ability to effectively distinguish among different attribute profiles. A Q-optimal procedure for MC items is also developed to improve the classification during the initial phase of a CD-CAT algorithm. The effectiveness and efficiency of the two proposed algorithms were confirmed by simulation studies.

#2704 Effects of high dietary salt on immunome and microbiome composition – results from a randomized clinical trial

Abstract Background and Aims Increased dietary salt intake ranks among the most prominent nutritional risk factors worldwide and has been tied to arterial hypertension and all-cause mortality. The mechanisms responsible are incompletely understood and have recently been extended to include inflammatory and microbiome-associated mechanisms. Building on our preliminary data, the present study investigates whether a moderate increase in salt intake in healthy subjects primes host physiology to a more transiently unstable state, potentially leading to changes in the microbiome-immune axis. Method We conducted a prospective, randomized, double-blinded trial to evaluate the effect of high dietary salt in healthy participants (NCT03024567), where the intervention group (n = 19) was given 6 g of salt (NaCl capsules) in addition to the habitual salt intake, resulting in an average doubling of the salt intake recommended by expert societies. A placebo group (n = 19) received gelatin capsules for 14 days. Clinical parameters, fecal samples and PBMC were collected at baseline and day 14. Shotgun metagenomic sequencing, metabolomics (stool and serum), and single-cell sequencing (Cellular Indexing of Transcriptomes and Epitopes by Sequencing (CITEseq)) of whole PBMCs and CD4+ sorted T cells (∼150 k single cells) was performed. Results Microbiome analysis confirms previous findings showing an increased dissimilarity of microbiome composition (Jensen-Shannon Divergence) under a high-salt. Differential abundance of the microbiome's functional space, as quantified by functional gut-specific modules, picked up a salt-specific module (M00494, NatK-NatR (sodium extrusion) two-component regulatory system) in the salt intervention group. CITEseq revealed a significant reduction of circulating NK cells and an increase in classical monocytes as well as changes to several T cell subsets. Furthermore, within several immune cell subsets, we detected differently expressed genes (DEGs) solely in the salt group. We found 602 DEG in naïve conventional T cells (Tconv) and 202 DEGs in non-naïve Tconv, e.g. belonging to NF-kB and T cell receptor signaling pathways. Conclusion Our placebo-controlled study is the first to utilize different omics techniques to investigate and identify high-salt-induced changes in the microbiome and immunome in healthy individuals that may be relevant for health-to-disease transition.

An effective multi-source data fusion approach based on [formula omitted]-divergence in belief functions theory with applications to air target recognition and fault diagnosis

Belief functions theory (BFT) plays a critical role in addressing uncertainty and imprecision in multi-source data fusion. Unfortunately, the application of Dempster’s rule in BFT often produces unexpected and counterintuitive results when it encounters highly conflicting evidence. In this paper, we propose an enhanced belief α-divergence to quantify the discrepancies between the evidences by considering both belief and plausibility functions. We further present two symmetric versions of enhanced belief α-divergence. We demonstrate that the proposed symmetric divergences exhibit key properties including non-negativity, non-degeneracy and symmetry. Also, we analyze that the proposed symmetric divergences can be converted to χ2 divergence, Jeffreys divergence, Hellinger distance, Jensen–Shannon divergence and arithmetic–geometric divergence under the framework of BFT in some special cases. Leveraging the proposed symmetric divergences, a novel multi-source data fusion approach is proposed, which evaluates the credibility and informational volume of each evidence, offering a more nuanced understanding of their significance. To validate the proposed approach, we apply it to air target recognition and fault diagnosis scenarios, where our proposed approach showcases superior performance.

Novel Distance Measures of q-Rung Orthopair Fuzzy Sets and Their Applications

The q-rung orthopair fuzzy sets (q-ROFSs), a novel concept for processing vague information, offer a more potent and all-encompassing method compared to traditional fuzzy sets, intuitionistic fuzzy sets, and Pythagorean fuzzy sets. The inclusion of the parameter q allows for the q-rung orthopair fuzzy sets to capture a broader range of uncertainty of information. In this paper, we present two novel distance measures for q-ROFSs inspired by the Jensen–Shannon divergence, called DJS_2D and DJS_3D, and we analyze some properties they satisfy, such as non-degeneracy, symmetry, boundedness, and triangular inequality. Then, the normalized distance measures, called DJS_2D˜ and DJS_3D˜, are proposed and we verify their rationality through numerical experiments. Finally, we apply the proposed distance measures to practical scenarios, including pattern recognition and multicriteria decision-making, and the results demonstrate the effectiveness of the proposed distance measures.

