ArXiv最新文献

{"title":"Unraveling the Geometry of Visual Relational Reasoning.","authors":"Jiaqi Shang, Gabriel Kreiman, Haim Sompolinsky","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>Humans and other animals readily generalize abstract relations, such as recognizing <i>constant</i> in shape or color, whereas neural networks struggle. To investigate how neural networks generalize abstract relations, we introduce <i>SimplifiedRPM</i>, a novel benchmark for systematic evaluation. In parallel, we conduct human experiments to benchmark relational difficulty, enabling direct model-human comparisons. Testing four architectures-ResNet-50, Vision Transformer, Wild Relation Network, and Scattering Compositional Learner (SCL)-we find that SCL best aligns with human behavior and generalizes best. Building on a geometric theory of neural representations, we show representational geometries that predict generalization. Layer-wise analysis reveals distinct relational reasoning strategies across models and suggests a trade-off where unseen rule representations compress into training-shaped subspaces. Guided by our geometric perspective, we propose and evaluate SNRloss, a novel objective balancing representation geometry. Our findings offer geometric insights into how neural networks generalize abstract relations, paving the way for more human-like visual reasoning in AI.</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11888560/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143588444","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From FAIR to CURE: Guidelines for Computational Models of Biological Systems.","authors":"Herbert M Sauro, Eran Agmon, Michael L Blinov, John H Gennari, Joe Hellerstein, Adel Heydarabadipour, Peter Hunter, Bartholomew E Jardine, Elebeoba May, David P Nickerson, Lucian P Smith, Gary D Bader, Frank Bergmann, Patrick M Boyle, Andreas Dräger, James R Faeder, Song Feng, Juliana Freire, Fabian Fröhlich, James A Glazier, Thomas E Gorochowski, Tomas Helikar, Stefan Hoops, Princess Imoukhuede, Sarah M Keating, Matthias Konig, Reinhard Laubenbacher, Leslie M Loew, Carlos F Lopez, William W Lytton, Andrew McCulloch, Pedro Mendes, Chris J Myers, Jerry G Myers, Lealem Mulugeta, Anna Niarakis, David D van Niekerk, Brett G Olivier, Alexander A Patrie, Ellen M Quardokus, Nicole Radde, Johann M Rohwer, Sven Sahle, James C Schaff, T J Sego, Janis Shin, Jacky L Snoep, Rajanikanth Vadigepalli, H Steve Wiley, Dagmar Waltemath, Ion Moraru","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>Guidelines for managing scientific data have been established under the FAIR principles requiring that data be Findable, Accessible, Interoperable, and Reusable. In many scientific disciplines, especially computational biology, both data and <i>models</i> are key to progress. For this reason, and recognizing that such models are a very special type of \"data\", we argue that computational models, especially mechanistic models prevalent in medicine, physiology and systems biology, deserve a complementary set of guidelines. We propose the CURE principles, emphasizing that models should be Credible, Understandable, Reproducible, and Extensible. We delve into each principle, discussing verification, validation, and uncertainty quantification for model credibility; the clarity of model descriptions and annotations for understandability; adherence to standards and open science practices for reproducibility; and the use of open standards and modular code for extensibility and reuse. We outline recommended and baseline requirements for each aspect of CURE, aiming to enhance the impact and trustworthiness of computational models, particularly in biomedical applications where credibility is paramount. Our perspective underscores the need for a more disciplined approach to modeling, aligning with emerging trends such as Digital Twins and emphasizing the importance of data and modeling standards for interoperability and reuse. Finally, we emphasize that given the non-trivial effort required to implement the guidelines, the community moves to automate as many of the guidelines as possible.</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11875277/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143544850","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Bayesian Parameter Inference and Uncertainty Quantification for a Computational Pulmonary Hemodynamics Model Using Gaussian Processes.","authors":"Amirreza Kachabi, Sofia Altieri Correa, Naomi C Chesler, Mitchel J Colebank","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>Patient-specific modeling is a valuable tool in cardiovascular disease research, offering insights beyond what current clinical equipment can measure. Given the limitations of available clinical data, models that incorporate uncertainty can provide clinicians with better guidance for tailored treatments. However, such modeling must align with clinical time frameworks to ensure practical applicability. In this study, we employ a one-dimensional fluid dynamics model integrated with data from a canine model of chronic thromboembolic pulmonary hypertension (CTEPH) to investigate microvascular disease, which is believed to involve complex mechanisms. To enhance computational efficiency during model calibration, we implement a Gaussian process emulator. This approach enables us to explore the relationship between disease severity and microvascular parameters, offering new insights into the progression and treatment of CTEPH in a timeframe that is compatible with a reasonable clinical timeframe.