In this short note, we prove a comparision theorem between Levine-Serpé's equivariant higher Chow... more In this short note, we prove a comparision theorem between Levine-Serpé's equivariant higher Chow groups of an algebraic variety equipped with an action of a finite group and ordinary higher Chow groups of its fixed points. As a consequence, we show that the equivariant motivic spectral sequence degenerates rationally. This yields a Riemann-Roch Theorem for equivariant algebraic K-theory.
This paper addresses a cornerstone of Automated Machine Learning: the problem of rapidly uncoveri... more This paper addresses a cornerstone of Automated Machine Learning: the problem of rapidly uncovering which machine learning algorithm performs best on a new dataset. Our approach leverages performances of such algorithms on datasets to which they have been previously exposed, i.e., implementing a form of meta-learning. More specifically, the problem is cast as a REVEAL Reinforcement Learning (RL) game: the meta-learning problem is wrapped into a RL environment in which an agent can start, pause, or resume training various machine learning algorithms to progressively “reveal” their learning curves. The learned policy is then applied to quickly uncover the best algorithm on a new dataset. While other similar approaches, such as Freeze-Thaw, were proposed in the past, using Bayesian optimization, our methodology is, to the best of our knowledge, the first that trains a RL agent to do this task on previous datasets. Using real and artificial data, we show that our new RL-based meta-learn...
There are many powerful techniques for automated termination analysis of term rewrite systems (TR... more There are many powerful techniques for automated termination analysis of term rewrite systems (TRSs). However, up to now they have hardly been used for real programming languages. In this talk, we describe recent results which permit the application of existing techniques from term rewriting in order to prove termination of programs. We discuss two possible approaches: 1. One could translate programs into TRSs and then use existing tools to verify termination of the resulting TRSs. 2. One could adapt TRS-techniques to the respective programming languages in order to analyze programs directly. We present such approaches for the functional language Haskell and the logic language Prolog. Our results have been implemented in the termination provers AProVE and Polytool. In order to handle termination problems resulting from real programs, these provers had to be coupled with modern SAT solvers, since the automation of the TRS-termination techniques had to improve significantly. Our resul...
The paper introduces the experimental research results of superplastic forming (SPF) of 15 AA7075... more The paper introduces the experimental research results of superplastic forming (SPF) of 15 AA7075 aluminum alloy sheet. The response surface methodology (RSM) based on a Box-Behnken 16 design (BBD) were used to study the influence of process parameters on the superplastic forming 17 ability. The analysis show the relationship between the relative height of the product and the main 18 process parameters: forming pressure of 0.7-0.9 MPa, deformation temperature of 500-5300C and 19 forming time of 20-40 minutes. The experimental results are consistent with the general trend of the 20 superplastic forming process: the relative height of the product increases with increasing pressure, 21 temperature, and forming time. However, there exist limit values of forming time, where the law of 22 the influence of temperature and forming pressure on relative height is reversed. Therefore, in each 23 specific machining case, it is necessary to select the range of appropriate process parameters to g...
In this work, the relative yields of aqueous secondary organic aerosols (aqSOAs) at the air–liqui... more In this work, the relative yields of aqueous secondary organic aerosols (aqSOAs) at the air–liquid (a–l) interface are investigated between photochemical and dark aging using in situ time-of-flight secondary ion mass spectrometry (ToF-SIMS). Our results show that dark aging is an important source of aqSOAs despite a lack of photochemical drivers. Photochemical reactions of glyoxal and hydroxyl radicals (•OH) produce oligomers and cluster ions at the aqueous surface. Interestingly, different oligomers and cluster ions form intensely in the dark at the a–l interface, contrary to the notion that oligomer formation mainly depends on light irradiation. Furthermore, cluster ions form readily during dark aging and have a higher water molecule adsorption ability. This finding is supported by the observation of more frequent organic water cluster ion formation. The relative yields of water clusters in the form of protonated and hydroxide ions are presented using van Krevelen diagrams to expl...
This paper analyzes volatility models and their risk forecasting abilities with the presence of j... more This paper analyzes volatility models and their risk forecasting abilities with the presence of jumps for the Vietnam Stock Exchange (VSE). We apply GARCH-type models, which capture short and long memory and the leverage effect, estimated from both raw and filtered returns. The data sample covers two VSE indexes, the VN index and HNX index, provided by the Ho Chi Minh City Stock Exchange (HOSE) and Hanoi Stock Exchange (HNX), respectively, during the period 2007 - 2015. The empirical results reveal that the FIAPARCH model is the most suitable model for the VN index and HNX index.