Abstract 1012: Single-Cell RNA Sequencing Defines Focal Adhesion-Matrix Interactions In Smooth Muscle Cells In Thoracic Aortic Disease Triggered By BAPN Inhibiting Lysyl Oxidase

Background: Thoracic aortic aneurysms and/or dissections (TAD) are associated with pathogenic variants in genes that involved in either smooth muscle cell (SMC) contraction or extracellular matrix (ECM). The inability to maintain homeostasis between SMCs and ECM is pivotal in TAD pathogenesis. Focal adhesions (FA) act as mechanosensors in cells, transferring mechanical signals from the ECM to the intracellular contractile filaments and regulating downstream Rho/ROCK signaling. However, the landscape of SMC-ECM interactions, has not been fully explored at the early stages of TAD. Methods: C57BL/6J wild-type mice aged 21 days postnatal (P21) were administered the ECM cross-link blocker, BAPN, for up to four weeks to induce TAD. Proximal aortas were harvested for single-cell RNA sequencing at P35. Immunofluorescence and immunoblots were used to validate findings. Results: Single-cell RNA sequencing classified SMCs into 5 clusters marked by Rgs5 (SMC Rgs5 + ), high levels of Palld (SMC Palld + ), Tnnt2 (SMC Tnnt2 + , marker of second heart field-derived cells), Atf3 (SMC Atf3+ cluster), or no clear marker (SMC NOS ). The analysis of how BAPN exposure affects different cell types, based on the Jensen-Shannon divergence in the probability distribution between groups, demonstrated that SMCs, and specifically SMC Atf3 + and SMC NOS , are the most perturbed cell cluster by BAPN. Gene set enrichment analysis identified significantly dysregulated pathways of FA and Rho/ROCK signaling. The expression of Itga5 and Fn1 (fibronectin) were increased and Itga8 was decreased. An SMC-SMC interaction analysis found Fn1 interactions were the most upregulated interaction in 4 SMC clusters. When analyzing downstream pathways due to the changes in interaction, 80% of the top 25 interactions were related to integrins, including increased Fn1 interactions with Itga5-Itgb1 and Itgav-Itgb3. In SMC Rgs5 + cluster, the most affected interactions are related to TGF-β signaling. Conclusions: These findings suggest that, in the majority of aortic SMCs, mechanosensing through FAs, driven by fibronectin interactions with specific integrin receptor pairs, is activated with BAPN treatment, potentially contributing to the pathogenesis of TAD.

Multiparameter shallow-seismic waveform inversion based on the Jensen–Shannon divergence

SUMMARY Seismic full-waveform inversion (FWI) or waveform inversion (WI) has gained extensive attention as a cutting-edge imaging method, which is expected to reveal the high-resolution images of complex geological structures. In this paper, we regard each 1-D signal in the inversion system as a 1-D probability distribution, then use the Jensen–Shannon divergence from information theory to measure the discrepancy between the predicted and observed signals, and finally implement a novel 2-D multiparameter shallow-seismic WI (MSWI). Essentially, the novel approach achieves an implicit weighting along the time-axis for each 1-D adjoint source defined by the classical WI (CWI), thus enhancing the extra illumination for a deeper medium compared with the CWI. By evaluating the inversion results of the two-layer model and fault model, the reconstruction accuracy for S-wave velocity and density of the new method is increased by about 30 and 20 per cent compared with that of the CWI under the same conditions, respectively. The reconstruction performance for P-wave velocity of these two methods is almost equal. In addition, the new 2-D MSWI is also resilient to white Gaussian noise in the data. Numerically, the inversion system has almost the strongest sensitivities to the S-wave velocity and density, performing the poorest sensitivity to the P-wave velocity. Finally, we test the novel method with a detection case for a power tunnel.

Jensen–Shannon Divergence You Only Look Once: A Real‐Time Robotic Grasp Detection Network

In this article, the arbitrary‐oriented object detection problem with application in robotic grasping is addressed. A novel Jensen–Shannon divergence (JSD)– You Only Look Once (YOLO) model is proposed, which enables real‐time grasp detection with high performance. The one‐stage object detection network YOLOv5 is modified with a decoupled head, which solves the angle classification problem and rectangle parameter regression problem separately, such that the YOLOv5 network is applicable for robotic grasping and the detection accuracy is significantly improved. A circular smooth label angle classification method is proposed to tackle the boundary discontinuity problem in angle regression, and the periodicity of the angle prediction is guaranteed. A novel Jensen–Shannon intersection of union is designed to calculate the intersection over union of oriented rectangles, which aims to better measure the discrepancies between the prediction and the ground truth and to avoid the singularity problem when two rectangles are not overlapped. Extensive evaluation on the Cornell and visual manipulation relationship dataset datasets demonstrates the effectiveness of the JSD–YOLO model in general robotic grasp operations, with 99.7% and 95.7% image‐wise split accuracy, respectively.

