Pub Date : 2024-08-20DOI: 10.1088/2632-2153/ad6be6
Suresh Bishnoi, Ravinder Bhattoo, Jayadeva3jayadeva@ee.iitd.ac.in, Sayan Ranu, N M Anoop Krishnan
The time evolution of physical systems is described by differential equations, which depend on abstract quantities like energy and force. Traditionally, these quantities are derived as functionals based on observables such as positions and velocities. Discovering these governing symbolic laws is the key to comprehending the interactions in nature. Here, we present a Hamiltonian graph neural network (Hgnn), a physics-enforced Gnn that learns the dynamics of systems directly from their trajectory. We demonstrate the performance of Hgnn on n−springs, n−pendulums, gravitational systems, and binary Lennard Jones systems; Hgnn learns the dynamics in excellent agreement with the ground truth from small amounts of data. We also evaluate the ability of Hgnn to generalize to larger system sizes, and to a hybrid spring-pendulum system that is a combination of two original systems (spring and pendulum) on which the models are trained independently. Finally, employing symbolic regression on the learned Hgnn, we infer the underlying equations relating to the energy functionals, even for complex systems such as the binary Lennard-Jones liquid. Our framework facilitates the interpretable discovery of interaction laws directly from physical system trajectories. Furthermore, this approach can be extended to other systems with topology-dependent dynamics, such as cells, polydisperse gels, or deformable bodies.
{"title":"Discovering symbolic laws directly from trajectories with hamiltonian graph neural networks","authors":"Suresh Bishnoi, Ravinder Bhattoo, Jayadeva3jayadeva@ee.iitd.ac.in, Sayan Ranu, N M Anoop Krishnan","doi":"10.1088/2632-2153/ad6be6","DOIUrl":"https://doi.org/10.1088/2632-2153/ad6be6","url":null,"abstract":"The time evolution of physical systems is described by differential equations, which depend on abstract quantities like energy and force. Traditionally, these quantities are derived as functionals based on observables such as positions and velocities. Discovering these governing symbolic laws is the key to comprehending the interactions in nature. Here, we present a Hamiltonian graph neural network (<sc>Hgnn</sc>), a physics-enforced <sc>Gnn</sc> that learns the dynamics of systems directly from their trajectory. We demonstrate the performance of <sc>Hgnn</sc> on <inline-formula>\u0000<tex-math><?CDATA $n-$?></tex-math><mml:math overflow=\"scroll\"><mml:mrow><mml:mi>n</mml:mi><mml:mo>−</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"mlstad6be6ieqn1.gif\"></inline-graphic></inline-formula>springs, <inline-formula>\u0000<tex-math><?CDATA $n-$?></tex-math><mml:math overflow=\"scroll\"><mml:mrow><mml:mi>n</mml:mi><mml:mo>−</mml:mo></mml:mrow></mml:math><inline-graphic xlink:href=\"mlstad6be6ieqn2.gif\"></inline-graphic></inline-formula>pendulums, gravitational systems, and binary Lennard Jones systems; <sc>Hgnn</sc> learns the dynamics in excellent agreement with the ground truth from small amounts of data. We also evaluate the ability of <sc>Hgnn</sc> to generalize to larger system sizes, and to a hybrid spring-pendulum system that is a combination of two original systems (spring and pendulum) on which the models are trained independently. Finally, employing symbolic regression on the learned <sc>Hgnn</sc>, we infer the underlying equations relating to the energy functionals, even for complex systems such as the binary Lennard-Jones liquid. Our framework facilitates the interpretable discovery of interaction laws directly from physical system trajectories. Furthermore, this approach can be extended to other systems with topology-dependent dynamics, such as cells, polydisperse gels, or deformable bodies.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"210 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142197705","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-14DOI: 10.1088/2632-2153/ad682f
Weizhen Li, Rui Carvalho
Identifying partial differential equations (PDEs) from data is crucial for understanding the governing mechanisms of natural phenomena, yet it remains a challenging task. We present an extension to the ARGOS framework, ARGOS-RAL, which leverages sparse regression with the recurrent adaptive lasso to identify PDEs from limited prior knowledge automatically. Our method automates calculating partial derivatives, constructing a candidate library, and estimating a sparse model. We rigorously evaluate the performance of ARGOS-RAL in identifying canonical PDEs under various noise levels and sample sizes, demonstrating its robustness in handling noisy and non-uniformly distributed data. We also test the algorithm’s performance on datasets consisting solely of random noise to simulate scenarios with severely compromised data quality. Our results show that ARGOS-RAL effectively and reliably identifies the underlying PDEs from data, outperforming the sequential threshold ridge regression method in most cases. We highlight the potential of combining statistical methods, machine learning, and dynamical systems theory to automatically discover governing equations from collected data, streamlining the scientific modeling process.
