Pub Date : 2024-01-05DOI: 10.3389/fcomp.2023.1286860
Sridhar Tayur, Ananth Tenneti
In this mini-review, we introduce and summarize research from the Quantum Technologies Group (QTG) at Carnegie Mellon University related to computational experience with quantum annealing, performed in collaboration with several other institutions including IIT-Madras and NASA (QuAIL). We present a novel hybrid quantum-classical heuristic algorithm (GAMA, Graver Augmented Multi-seed Algorithm) for non-linear, integer optimization, and illustrate it on an application (in cancer genomics). We then present an algebraic geometry-based algorithm for embedding a problem onto a hardware that is not fully connected, along with a companion Integer Programming (IP) approach. Next, we discuss the performance of two photonic devices - the Temporal Multiplexed Ising Machine (TMIM) and the Spatial Photonic Ising Machine (SPIM) - on Max-Cut and Number Partitioning instances. We close with an outline of the current work.
{"title":"Quantum annealing research at CMU: algorithms, hardware, applications","authors":"Sridhar Tayur, Ananth Tenneti","doi":"10.3389/fcomp.2023.1286860","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1286860","url":null,"abstract":"In this mini-review, we introduce and summarize research from the Quantum Technologies Group (QTG) at Carnegie Mellon University related to computational experience with quantum annealing, performed in collaboration with several other institutions including IIT-Madras and NASA (QuAIL). We present a novel hybrid quantum-classical heuristic algorithm (GAMA, Graver Augmented Multi-seed Algorithm) for non-linear, integer optimization, and illustrate it on an application (in cancer genomics). We then present an algebraic geometry-based algorithm for embedding a problem onto a hardware that is not fully connected, along with a companion Integer Programming (IP) approach. Next, we discuss the performance of two photonic devices - the Temporal Multiplexed Ising Machine (TMIM) and the Spatial Photonic Ising Machine (SPIM) - on Max-Cut and Number Partitioning instances. We close with an outline of the current work.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"2 10","pages":""},"PeriodicalIF":2.6,"publicationDate":"2024-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139381005","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-05DOI: 10.3389/fcomp.2023.1286657
Sai Sakunthala Guddanti, Apurva Padhye, Anil Prabhakar, Sridhar Tayur
Early diagnosis of pneumonia is crucial to increase the chances of survival and reduce the recovery time of the patient. Chest X-ray images, the most widely used method in practice, are challenging to classify. Our aim is to develop a machine learning tool that can accurately classify images as belonging to normal or infected individuals. A support vector machine (SVM) is attractive because binary classification can be represented as an optimization problem, in particular as a Quadratic Unconstrained Binary Optimization (QUBO) model, which, in turn, maps naturally to an Ising model, thereby making annealing—classical, quantum, and hybrid—an attractive approach to explore. In this study, we offer a comparison between different methods: (1) a classical state-of-the-art implementation of SVM (LibSVM); (2) solving SVM with a classical solver (Gurobi), with and without decomposition; (3) solving SVM with simulated annealing; (4) solving SVM with quantum annealing (D-Wave); and (5) solving SVM using Graver Augmented Multi-seed Algorithm (GAMA). GAMA is tried with several different numbers of Graver elements and a number of seeds using both simulating annealing and quantum annealing. We found that simulated annealing and GAMA (with simulated annealing) are comparable, provide accurate results quickly, competitive with LibSVM, and superior to Gurobi and quantum annealing.