A Unifying Generator Loss Function for Generative Adversarial Networks.

A unifying α-parametrized generator loss function is introduced for a dual-objective generative adversarial network (GAN) that uses a canonical (or classical) discriminator loss function such as the one in the original GAN (VanillaGAN) system. The generator loss function is based on a symmetric class probability estimation type function, Lα, and the resulting GAN system is termed Lα-GAN. Under an optimal discriminator, it is shown that the generator's optimization problem consists of minimizing a Jensen-fα-divergence, a natural generalization of the Jensen-Shannon divergence, where fα is a convex function expressed in terms of the loss function Lα. It is also demonstrated that this Lα-GAN problem recovers as special cases a number of GAN problems in the literature, including VanillaGAN, least squares GAN (LSGAN), least kth-order GAN (LkGAN), and the recently introduced (αD,αG)-GAN with αD=1. Finally, experimental results are provided for three datasets-MNIST, CIFAR-10, and Stacked MNIST-to illustrate the performance of various examples of the Lα-GAN system.

Abstract MP01: Life-Course Socioeconomic Status and the Gut Microbiome in the Hispanic Community Health Study/Study of Latinos (HCHS/SOL)

Introduction: Socioeconomic status (SES) in childhood and beyond may influence the gut microbiome (GMB), with implications for disease risk. Studies evaluating the relationship between life-course SES and the gut microbiome (GMB) are sparse, particularly among Hispanic/Latino individuals, who have a high prevalence of low SES. Hypothesis: Low life-course SES indicators are associated with less diversity and overall composition of the gut microbiome in United States (U.S.) Hispanic/Latino adults. Methods: The Hispanic Community Health Study/Study of Latinos (HCHS/SOL) is a multi-site population-based cohort study conducted in the U.S. Childhood sanitation conditions (plumbing, sewer/septic tank), childhood SES (parental education and childhood hardship), adulthood neighborhood SES, and adulthood SES (education, income, and economic hardship) were used as life-course SES indicators. Shotgun sequencing was performed on stool samples (n=1057). Alpha-diversity was measured by the Shannon diversity index. Permutational multivariate analysis of variance was used to assess the association of SES indicators with overall microbiome composition, as measured using the Jensen-Shannon Divergence distance. Analysis of Compositions of Microbiomes was used to identify associations of life-course SES indicators with GMB species, adjusting for age, sex, BMI, U.S. or foreign-born, study center, Hispanic/Latino background, type of stool, diet quality, smoking, alcohol consumption, depressive and anxiety symptoms, diabetes, hypertension, cardiovascular disease, dyslipidemia, and antibiotic use. Results: Low sanitation and low SES during childhood were reported by 32% (342 of 1053) and 37% (330 of 884) of participants, respectively. In adulthood, 56% (585 of 1053) lived in a low SES neighborhood and 45% (447 of 1002) had low SES. Life-course SES indicators were not associated with GMB diversity. However, overall microbiome composition differed significantly according to childhood SES (R2 =0.28%, P = 0.04). Low childhood sanitation was associated with higher abundance of Ruminococcus torques and Clostridium lactatifermentans , and depleted abundance of Prevotella species. Low childhood SES was associated with higher abundance of species from genus Streptococcus (S. salivarius, S. parasanguinis, and S. infantis), Veillonella parvula, Prevotella stercorea, and Roseburia intestinalis. Low adulthood SES was associated with depleted abundance of species from Bacteroides, and living in a low adulthood neighborhood SES was associated with depleted abundance of species from Rikenella . Conclusion: Low childhood SES was associated with an altered GMB composition in U.S. Hispanic/Latino adults. Early-life SES may have long-term effects on GMB composition, underscoring another biological mechanism linking early childhood factors to adulthood disease.

Choice of Metric Divergence in Genome Sequence Comparison.