{"title":"Automating the discovery of partial differential equations in dynamical systems","authors":"Weizhen Li, Rui Carvalho","doi":"10.1088/2632-2153/ad682f","DOIUrl":"https://doi.org/10.1088/2632-2153/ad682f","url":null,"abstract":"Identifying partial differential equations (PDEs) from data is crucial for understanding the governing mechanisms of natural phenomena, yet it remains a challenging task. We present an extension to the ARGOS framework, ARGOS-RAL, which leverages sparse regression with the recurrent adaptive lasso to identify PDEs from limited prior knowledge automatically. Our method automates calculating partial derivatives, constructing a candidate library, and estimating a sparse model. We rigorously evaluate the performance of ARGOS-RAL in identifying canonical PDEs under various noise levels and sample sizes, demonstrating its robustness in handling noisy and non-uniformly distributed data. We also test the algorithm’s performance on datasets consisting solely of random noise to simulate scenarios with severely compromised data quality. Our results show that ARGOS-RAL effectively and reliably identifies the underlying PDEs from data, outperforming the sequential threshold ridge regression method in most cases. We highlight the potential of combining statistical methods, machine learning, and dynamical systems theory to automatically discover governing equations from collected data, streamlining the scientific modeling process.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"14 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142197702","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-14DOI: 10.1088/2632-2153/ad6a00
Jieun Yoo, Jennet Dickinson, Morris Swartz, Giuseppe Di Guglielmo, Alice Bean, Douglas Berry, Manuel Blanco Valentin, Karri DiPetrillo, Farah Fahim, Lindsey Gray, James Hirschauer, Shruti R Kulkarni, Ron Lipton, Petar Maksimovic, Corrinne Mills, Mark S Neubauer, Benjamin Parpillon, Gauri Pradhan, Chinar Syal, Nhan Tran, Dahai Wen, Aaron Young
Highly granular pixel detectors allow for increasingly precise measurements of charged particle tracks. Next-generation detectors require that pixel sizes will be further reduced, leading to unprecedented data rates exceeding those foreseen at the High- Luminosity Large Hadron Collider. Signal processing that handles data incoming at a rate of O(40 MHz) and intelligently reduces the data within the pixelated region of the detector at rate will enhance physics performance at high luminosity and enable physics analyses that are not currently possible. Using the shape of charge clusters deposited in an array of small pixels, the physical properties of the traversing particle can be extracted with locally customized neural networks. In this first demonstration, we present a neural network that can be embedded into the on-sensor readout and filter out hits from low momentum tracks, reducing the detector’s data volume by 57.1%–75.7%. The network is designed and simulated as a custom readout integrated circuit with 28 nm CMOS technology and is expected to operate at less than 300 μW with an area of less than 0.2 mm2. The temporal development of charge clusters is investigated to demonstrate possible future performance gains, and there is also a discussion of future algorithmic and technological improvements that could enhance efficiency, data reduction, and power per area.