{"title":"Pneumonia detection by binary classification: classical, quantum, and hybrid approaches for support vector machine (SVM)","authors":"Sai Sakunthala Guddanti, Apurva Padhye, Anil Prabhakar, Sridhar Tayur","doi":"10.3389/fcomp.2023.1286657","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1286657","url":null,"abstract":"Early diagnosis of pneumonia is crucial to increase the chances of survival and reduce the recovery time of the patient. Chest X-ray images, the most widely used method in practice, are challenging to classify. Our aim is to develop a machine learning tool that can accurately classify images as belonging to normal or infected individuals. A support vector machine (SVM) is attractive because binary classification can be represented as an optimization problem, in particular as a Quadratic Unconstrained Binary Optimization (QUBO) model, which, in turn, maps naturally to an Ising model, thereby making annealing—classical, quantum, and hybrid—an attractive approach to explore. In this study, we offer a comparison between different methods: (1) a classical state-of-the-art implementation of SVM (LibSVM); (2) solving SVM with a classical solver (Gurobi), with and without decomposition; (3) solving SVM with simulated annealing; (4) solving SVM with quantum annealing (D-Wave); and (5) solving SVM using Graver Augmented Multi-seed Algorithm (GAMA). GAMA is tried with several different numbers of Graver elements and a number of seeds using both simulating annealing and quantum annealing. We found that simulated annealing and GAMA (with simulated annealing) are comparable, provide accurate results quickly, competitive with LibSVM, and superior to Gurobi and quantum annealing.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"13 10","pages":""},"PeriodicalIF":2.6,"publicationDate":"2024-01-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139382846","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-04DOI: 10.3389/fcomp.2023.1237531
Cleo Valentine
Technological advancements in physiological body sensor networks (i.e., biometric tracking wearables) and simulated environments (i.e., VR) have led to increased research in the field of neuroarchitecture, specifically investigating the effects of architectural forms, defined here as subtle variations in the shape or configuration of the interior built environment, on neurological responses. While this research field is still in its nascent stages, early findings suggest that certain architectural forms may impact physiological stress responses. Physiological stress has, in turn, been implicated in the development of certain diseases, including cardiovascular disease, cancer, chronic kidney disease, non-alcoholic fatty liver disease and autoimmune and neurodegenerative disorders. To aid future research, particularly into the relationship between media architecture and physiological stress, this paper conducts a systematic review following PRISMA-P guidelines on studies that evaluated physiological stress responses to architectural form using clinical biomarkers. The review identifies the specific clinical biomarkers used to evaluate physiological stress responses to architectural forms and the distinct categories of architectural forms that have, to date, been correlated with elevated stress responses: curvature, enclosure and proportion. Although these studies' findings imply that the identified architectural forms influence physiological stress, their generalisability is arguably constrained by several factors. These constraints include the paucity of research in this area, the lack of uniformity in the definition and measurement of these architectural forms, the varying contextual settings, the unisensory approach of research methodologies, and the duration of exposure under evaluation. The review concludes that clinical biomarkers may be used to measure the impact of architectural form on physiological stress; however, future research should strive for standardized approaches in defining and measuring architectural forms in order to increase the transferability and robustness of results.
{"title":"The impact of architectural form on physiological stress: a systematic review","authors":"Cleo Valentine","doi":"10.3389/fcomp.2023.1237531","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1237531","url":null,"abstract":"Technological advancements in physiological body sensor networks (i.e., biometric tracking wearables) and simulated environments (i.e., VR) have led to increased research in the field of neuroarchitecture, specifically investigating the effects of architectural forms, defined here as subtle variations in the shape or configuration of the interior built environment, on neurological responses. While this research field is still in its nascent stages, early findings suggest that certain architectural forms may impact physiological stress responses. Physiological stress has, in turn, been implicated in the development of certain diseases, including cardiovascular disease, cancer, chronic kidney disease, non-alcoholic fatty liver disease and autoimmune and neurodegenerative disorders. To aid future research, particularly into the relationship between media architecture and physiological stress, this paper conducts a systematic review following PRISMA-P guidelines on studies that evaluated physiological stress responses to architectural form using clinical biomarkers. The review identifies the specific clinical biomarkers used to evaluate physiological stress responses to architectural forms and the distinct categories of architectural forms that have, to date, been correlated with elevated stress responses: curvature, enclosure and proportion. Although these studies' findings imply that the identified architectural forms influence physiological stress, their generalisability is arguably constrained by several factors. These constraints include the paucity of research in this area, the lack of uniformity in the definition and measurement of these architectural forms, the varying contextual settings, the unisensory approach of research methodologies, and the duration of exposure under evaluation. The review concludes that clinical biomarkers may be used to measure the impact of architectural form on physiological stress; however, future research should strive for standardized approaches in defining and measuring architectural forms in order to increase the transferability and robustness of results.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"10 6","pages":""},"PeriodicalIF":2.6,"publicationDate":"2024-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139386357","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-01-04DOI: 10.3389/fcomp.2023.1233904
Eléni Economidou, Alina Itzlinger, Christopher Frauenberger
The emerging field of human-building interaction (HBI) has its roots in the historical trends of the development of architecture and human-computer interaction (HCI). Advancements in building information modelling (BIM), sensing, and actuation technologies as well as the commodification and miniaturisation of microprocessors over the past two decades are transforming what once were quixotic visions of a cybernetic architecture into reality. This new reality which integrates computation with architecture opens up different kinds of engagements in the ways we design, use, and inhabit our built environments. A question that follows this new reality is: how can we conceptualise human experience in such environments? Thus far, the lived human experience of such interactions has been an overlooked aspect in HBI-related research. In this article, we provide an initial experience framework for HBI underpinned by existing literature from the HCI and architecture domains on the subjective, lived-in experience of architecture and findings derived from a case study of a field-deployed HBI interface. The research objective of our framework is to outline aspects of HBI lived experiences that can be used as guiding lenses for HBI designers and practitioners who wish to design for and assess such experiences.