</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11875295/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143544213","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Finite Element Analysis Model for Magnetomotive Ultrasound Elastometry Magnet Design with Experimental Validation.","authors":"Jacquelline Nyakunu, Christopher T Piatnichouk, Henry C Russell, Niels J van Duijnhoven, Benjamin E Levy","doi":"","DOIUrl":"","url":null,"abstract":"<p><strong>Objective: </strong>Magnetomotive ultrasound (MMUS) using magnetic nanoparticle contrast agents has shown promise for thrombosis imaging and quantitative elastometry via magnetomotive resonant acoustic spectroscopy (MRAS). Young's modulus measurements of smaller, stiffer thrombi require an MRAS system capable of generating forces at higher temporal frequencies. Solenoids with fewer turns, and thus less inductance, could improve high frequency performance, but the reduced force may compromise results. In this work, a computational model capable of assessing the effectiveness of MRAS elastometry magnet configurations is presented and validated.</p><p><strong>Approach: </strong>Finite element analysis (FEA) was used to model the force and inductance of MRAS systems. The simulations incorporated both solenoid electromagnets and permanent magnets in three-dimensional steady-state, frequency domain, and time domain studies.</p><p><strong>Main results: </strong>The model successfully predicted that a configuration in which permanent magnets were added to an existing MRAS system could be used to increase the force supplied. Accordingly, the displacement measured in a magnetically labeled validation phantom increased by a factor of 2.2 ± 0.3 when the force was predicted to increase by a factor of 2.2 ± 0.2. The model additionally identified a new solenoid configuration consisting of four smaller coils capable of providing sufficient force at higher driving frequencies.</p><p><strong>Significance: </strong>These results indicate two methods by which MRAS systems could be designed to deliver higher frequency magnetic forces without the need for experimental trial and error. Either the number of turns within each solenoid could be reduced while permanent magnets are added at precise locations, or a larger number of smaller solenoids could be used. These findings overcome a key challenge toward the goal of MMUS thrombosis elastometry, and simulation files are provided online for broader experimentation.</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11343222/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142057527","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Learning to Discover Regulatory Elements for Gene Expression Prediction.","authors":"Xingyu Su, Haiyang Yu, Degui Zhi, Shuiwang Ji","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>We consider the problem of predicting gene expressions from DNA sequences. A key challenge of this task is to find the regulatory elements that control gene expressions. Here, we introduce Seq2Exp, a Sequence to Expression network explicitly designed to discover and extract regulatory elements that drive target gene expression, enhancing the accuracy of the gene expression prediction. Our approach captures the causal relationship between epigenomic signals, DNA sequences and their associated regulatory elements. Specifically, we propose to decompose the epigenomic signals and the DNA sequence conditioned on the causal active regulatory elements, and apply an information bottleneck with the Beta distribution to combine their effects while filtering out non-causal components. Our experiments demonstrate that Seq2Exp outperforms existing baselines in gene expression prediction tasks and discovers influential regions compared to commonly used statistical methods for peak detection such as MACS3. The source code is released as part of the AIRS library (https://github.com/divelab/AIRS/).</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11875287/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143545080","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Regularization by Neural Style Transfer for MRI Field-Transfer Reconstruction with Limited Data.","authors":"Guoyao Shen, Yancheng Zhu, Mengyu Li, Ryan McNaughton, Hernan Jara, Sean B Andersson, Chad W Farris, Stephan Anderson, Xin Zhang","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>Recent advances in MRI reconstruction have demonstrated remarkable success through deep learning-based models. However, most existing methods rely heavily on large-scale, task-specific datasets, making reconstruction in data-limited settings a critical yet underexplored challenge. While regularization by denoising (RED) leverages denoisers as priors for reconstruction, we propose Regularization by Neural Style Transfer (RNST), a novel framework that integrates a neural style transfer (NST) engine with a denoiser to enable magnetic field-transfer reconstruction. RNST generates high-field-quality images from low-field inputs without requiring paired training data, leveraging style priors to address limited-data settings. Our experiment results demonstrate RNST's ability to reconstruct high-quality images across diverse anatomical planes (axial, coronal, sagittal) and noise levels, achieving superior clarity, contrast, and structural fidelity compared to lower-field references. Crucially, RNST maintains robustness even when style and content images lack exact alignment, broadening its applicability in clinical environments where precise reference matches are unavailable. By combining the strengths of NST and denoising, RNST offers a scalable, data-efficient solution for MRI field-transfer reconstruction, demonstrating significant potential for resource-limited settings.