The paper introduces a novel probability descriptor for genome sequence comparison, employing a generalized form of Jensen-Shannon divergence. This divergence metric stems from a one-parameter family, comprising fractions up to a maximum value of half. Utilizing this metric as a distance measure, a distance matrix is computed for the new probability descriptor, shaping Phylogenetic trees via the neighbor-joining method. Initial exploration involves setting the parameter at half for various species. Assessing the impact of parameter variation, trees drawn at different parameter values (half, one-fourth, one-eighth). However, measurement scales decrease with parameter value increments, with higher similarity accuracy corresponding to lower scale values. Ultimately, the highest accuracy aligns with the maximum parameter value of half. Comparative analyses against previous methods, evaluating via Symmetric Distance (SD) values and rationalized perception, consistently favor the present approach's results. Notably, outcomes at the maximum parameter value exhibit the most accuracy, validating the method's efficacy against earlier approaches.

Sample Size Optimization for Digital Soil Mapping: An Empirical Example

In the evolving field of digital soil mapping (DSM), the determination of sample size remains a pivotal challenge, particularly for large-scale regional projects. We introduced the Jensen-Shannon Divergence (DJS), a novel tool recently applied to DSM, to determine optimal sample sizes for a 2790 km2 area in Ontario, Canada. Utilizing 1791 observations, we generated maps for cation exchange capacity (CEC), clay content, pH, and soil organic carbon (SOC). We then assessed sample sets ranging from 50 to 4000 through conditioned Latin hypercube sampling (cLHS), feature space coverage sampling (FSCS), and simple random sampling (SRS) to calibrate random forest models, analyzing performance via concordance correlation coefficient and root mean square error. Findings reveal DJS as a robust estimator for optimal sample sizes—865 for cLHS, 874 for FSCS, and 869 for SRS, with property-specific optimal sizes indicating the potential for enhanced DSM accuracy. This methodology facilitates a strategic approach to sample size determination, significantly improving the precision of large-scale soil mapping. Conclusively, our research validates the utility of DJS in DSM, offering a scalable solution. This advancement holds considerable promise for improving soil management and sustainability practices, underpinning the critical role of precise soil data in agricultural productivity and environmental conservation.

Evaluating the quality of visual explanations on chest X-ray images for thorax diseases classification

AbstractDeep learning models are extensively used but often lack transparency due to their complex internal mechanics. To bridge this gap, the field of explainable AI (XAI) strives to make these models more interpretable. However, a significant obstacle in XAI is the absence of quantifiable metrics for evaluating explanation quality. Existing techniques, reliant on manual assessment or inadequate metrics, face limitations in scalability, reproducibility, and trustworthiness. Recognizing these issues, the current study specifically addresses the quality assessment of visual explanations in medical imaging, where interpretability profoundly influences diagnostic accuracy and trust in AI-assisted decisions. Introducing novel criteria such as informativeness, localization, coverage, multi-target capturing, and proportionality, this work presents a comprehensive method for the objective assessment of various explainability algorithms. These newly introduced criteria aid in identifying optimal evaluation metrics. The study expands the domain’s analytical toolkit by examining existing metrics, which have been prevalent in recent works for similar applications, and proposing new ones. Rigorous analysis led to selecting Jensen–Shannon divergence (JS_DIV) as the most effective metric for visual explanation quality. Applied to the multi-label, multi-class diagnosis of thoracic diseases using a trained classifier on the CheXpert dataset, local interpretable model-agnostic explanations (LIME) with diverse segmentation strategies interpret the classifier’s decisions. A qualitative analysis on an unseen subset of the VinDr-CXR dataset evaluates these metrics, confirming JS_DIV’s superiority. The subsequent quantitative analysis optimizes LIME’s hyper-parameters and benchmarks its performance across various segmentation algorithms, underscoring the utility of an objective assessment metric in practical applications.

On Inactivation of the Coronavirus Main Protease.

A deeper understanding of the inactive conformations of the coronavirus main protease (MPro) could inform the design of allosteric drugs. Based on extensive molecular dynamics simulations, we built a Markov State Model to investigate structural changes that can inactivate the SARS-CoV-2 MPro. In a subset of structures, one subunit of the homodimer assumes an inactive conformation that resembles an inactive crystal structure. However, contradicting the widely held half-of-sites activity hypothesis, the most populated enzyme structures have two active subunits. We then used transition path theory (TPT) and the Jensen-Shannon Divergence (JSD) to pinpoint residues involved in the inactivation process. A π stack between Phe140 and His163 is a key feature that can distinguish active and inactive conformations of MPro. Each subunit has unique inactive conformations stabilized by π stacking interactions involving residues Phe140, Tyr118, His163, and His172, a hydrogen bonding network centered around His163 and His172, and a modified network of interactions in the dimer interface. The importance of these residues in maintaining an active structure explains the sensitivity of enzymatic activity to site-directed mutagenesis.