{"title":"Lived experience in human-building interaction (HBI): an initial framework","authors":"Eléni Economidou, Alina Itzlinger, Christopher Frauenberger","doi":"10.3389/fcomp.2023.1233904","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1233904","url":null,"abstract":"The emerging field of human-building interaction (HBI) has its roots in the historical trends of the development of architecture and human-computer interaction (HCI). Advancements in building information modelling (BIM), sensing, and actuation technologies as well as the commodification and miniaturisation of microprocessors over the past two decades are transforming what once were quixotic visions of a cybernetic architecture into reality. This new reality which integrates computation with architecture opens up different kinds of engagements in the ways we design, use, and inhabit our built environments. A question that follows this new reality is: how can we conceptualise human experience in such environments? Thus far, the lived human experience of such interactions has been an overlooked aspect in HBI-related research. In this article, we provide an initial experience framework for HBI underpinned by existing literature from the HCI and architecture domains on the subjective, lived-in experience of architecture and findings derived from a case study of a field-deployed HBI interface. The research objective of our framework is to outline aspects of HBI lived experiences that can be used as guiding lenses for HBI designers and practitioners who wish to design for and assess such experiences.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"85 4","pages":""},"PeriodicalIF":2.6,"publicationDate":"2024-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"139386097","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-22DOI: 10.3389/fcomp.2023.1230284
Velvet Spors, Martin Flintham, Pat Brundell, David Murphy
The days of dusty couches in therapists' offices behind closed doors are long gone. Now, personalized mood tracking, therapy appointments and breathing exercises are just mere clicks (or taps) away: Technologies for self-care (SCTs) that focus on mental health are both a flourishing industry and an academic field of interest. As societal, and cultural artifacts, SCTs for mental health are imbued with values, worldviews, and assumptions about these concepts by their designers and developers. Here, current SCTs tend to lean toward a more medical(ised) approach due to being shaped by dominant views of mental health as an individualized issue. However, this approach is only one of many potential pedagogies and approaches. As an alternative, we explore what SCTs for mental health could be like, from a humanistic, person-centered standpoint: We conceptualize mental health in holistic terms, as an experiential quality of everyday life.To this end, we report on two engagements with humanistic practitioners and the person-centered approach as a guiding principle: First, we ran a workshop informed by the Rogerian “encounter group”. This approach is focused on providing the space to meaningfully meet and relate to people. Inspired by this concept, we brought together humanistic practitioners to openly explore what technology for (self-)care means for them. Second, we build on the insights from the aforementioned study by organizing an asynchronous, online whiteboard for humanistic practitioners—counselors, students-in-training, therapists, and researchers—to explore their utopian, realistic and dystopian visions of SCTs.Through thematic analysis and affinity-clustering these engagements, we construct an understanding that technology within a person-centered, humanistic context is a constrained, ambiguous undertaking, yet also one full of potential.We conclude the paper by sketching out three design opportunities for how the person-centered approach, and humanistic psychology in general could be integrated into caring technologies.