</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11875279/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143545086","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Geometry of the cumulant series in neuroimaging.","authors":"Santiago Coelho, Filip Szczepankiewicz, Els Fieremans, Dmitry S Novikov","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>Water diffusion gives rise to micrometer-scale sensitivity of diffusion MRI (dMRI) to cellular-level tissue structure. The advent of precision medicine and quantitative imaging hinges on revealing the information content of dMRI, and providing its parsimonious basis- and hardware-independent \"fingerprint\". Here we reveal the geometry of a multi-dimensional dMRI signal, classify all 21 invariants of diffusion and covariance tensors in terms of irreducible representations of the group of rotations, and relate them to tissue properties. Previously studied dMRI contrasts are expressed via 7 invariants, while the remaining 14 provide novel complementary information. We design acquisitions based on icosahedral vertices guaranteeing minimal number of measurements to determine 3-4 most used invariants in only 1-2 minutes for the whole brain. Representing dMRI signals via scalar invariant maps with definite symmetries will underpin machine learning classifiers of brain pathology, development, and aging, while fast protocols will enable translation of advanced dMRI into clinical practice.</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11398539/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142303253","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Instability of a fluctuating biomimetic membrane driven by an applied uniform DC electric field.","authors":"Zongxin Yu, Shuozhen Zhao, Michael J Miksis, Petia M Vlahovska","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>The linear stability of a lipid membrane under a DC electric field, applied perpendicularly to the interface, is investigated in the electrokinetic framework, taking account to the dynamics of the Debye layers formed near the membrane. The perturbed charge in the Debye layer redistributes and destabilizes the membrane via electrical surface stress interior and exterior to the membrane. The instability is suppressed as the difference in the electrolyte concentration of the solutions separated by the membrane increases, due to a weakened base state electric field near the membrane. This result contrasts with the destabilizing effect predicted using the leaky dielectric model in cases of asymmetric conductivity. We attribute this difference to the varying assumptions about the perturbation amplitude relative to the Debye length, which result in different regimes of validity for the linear stability analysis within these two frameworks.</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11875285/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143545079","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Diffuse-charge dynamics across a capacitive interface in a DC electric field.","authors":"Shuozhen Zhao, Bhavya Balu, Zongxin Yu, Michael J Miksis, Petia M Vlahovska","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>Cells and cellular organelles are encapsulated by nanometrically thin membranes whose main component is a lipid bilayer. In the presence of electric fields, the ion-impermeable lipid bilayer acts as a capacitor and supports a potential difference across the membrane. We analyze the charging dynamics of a planar membrane separating bulk solutions with different electrolyte concentrations upon the application of an applied uniform DC electric field. The membrane is modeled as a zero-thickness capacitive interface. The evolution of the electric potential and ions distributions in the bulk are solved for using the Poisson-Nernst-Planck (PNP) equations. Asymptotic solutions are derived in the limit of thin Debye layers and weak fields (compared to the thermal electric potential).</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11875280/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143544788","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Time-series attribution maps with regularized contrastive learning.","authors":"Steffen Schneider, Rodrigo González Laiz, Anastasiia Filippova, Markus Frey, Mackenzie Weygandt Mathis","doi":"","DOIUrl":"","url":null,"abstract":"<p><p>Gradient-based attribution methods aim to explain decisions of deep learning models but so far lack identifiability guarantees. Here, we propose a method to generate attribution maps with identifiability guarantees by developing a regularized contrastive learning algorithm trained on time-series data plus a new attribution method called Inverted Neuron Gradient (collectively named xCEBRA). We show theoretically that xCEBRA has favorable properties for identifying the Jacobian matrix of the data generating process. Empirically, we demonstrate robust approximation of zero vs. non-zero entries in the ground-truth attribution map on synthetic datasets, and significant improvements across previous attribution methods based on feature ablation, Shapley values, and other gradient-based methods. Our work constitutes a first example of identifiable inference of time-series attribution maps and opens avenues to a better understanding of time-series data, such as for neural dynamics and decision-processes within neural networks.</p>","PeriodicalId":93888,"journal":{"name":"ArXiv","volume":" ","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11875283/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143545154","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}