{"title":"Smart pixel sensors: towards on-sensor filtering of pixel clusters with deep learning","authors":"Jieun Yoo, Jennet Dickinson, Morris Swartz, Giuseppe Di Guglielmo, Alice Bean, Douglas Berry, Manuel Blanco Valentin, Karri DiPetrillo, Farah Fahim, Lindsey Gray, James Hirschauer, Shruti R Kulkarni, Ron Lipton, Petar Maksimovic, Corrinne Mills, Mark S Neubauer, Benjamin Parpillon, Gauri Pradhan, Chinar Syal, Nhan Tran, Dahai Wen, Aaron Young","doi":"10.1088/2632-2153/ad6a00","DOIUrl":"https://doi.org/10.1088/2632-2153/ad6a00","url":null,"abstract":"Highly granular pixel detectors allow for increasingly precise measurements of charged particle tracks. Next-generation detectors require that pixel sizes will be further reduced, leading to unprecedented data rates exceeding those foreseen at the High- Luminosity Large Hadron Collider. Signal processing that handles data incoming at a rate of <inline-formula>\u0000<tex-math><?CDATA $mathcal{O}$?></tex-math><mml:math overflow=\"scroll\"><mml:mrow><mml:mrow><mml:mi>O</mml:mi></mml:mrow></mml:mrow></mml:math><inline-graphic xlink:href=\"mlstad6a00ieqn1.gif\"></inline-graphic></inline-formula>(40 MHz) and intelligently reduces the data within the pixelated region of the detector <italic toggle=\"yes\">at rate</italic> will enhance physics performance at high luminosity and enable physics analyses that are not currently possible. Using the shape of charge clusters deposited in an array of small pixels, the physical properties of the traversing particle can be extracted with locally customized neural networks. In this first demonstration, we present a neural network that can be embedded into the on-sensor readout and filter out hits from low momentum tracks, reducing the detector’s data volume by 57.1%–75.7%. The network is designed and simulated as a custom readout integrated circuit with 28 nm CMOS technology and is expected to operate at less than 300 <inline-formula>\u0000<tex-math><?CDATA $mu W$?></tex-math><mml:math overflow=\"scroll\"><mml:mrow><mml:mi>μ</mml:mi><mml:mi>W</mml:mi></mml:mrow></mml:math><inline-graphic xlink:href=\"mlstad6a00ieqn2.gif\"></inline-graphic></inline-formula> with an area of less than 0.2 mm<sup>2</sup>. The temporal development of charge clusters is investigated to demonstrate possible future performance gains, and there is also a discussion of future algorithmic and technological improvements that could enhance efficiency, data reduction, and power per area.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"1 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142197703","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-13DOI: 10.1088/2632-2153/ad69ff
Jonas Elsborg, Arghya Bhowmik
Finding low-energy atomic ordering in compositionally complex materials is one of the hardest problems in materials discovery, the solution of which can lead to breakthroughs in functional materials—from alloys to ceramics. In this work, we present the Artificial Structure Arranging Net (ArtiSAN)—a reinforcement learning agent utilizing graph representation that is trained to find low-energy atomic configurations of multicomponent systems through a series of atomic switch operations. ArtiSAN is trained on small alloy supercells ranging from binary to septenary. Strikingly, ArtiSAN generalizes to much larger systems of more than a thousand atoms, which are inaccessible with state-of-the-art methods due to the combinatorially larger search space. The performance of the current ArtiSAN agent is tested and deployed on several compositions that can be correlated with known experimental and high-fidelity computational structures. ArtiSAN demonstrates transfer across size and composition and finds physically meaningful structures using no energy evaluation calls once fully trained. While ArtiSAN will require further modifications to capture all variability in structure search, it is a remarkable step towards solving the structural part of the problem of disordered materials discovery.
{"title":"ArtiSAN: navigating the complexity of material structures with deep reinforcement learning","authors":"Jonas Elsborg, Arghya Bhowmik","doi":"10.1088/2632-2153/ad69ff","DOIUrl":"https://doi.org/10.1088/2632-2153/ad69ff","url":null,"abstract":"Finding low-energy atomic ordering in compositionally complex materials is one of the hardest problems in materials discovery, the solution of which can lead to breakthroughs in functional materials—from alloys to ceramics. In this work, we present the <bold>Arti</bold>ficial <bold>S</bold>tructure <bold>A</bold>rranging <bold>N</bold>et (<bold>ArtiSAN</bold>)—a reinforcement learning agent utilizing graph representation that is trained to find low-energy atomic configurations of multicomponent systems through a series of atomic switch operations. ArtiSAN is trained on small alloy supercells ranging from binary to septenary. Strikingly, ArtiSAN generalizes to much larger systems of more than a thousand atoms, which are inaccessible with state-of-the-art methods due to the combinatorially larger search space. The performance of the current ArtiSAN agent is tested and deployed on several compositions that can be correlated with known experimental and high-fidelity computational structures. ArtiSAN demonstrates transfer across size and composition and finds physically meaningful structures using no energy evaluation calls once fully trained. While ArtiSAN will require further modifications to capture all variability in structure search, it is a remarkable step towards solving the structural part of the problem of disordered materials discovery.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"11 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142197706","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-13DOI: 10.1088/2632-2153/ad6ad2
Sangwon Kim, Byoung Chul Ko