{"title":"Care-full data, care-less systems: making sense of self-care technologies for mental health with humanistic practitioners in the United Kingdom","authors":"Velvet Spors, Martin Flintham, Pat Brundell, David Murphy","doi":"10.3389/fcomp.2023.1230284","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1230284","url":null,"abstract":"The days of dusty couches in therapists' offices behind closed doors are long gone. Now, personalized mood tracking, therapy appointments and breathing exercises are just mere clicks (or taps) away: Technologies for self-care (SCTs) that focus on mental health are both a flourishing industry and an academic field of interest. As societal, and cultural artifacts, SCTs for mental health are imbued with values, worldviews, and assumptions about these concepts by their designers and developers. Here, current SCTs tend to lean toward a more medical(ised) approach due to being shaped by dominant views of mental health as an individualized issue. However, this approach is only one of many potential pedagogies and approaches. As an alternative, we explore what SCTs for mental health could be like, from a humanistic, person-centered standpoint: We conceptualize mental health in holistic terms, as an experiential quality of everyday life.To this end, we report on two engagements with humanistic practitioners and the person-centered approach as a guiding principle: First, we ran a workshop informed by the Rogerian “encounter group”. This approach is focused on providing the space to meaningfully meet and relate to people. Inspired by this concept, we brought together humanistic practitioners to openly explore what technology for (self-)care means for them. Second, we build on the insights from the aforementioned study by organizing an asynchronous, online whiteboard for humanistic practitioners—counselors, students-in-training, therapists, and researchers—to explore their utopian, realistic and dystopian visions of SCTs.Through thematic analysis and affinity-clustering these engagements, we construct an understanding that technology within a person-centered, humanistic context is a constrained, ambiguous undertaking, yet also one full of potential.We conclude the paper by sketching out three design opportunities for how the person-centered approach, and humanistic psychology in general could be integrated into caring technologies.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"32 22","pages":""},"PeriodicalIF":2.6,"publicationDate":"2023-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138946790","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-21DOI: 10.3389/fcomp.2023.1289869
Caitlin Morris, Valdemar Danry, Pattie Maes
Technologies on the body that require explicit awareness to be operated or monitored often risk disrupting human awareness and induce stress and excessive cognitive load. With the increasing interest in body-centric technologies, it is thus essential to understand how to build technologies that interface with human awareness without disrupting or requiring too many cognitive resources. In this paper, we build and evaluate a wearable system that uses different feedback types to alter human awareness (of the device). We further demonstrate how this awareness impacts cognitive load, sense of body-ownership, and sense of agency, which are often essential antecedents to successful and continued use. Moreover, we further investigate physiological signals, such as physiological synchrony, as well as qualitative reports in a multimodal analysis. Our results show that devices that provide feedback that deviate from expected behavior tend to generate higher amounts of explicit awareness, and that such increased awareness correlates with increased cognitive load, lower sense of agency and lower sense of body-ownership. Moreover, we find that interoceptive acuity correlates with diminished sense of agency. We discuss their implications for designing wearable body-centric systems that induce or disrupt different levels of awareness to deliver or diminish a sense of body-ownership and agency over the system.
{"title":"Wearable systems without experiential disruptions: exploring the impact of device feedback changes on explicit awareness, physiological synchrony, sense of agency, and device-body ownership","authors":"Caitlin Morris, Valdemar Danry, Pattie Maes","doi":"10.3389/fcomp.2023.1289869","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1289869","url":null,"abstract":"Technologies on the body that require explicit awareness to be operated or monitored often risk disrupting human awareness and induce stress and excessive cognitive load. With the increasing interest in body-centric technologies, it is thus essential to understand how to build technologies that interface with human awareness without disrupting or requiring too many cognitive resources. In this paper, we build and evaluate a wearable system that uses different feedback types to alter human awareness (of the device). We further demonstrate how this awareness impacts cognitive load, sense of body-ownership, and sense of agency, which are often essential antecedents to successful and continued use. Moreover, we further investigate physiological signals, such as physiological synchrony, as well as qualitative reports in a multimodal analysis. Our results show that devices that provide feedback that deviate from expected behavior tend to generate higher amounts of explicit awareness, and that such increased awareness correlates with increased cognitive load, lower sense of agency and lower sense of body-ownership. Moreover, we find that interoceptive acuity correlates with diminished sense of agency. We discuss their implications for designing wearable body-centric systems that induce or disrupt different levels of awareness to deliver or diminish a sense of body-ownership and agency over the system.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"134 41","pages":""},"PeriodicalIF":2.6,"publicationDate":"2023-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138953434","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-21DOI: 10.3389/fcomp.2023.1305413
V. Bhuvana Kumar, M. Kathiravan
The complex synthesis of emotions seen in music is meticulously composed using a wide range of aural components. Given the expanding soundscape and abundance of online music resources, creating a music recommendation system is significant. The area of music file emotion recognition is particularly fascinating. The RGRU (Enhanced Residual Gated Recurrent Unit), a complex architecture, is used in our study to look at MIDI (Musical Instrument and Digital Interface) compositions for detecting emotions. This involves extracting diverse features from the MIDI dataset, encompassing harmony, rhythm, dynamics, and statistical attributes. These extracted features subsequently serve as input to our emotion recognition model for emotion detection. We use an improved RGRU version to identify emotions and the Adaptive Red Fox Algorithm (ARFA) to optimize the RGRU hyperparameters. Our suggested model offers a sophisticated classification framework that effectively divides emotional content into four separate quadrants: positive-high, positive-low, negative-high, and negative-low. The Python programming environment is used to implement our suggested approach. We use the EMOPIA dataset to compare its performance to the traditional approach and assess its effectiveness experimentally. The trial results show better performance compared to traditional methods, with higher accuracy, recall, F-measure, and precision.