In fields requiring high accountability, it is necessary to understand how deep-learning models make decisions when analyzing the causes of image classification. Concept-based interpretation methods have recently been introduced to reveal the internal mechanisms of deep learning models using high-level concepts. However, such methods are constrained by a trade-off between accuracy and interpretability. For instance, in real-world environments, unlike in well-curated training data, the accurate prediction of expected concepts becomes a challenge owing to the various distortions and complexities introduced by different objects. To overcome this tradeoff, we propose concept graph embedding models (CGEM), reflecting the complex dependencies and structures among concepts through the learning of mutual directionalities. The concept graph convolutional neural network (Concept GCN), a downstream task of CGEM, differs from previous methods that solely determine the presence of concepts because it performs a final classification based on the relationships between con- cepts learned through graph embedding. This process endows the model with high resilience even in the presence of incorrect concepts. In addition, we utilize a deformable bipartite GCN for object- centric concept encoding in the earlier stages, which enhances the homogeneity of the concepts. The experimental results show that, based on deformable concept encoding, the CGEM mitigates the trade-off between task accuracy and interpretability. Moreover, it was confirmed that this approach allows the model to increase the resilience and interpretability while maintaining robustness against various real-world concept distortions and incorrect concept interventions. Our code is available at https://github.com/jumpsnack/cgem.
{"title":"Concept graph embedding models for enhanced accuracy and interpretability","authors":"Sangwon Kim, Byoung Chul Ko","doi":"10.1088/2632-2153/ad6ad2","DOIUrl":"https://doi.org/10.1088/2632-2153/ad6ad2","url":null,"abstract":"In fields requiring high accountability, it is necessary to understand how deep-learning models make decisions when analyzing the causes of image classification. Concept-based interpretation methods have recently been introduced to reveal the internal mechanisms of deep learning models using high-level concepts. However, such methods are constrained by a trade-off between accuracy and interpretability. For instance, in real-world environments, unlike in well-curated training data, the accurate prediction of expected concepts becomes a challenge owing to the various distortions and complexities introduced by different objects. To overcome this tradeoff, we propose concept graph embedding models (CGEM), reflecting the complex dependencies and structures among concepts through the learning of mutual directionalities. The concept graph convolutional neural network (Concept GCN), a downstream task of CGEM, differs from previous methods that solely determine the presence of concepts because it performs a final classification based on the relationships between con- cepts learned through graph embedding. This process endows the model with high resilience even in the presence of incorrect concepts. In addition, we utilize a deformable bipartite GCN for object- centric concept encoding in the earlier stages, which enhances the homogeneity of the concepts. The experimental results show that, based on deformable concept encoding, the CGEM mitigates the trade-off between task accuracy and interpretability. Moreover, it was confirmed that this approach allows the model to increase the resilience and interpretability while maintaining robustness against various real-world concept distortions and incorrect concept interventions. Our code is available at <ext-link ext-link-type=\"uri\" xlink:href=\"https://github.com/jumpsnack/cgem\">https://github.com/jumpsnack/cgem</ext-link>.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"69 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142197707","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-13DOI: 10.1088/2632-2153/ad5fdb
Philipp Schmidt, Sören Arlt, Carlos Ruiz-Gonzalez, Xuemei Gu, Carla Rodríguez, Mario Krenn
Generative Artificial Intelligence (AI) models can propose solutions to scientific problems beyond human capability. To truly make conceptual contributions, researchers need to be capable of understanding the AI-generated structures and extracting the underlying concepts and ideas. When algorithms provide little explanatory reasoning alongside the output, scientists have to reverse-engineer the fundamental insights behind proposals based solely on examples. This task can be challenging as the output is often highly complex and thus not immediately accessible to humans. In this work we show how transferring part of the analysis process into an immersive virtual reality (VR) environment can assist researchers in developing an understanding of AI-generated solutions. We demonstrate the usefulness of VR in finding interpretable configurations of abstract graphs, representing Quantum Optics experiments. Thereby, we can manually discover new generalizations of AI-discoveries as well as new understanding in experimental quantum optics. Furthermore, it allows us to customize the search space in an informed way—as a human-in-the-loop—to achieve significantly faster subsequent discovery iterations. As concrete examples, with this technology, we discover a new resource-efficient 3-dimensional entanglement swapping scheme, as well as a 3-dimensional 4-particle Greenberger–Horne–Zeilinger-state analyzer. Our results show the potential of VR to enhance a researcher’s ability to derive knowledge from graph-based generative AI. This type of AI is a widely used abstract data representation in various scientific fields.
{"title":"Virtual reality for understanding artificial-intelligence-driven scientific discovery with an application in quantum optics","authors":"Philipp Schmidt, Sören Arlt, Carlos Ruiz-Gonzalez, Xuemei Gu, Carla Rodríguez, Mario Krenn","doi":"10.1088/2632-2153/ad5fdb","DOIUrl":"https://doi.org/10.1088/2632-2153/ad5fdb","url":null,"abstract":"Generative Artificial Intelligence (AI) models can propose solutions to scientific problems beyond human capability. To truly make conceptual contributions, researchers need to be capable of understanding the AI-generated structures and extracting the underlying concepts and ideas. When algorithms provide little explanatory reasoning alongside the output, scientists have to reverse-engineer the fundamental insights behind proposals based solely on examples. This task can be challenging as the output is often highly complex and thus not immediately accessible to humans. In this work we show how transferring part of the analysis process into an immersive virtual reality (VR) environment can assist researchers in developing an understanding of AI-generated solutions. We demonstrate the usefulness of VR in finding interpretable configurations of abstract graphs, representing Quantum Optics experiments. Thereby, we can manually discover new generalizations of AI-discoveries as well as new understanding in experimental quantum optics. Furthermore, it allows us to customize the search space in an informed way—as a human-in-the-loop—to achieve significantly faster subsequent discovery iterations. As concrete examples, with this technology, we discover a new resource-efficient 3-dimensional entanglement swapping scheme, as well as a 3-dimensional 4-particle Greenberger–Horne–Zeilinger-state analyzer. Our results show the potential of VR to enhance a researcher’s ability to derive knowledge from graph-based generative AI. This type of AI is a widely used abstract data representation in various scientific fields.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"15 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142197704","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-11DOI: 10.1088/2632-2153/ad65b5
M Paiano, S Martina, C Giannelli and F Caruso
The availability of data is limited in some fields, especially for object detection tasks, where it is necessary to have correctly labeled bounding boxes around each object. A notable example of such data scarcity is found in the domain of marine biology, where it is useful to develop methods to automatically detect submarine species for environmental monitoring. To address this data limitation, the state-of-the-art machine learning strategies employ two main approaches. The first involves pretraining models on existing datasets before generalizing to the specific domain of interest. The second strategy is to create synthetic datasets specifically tailored to the target domain using methods like copy-paste techniques or ad-hoc simulators. The first strategy often faces a significant domain shift, while the second demands custom solutions crafted for the specific task. In response to these challenges, here we propose a transfer learning framework that is valid for a generic scenario. In this framework, generated images help to improve the performances of an object detector in a few-real data regime. This is achieved through a diffusion-based generative model that was pretrained on large generic datasets. With respect to the state-of-the-art, we find that it is not necessary to fine tune the generative model on the specific domain of interest. We believe that this is an important advance because it mitigates the labor-intensive task of manual labeling the images in object detection tasks. We validate our approach focusing on fishes in an underwater environment, and on the more common domain of cars in an urban setting. Our method achieves detection performance comparable to models trained on thousands of images, using only a few hundreds of input data. Our results pave the way for new generative AI-based protocols for machine learning applications in various domains, for instance ranging from geophysics to biology and medicine.