{"title":"Emotion recognition from MIDI musical file using Enhanced Residual Gated Recurrent Unit architecture","authors":"V. Bhuvana Kumar, M. Kathiravan","doi":"10.3389/fcomp.2023.1305413","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1305413","url":null,"abstract":"The complex synthesis of emotions seen in music is meticulously composed using a wide range of aural components. Given the expanding soundscape and abundance of online music resources, creating a music recommendation system is significant. The area of music file emotion recognition is particularly fascinating. The RGRU (Enhanced Residual Gated Recurrent Unit), a complex architecture, is used in our study to look at MIDI (Musical Instrument and Digital Interface) compositions for detecting emotions. This involves extracting diverse features from the MIDI dataset, encompassing harmony, rhythm, dynamics, and statistical attributes. These extracted features subsequently serve as input to our emotion recognition model for emotion detection. We use an improved RGRU version to identify emotions and the Adaptive Red Fox Algorithm (ARFA) to optimize the RGRU hyperparameters. Our suggested model offers a sophisticated classification framework that effectively divides emotional content into four separate quadrants: positive-high, positive-low, negative-high, and negative-low. The Python programming environment is used to implement our suggested approach. We use the EMOPIA dataset to compare its performance to the traditional approach and assess its effectiveness experimentally. The trial results show better performance compared to traditional methods, with higher accuracy, recall, F-measure, and precision.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"51 26","pages":""},"PeriodicalIF":2.6,"publicationDate":"2023-12-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138949538","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-19DOI: 10.3389/fcomp.2023.1275026
W. Pickard, Kelsey Sikes, Huma Jamil, Nicholas Chaffee, Nathaniel Blanchard, Michael Kirby, Christopher Peterson
Artificial neural networks (ANNs) are sensitive to perturbations and adversarial attacks. One hypothesized solution to adversarial robustness is to align manifolds in the embedded space of neural networks with biologically grounded manifolds. Recent state-of-the-art works that emphasize learning robust neural representations, rather than optimizing for a specific target task like classification, support the idea that researchers should investigate this hypothesis. While works have shown that fine-tuning ANNs to coincide with biological vision does increase robustness to both perturbations and adversarial attacks, these works have relied on proprietary datasets—the lack of publicly available biological benchmarks makes it difficult to evaluate the efficacy of these claims. Here, we deliver a curated dataset consisting of biological representations of images taken from two commonly used computer vision datasets, ImageNet and COCO, that can be easily integrated into model training and evaluation. Specifically, we take a large functional magnetic resonance imaging (fMRI) dataset (BOLD5000), preprocess it into representational dissimilarity matrices (RDMs), and establish an infrastructure that anyone can use to train models with biologically grounded representations. Using this infrastructure, we investigate the representations of several popular neural networks and find that as networks have been optimized for tasks, their correspondence with biological fidelity has decreased. Additionally, we use a previously unexplored graph-based technique, Fiedler partitioning, to showcase the viability of the biological data, and the potential to extend these analyses by extending RDMs into Laplacian matrices. Overall, our findings demonstrate the potential of utilizing our new biological benchmark to effectively enhance the robustness of models.