{"title":"Transfer learning with generative models for object detection on limited datasets","authors":"M Paiano, S Martina, C Giannelli and F Caruso","doi":"10.1088/2632-2153/ad65b5","DOIUrl":"https://doi.org/10.1088/2632-2153/ad65b5","url":null,"abstract":"The availability of data is limited in some fields, especially for object detection tasks, where it is necessary to have correctly labeled bounding boxes around each object. A notable example of such data scarcity is found in the domain of marine biology, where it is useful to develop methods to automatically detect submarine species for environmental monitoring. To address this data limitation, the state-of-the-art machine learning strategies employ two main approaches. The first involves pretraining models on existing datasets before generalizing to the specific domain of interest. The second strategy is to create synthetic datasets specifically tailored to the target domain using methods like copy-paste techniques or ad-hoc simulators. The first strategy often faces a significant domain shift, while the second demands custom solutions crafted for the specific task. In response to these challenges, here we propose a transfer learning framework that is valid for a generic scenario. In this framework, generated images help to improve the performances of an object detector in a few-real data regime. This is achieved through a diffusion-based generative model that was pretrained on large generic datasets. With respect to the state-of-the-art, we find that it is not necessary to fine tune the generative model on the specific domain of interest. We believe that this is an important advance because it mitigates the labor-intensive task of manual labeling the images in object detection tasks. We validate our approach focusing on fishes in an underwater environment, and on the more common domain of cars in an urban setting. Our method achieves detection performance comparable to models trained on thousands of images, using only a few hundreds of input data. Our results pave the way for new generative AI-based protocols for machine learning applications in various domains, for instance ranging from geophysics to biology and medicine.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"192 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141931252","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-05DOI: 10.1088/2632-2153/ad6830
André Sequeira, Luis Paulo Santos and Luis Soares Barbosa
This research explores the trainability of Parameterized Quantum Circuit-based policies in Reinforcement Learning, an area that has recently seen a surge in empirical exploration. While some studies suggest improved sample complexity using quantum gradient estimation, the efficient trainability of these policies remains an open question. Our findings reveal significant challenges, including standard Barren Plateaus with exponentially small gradients and gradient explosion. These phenomena depend on the type of basis-state partitioning and the mapping of these partitions onto actions. For a polynomial number of actions, a trainable window can be ensured with a polynomial number of measurements if a contiguous-like partitioning of basis-states is employed. These results are empirically validated in a multi-armed bandit environment.
{"title":"Trainability issues in quantum policy gradients","authors":"André Sequeira, Luis Paulo Santos and Luis Soares Barbosa","doi":"10.1088/2632-2153/ad6830","DOIUrl":"https://doi.org/10.1088/2632-2153/ad6830","url":null,"abstract":"This research explores the trainability of Parameterized Quantum Circuit-based policies in Reinforcement Learning, an area that has recently seen a surge in empirical exploration. While some studies suggest improved sample complexity using quantum gradient estimation, the efficient trainability of these policies remains an open question. Our findings reveal significant challenges, including standard Barren Plateaus with exponentially small gradients and gradient explosion. These phenomena depend on the type of basis-state partitioning and the mapping of these partitions onto actions. For a polynomial number of actions, a trainable window can be ensured with a polynomial number of measurements if a contiguous-like partitioning of basis-states is employed. These results are empirically validated in a multi-armed bandit environment.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"76 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141931176","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-05DOI: 10.1088/2632-2153/ad652c
Khaled Kahouli, Stefaan Simon Pierre Hessmann, Klaus-Robert Müller, Shinichi Nakajima, Stefan Gugler and Niklas Wolf Andreas Gebauer
Molecular relaxation, finding the equilibrium state of a non-equilibrium structure, is an essential component of computational chemistry to understand reactivity. Classical force field (FF) methods often rely on insufficient local energy minimization, while neural network FF models require large labeled datasets encompassing both equilibrium and non-equilibrium structures. As a remedy, we propose MoreRed, molecular relaxation by reverse diffusion, a conceptually novel and purely statistical approach where non-equilibrium structures are treated as noisy instances of their corresponding equilibrium states. To enable the denoising of arbitrarily noisy inputs via a generative diffusion model, we further introduce a novel diffusion time step predictor. Notably, MoreRed learns a simpler pseudo potential energy surface (PES) instead of the complex physical PES. It is trained on a significantly smaller, and thus computationally cheaper, dataset consisting of solely unlabeled equilibrium structures, avoiding the computation of non-equilibrium structures altogether. We compare MoreRed to classical FFs, equivariant neural network FFs trained on a large dataset of equilibrium and non-equilibrium data, as well as a semi-empirical tight-binding model. To assess this quantitatively, we evaluate the root-mean-square deviation between the found equilibrium structures and the reference equilibrium structures as well as their energies.