{"title":"Exploring fMRI RDMs: enhancing model robustness through neurobiological data","authors":"W. Pickard, Kelsey Sikes, Huma Jamil, Nicholas Chaffee, Nathaniel Blanchard, Michael Kirby, Christopher Peterson","doi":"10.3389/fcomp.2023.1275026","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1275026","url":null,"abstract":"Artificial neural networks (ANNs) are sensitive to perturbations and adversarial attacks. One hypothesized solution to adversarial robustness is to align manifolds in the embedded space of neural networks with biologically grounded manifolds. Recent state-of-the-art works that emphasize learning robust neural representations, rather than optimizing for a specific target task like classification, support the idea that researchers should investigate this hypothesis. While works have shown that fine-tuning ANNs to coincide with biological vision does increase robustness to both perturbations and adversarial attacks, these works have relied on proprietary datasets—the lack of publicly available biological benchmarks makes it difficult to evaluate the efficacy of these claims. Here, we deliver a curated dataset consisting of biological representations of images taken from two commonly used computer vision datasets, ImageNet and COCO, that can be easily integrated into model training and evaluation. Specifically, we take a large functional magnetic resonance imaging (fMRI) dataset (BOLD5000), preprocess it into representational dissimilarity matrices (RDMs), and establish an infrastructure that anyone can use to train models with biologically grounded representations. Using this infrastructure, we investigate the representations of several popular neural networks and find that as networks have been optimized for tasks, their correspondence with biological fidelity has decreased. Additionally, we use a previously unexplored graph-based technique, Fiedler partitioning, to showcase the viability of the biological data, and the potential to extend these analyses by extending RDMs into Laplacian matrices. Overall, our findings demonstrate the potential of utilizing our new biological benchmark to effectively enhance the robustness of models.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"10 16","pages":""},"PeriodicalIF":2.6,"publicationDate":"2023-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138959963","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-19DOI: 10.3389/fcomp.2023.1271899
Muhammad Fraz, M. Muslam, Mudassar Hussain, Rashid Amin, Jiang Xie
Cognitive Radio Networks (CRNs) have ushered in a transformative era in wireless communication, reshaping the landscape of radio spectrum utilization and management. At the core of CRNs lies the pivotal capability to sense the radio frequency spectrum dynamically and adapt transmission parameters to preemptively address interference and optimize spectrum utilization. This article addresses the escalating challenges associated with Quality of Service (QoS) management in CRNs, exacerbated by their dynamic nature, especially in scenarios characterized by high mobility. Concurrently, the article underscores the critical significance of energy efficiency, given its direct implications on network operational costs and sustainability. To effectively navigate the intricate interplay between QoS and energy management in CRNs, we propose a Smart Sensing Enabled Dynamic Spectrum Management scheme (SSDSM). Within the SSDSM framework, cognitive user energy undergoes intelligent sensing, while QoS is governed through dynamic spectrum management. The proposed scheme optimizes service response time by refining fuzzy-based controllers and curtails energy consumption through periodic sensing triggered by predefined rules. Operationalizing within a centralized paradigm, the entire network is overseen by a central controlling node, tasked with formulating an optimal channel list using the SSDSM scheme and allocating it to cognitive users. The efficacy of the proposed scheme is evaluated and validated through rigorous testing using MATLAB. Results reveal tangible enhancements in system efficiency, encompassing maximized throughput, reduced handoff ratio, and minimized service response delay. This research contributes to the ongoing discourse on advancing the performance metrics of cognitive radio networks in the pursuit of reliable and sustainable wireless communication services.
{"title":"Smart sensing enabled dynamic spectrum management for cognitive radio networks","authors":"Muhammad Fraz, M. Muslam, Mudassar Hussain, Rashid Amin, Jiang Xie","doi":"10.3389/fcomp.2023.1271899","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1271899","url":null,"abstract":"Cognitive Radio Networks (CRNs) have ushered in a transformative era in wireless communication, reshaping the landscape of radio spectrum utilization and management. At the core of CRNs lies the pivotal capability to sense the radio frequency spectrum dynamically and adapt transmission parameters to preemptively address interference and optimize spectrum utilization. This article addresses the escalating challenges associated with Quality of Service (QoS) management in CRNs, exacerbated by their dynamic nature, especially in scenarios characterized by high mobility. Concurrently, the article underscores the critical significance of energy efficiency, given its direct implications on network operational costs and sustainability. To effectively navigate the intricate interplay between QoS and energy management in CRNs, we propose a Smart Sensing Enabled Dynamic Spectrum Management scheme (SSDSM). Within the SSDSM framework, cognitive user energy undergoes intelligent sensing, while QoS is governed through dynamic spectrum management. The proposed scheme optimizes service response time by refining fuzzy-based controllers and curtails energy consumption through periodic sensing triggered by predefined rules. Operationalizing within a centralized paradigm, the entire network is overseen by a central controlling node, tasked with formulating an optimal channel list using the SSDSM scheme and allocating it to cognitive users. The efficacy of the proposed scheme is evaluated and validated through rigorous testing using MATLAB. Results reveal tangible enhancements in system efficiency, encompassing maximized throughput, reduced handoff ratio, and minimized service response delay. This research contributes to the ongoing discourse on advancing the performance metrics of cognitive radio networks in the pursuit of reliable and sustainable wireless communication services.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":" 30","pages":""},"PeriodicalIF":2.6,"publicationDate":"2023-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138961727","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-12-14DOI: 10.3389/fcomp.2023.1293209
Muhammad Saad Sheikh, Rabia Noor Enam, R. Qureshi
Fog Computing has emerged as a pivotal technology for enabling low-latency, context-aware, and efficient computing at the edge of the network. Effective task scheduling plays a vital role in optimizing the performance of fog computing systems. Traditional task scheduling algorithms, primarily designed for centralized cloud environments, often fail to cater to the dynamic, heterogeneous, and resource-constrained nature of Fog nodes. To overcome these limitations, we introduce a sophisticated machine learning-driven methodology that adapts task allocation to the ever-changing Fog environment's conditions. Our approach amalgamates K-Means clustering algorithm enhanced with fuzzy logic, a robust unsupervised learning technique, to efficiently group Fog nodes based on their resource characteristics and workload patterns. The proposed method combines the clustering capabilities of K-means with the adaptability of fuzzy logic to dynamically allocate tasks to fog nodes. By leveraging machine learning techniques, we demonstrate how tasks can be intelligently allocated to fog nodes, resulting in reducing execution time, response time and network usage. Through extensive experiments, we showcase the effectiveness and adaptability of our proposed approach in dynamic fog environments. Clustering proves to be a time-effective method for identifying groups of jobs per virtual machine (VM) efficiently. To model and evaluate our proposed approach, we have utilized iFogSim. The simulation results affirm the effectiveness of our scheduling technique, showcasing significant enhancements in execution time reduction, minimized network utilization, and improved response time when compared to existing machine learning and non-machine learning based scheduling methods within the iFogSim framework.
{"title":"Machine learning-driven task scheduling with dynamic K-means based clustering algorithm using fuzzy logic in FOG environment","authors":"Muhammad Saad Sheikh, Rabia Noor Enam, R. Qureshi","doi":"10.3389/fcomp.2023.1293209","DOIUrl":"https://doi.org/10.3389/fcomp.2023.1293209","url":null,"abstract":"Fog Computing has emerged as a pivotal technology for enabling low-latency, context-aware, and efficient computing at the edge of the network. Effective task scheduling plays a vital role in optimizing the performance of fog computing systems. Traditional task scheduling algorithms, primarily designed for centralized cloud environments, often fail to cater to the dynamic, heterogeneous, and resource-constrained nature of Fog nodes. To overcome these limitations, we introduce a sophisticated machine learning-driven methodology that adapts task allocation to the ever-changing Fog environment's conditions. Our approach amalgamates K-Means clustering algorithm enhanced with fuzzy logic, a robust unsupervised learning technique, to efficiently group Fog nodes based on their resource characteristics and workload patterns. The proposed method combines the clustering capabilities of K-means with the adaptability of fuzzy logic to dynamically allocate tasks to fog nodes. By leveraging machine learning techniques, we demonstrate how tasks can be intelligently allocated to fog nodes, resulting in reducing execution time, response time and network usage. Through extensive experiments, we showcase the effectiveness and adaptability of our proposed approach in dynamic fog environments. Clustering proves to be a time-effective method for identifying groups of jobs per virtual machine (VM) efficiently. To model and evaluate our proposed approach, we have utilized iFogSim. The simulation results affirm the effectiveness of our scheduling technique, showcasing significant enhancements in execution time reduction, minimized network utilization, and improved response time when compared to existing machine learning and non-machine learning based scheduling methods within the iFogSim framework.","PeriodicalId":52823,"journal":{"name":"Frontiers in Computer Science","volume":"95 1","pages":""},"PeriodicalIF":2.6,"publicationDate":"2023-12-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"138975484","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：