{"title":"Molecular relaxation by reverse diffusion with time step prediction","authors":"Khaled Kahouli, Stefaan Simon Pierre Hessmann, Klaus-Robert Müller, Shinichi Nakajima, Stefan Gugler and Niklas Wolf Andreas Gebauer","doi":"10.1088/2632-2153/ad652c","DOIUrl":"https://doi.org/10.1088/2632-2153/ad652c","url":null,"abstract":"Molecular relaxation, finding the equilibrium state of a non-equilibrium structure, is an essential component of computational chemistry to understand reactivity. Classical force field (FF) methods often rely on insufficient local energy minimization, while neural network FF models require large labeled datasets encompassing both equilibrium and non-equilibrium structures. As a remedy, we propose MoreRed, molecular relaxation by reverse diffusion, a conceptually novel and purely statistical approach where non-equilibrium structures are treated as noisy instances of their corresponding equilibrium states. To enable the denoising of arbitrarily noisy inputs via a generative diffusion model, we further introduce a novel diffusion time step predictor. Notably, MoreRed learns a simpler pseudo potential energy surface (PES) instead of the complex physical PES. It is trained on a significantly smaller, and thus computationally cheaper, dataset consisting of solely unlabeled equilibrium structures, avoiding the computation of non-equilibrium structures altogether. We compare MoreRed to classical FFs, equivariant neural network FFs trained on a large dataset of equilibrium and non-equilibrium data, as well as a semi-empirical tight-binding model. To assess this quantitatively, we evaluate the root-mean-square deviation between the found equilibrium structures and the reference equilibrium structures as well as their energies.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"303 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141931253","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-04DOI: 10.1088/2632-2153/ad66af
Xiaobo Zhu, Yan Wu, Jin Che, Chao Wang, Liying Wang and Zhanheng Chen
Representation learning over graph networks has recently gained popularity, with many models showing promising results. However, several challenges remain: (1) most methods are designed for static or discrete-time dynamic graphs; (2) existing continuous-time dynamic graph algorithms focus on a single evolving perspective; and (3) many continuous-time dynamic graph approaches necessitate numerous temporal neighbors to capture long-term dependencies. In response, this paper introduces a Multi-Perspective Feedback-Attention Coupling (MPFA) model. MPFA incorporates information from both evolving and original perspectives to effectively learn the complex dynamics of dynamic graph evolution processes. The evolving perspective considers the current state of historical interaction events of nodes and uses a temporal attention module to aggregate current state information. This perspective also makes it possible to capture long-term dependencies of nodes using a small number of temporal neighbors. Meanwhile, the original perspective utilizes a feedback attention module with growth characteristic coefficients to aggregate the original state information of node interactions. Experimental results on one dataset organized by ourselves and seven public datasets validate the effectiveness and competitiveness of our proposed model.
{"title":"Multi-perspective feedback-attention coupling model for continuous-time dynamic graphs","authors":"Xiaobo Zhu, Yan Wu, Jin Che, Chao Wang, Liying Wang and Zhanheng Chen","doi":"10.1088/2632-2153/ad66af","DOIUrl":"https://doi.org/10.1088/2632-2153/ad66af","url":null,"abstract":"Representation learning over graph networks has recently gained popularity, with many models showing promising results. However, several challenges remain: (1) most methods are designed for static or discrete-time dynamic graphs; (2) existing continuous-time dynamic graph algorithms focus on a single evolving perspective; and (3) many continuous-time dynamic graph approaches necessitate numerous temporal neighbors to capture long-term dependencies. In response, this paper introduces a Multi-Perspective Feedback-Attention Coupling (MPFA) model. MPFA incorporates information from both evolving and original perspectives to effectively learn the complex dynamics of dynamic graph evolution processes. The evolving perspective considers the current state of historical interaction events of nodes and uses a temporal attention module to aggregate current state information. This perspective also makes it possible to capture long-term dependencies of nodes using a small number of temporal neighbors. Meanwhile, the original perspective utilizes a feedback attention module with growth characteristic coefficients to aggregate the original state information of node interactions. Experimental results on one dataset organized by ourselves and seven public datasets validate the effectiveness and competitiveness of our proposed model.","PeriodicalId":33757,"journal":{"name":"Machine Learning Science and Technology","volume":"14 1","pages":""},"PeriodicalIF":6.8,"publicationDate":"2024-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141968649","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"物理与天体物理","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：