La classification (supervisee, non supervisee et semi-supervisee) est une thematique importante d... more La classification (supervisee, non supervisee et semi-supervisee) est une thematique importante de la fouille de donnees. Dans cette these, nous nous concentrons sur le developpement d'approches d'optimisation pour resoudre certains types des problemes issus de la classification de donnees. Premierement, nous avons examine et developpe des algorithmes pour resoudre deux problemes classiques en apprentissage non supervisee : la maximisation du critere de modularite pour la detection de communautes dans des reseaux complexes et les cartes auto-organisatrices. Deuxiemement, pour l'apprentissage semi-supervisee, nous proposons des algorithmes efficaces pour le probleme de selection de variables en semi-supervisee Machines a vecteurs de support. Finalement, dans la derniere partie de la these, nous considerons le probleme de selection de variables en Machines a vecteurs de support multi-classes. Tous ces problemes d'optimisation sont non convexe de tres grande dimension e...
Actuator fault estimation problem is tackled in this paper. The actuator faults are modeled in th... more Actuator fault estimation problem is tackled in this paper. The actuator faults are modeled in the form of multiplicative faults by using effectiveness factors representing the loss of efficiency of the actuators. The main contribution of this paper lies in the capability of dealing with the presented problem by using a switched LPV observer approach. The LTI system in the presence of faulty actuators is rewritten as a switched LPV system by considering the control inputs as scheduling parameters. Then, the actuator faults and the system states are estimated using a switched LPV extended observer. The observer gain is derived, based on the LMIs solution for the switched LPV systems. The presented actuator fault estimation approach is validated by two illustrative examples, the first one about a damper fault estimation of a semi-active suspension system, and the second one concerned to fault estimations on a multiple actuators system.
The aims of this study are to examine the effects of 24-form Tai Chi exercise in six months on ph... more The aims of this study are to examine the effects of 24-form Tai Chi exercise in six months on physical fitness, blood pressure and perceived health as well as sleep quality, visual attention, and balance ability of older people living in dwelling community. This study is also aimed to compare differences in physical fitness and subjectively rated health between German and Vietnamese older adults in terms cross-cultural study between Vietnamese and German samples. The subjects were divided randomly into two groups, Training group and Control group. The subjects were expected to consent or volunteer. Participants in Training group (forty eight subjects ranging in age from 60 years to 80 (69.02±5.16) were assigned 6-months Tai Chi training in Vinh city, Vietnam. Participants in the control group (forty eight subjects ranging in age from 60 years to 79 (68.72±4.94) were instructed to maintain their routine daily activities and not to begin any new exercise programs. The SFT were used i...
In the past, much effort of healthcare decision support systems were focused on the data acquisit... more In the past, much effort of healthcare decision support systems were focused on the data acquisition and storage, in order to allow the use of this data at some later point in time. Medical data was used in static manner, for analytical purposes, in order to verify the undertaken decisions. Due to the massive economical impact of today’s health system, great changes in medical treatments are notable. Apart of the humanitarian and healing nature of medicine, this industry is becoming more and more business like. The exploitation of evidence-based guidelines becomes a priority concern, as the awareness of the importance of knowledge management rises. Consequently, interoperability between medical information systems is becoming a necessity in modern health care. Under strong security measures, health care organisations are striking to unite and share their (partly very high sensitive) data assets in order to achieve a wider knowledge base and to provide a matured decision support serv...
Due to the work of many authors in the last decades, given an algebraic orbifold (smooth proper D... more Due to the work of many authors in the last decades, given an algebraic orbifold (smooth proper Deligne-Mumford stack with trivial generic stabilizer), one can construct its orbifold Chow ring and orbifold Grothendieck ring, and relate them by the orbifold Chern character map, generalizing the fundamental work of Chen-Ruan on orbifold cohomology. In this paper, we extend this theory naturally to higher Chow groups and higher algebraic K-theory, mainly following the work of Jarvis-Kaufmann-Kimura and Edidin-Jarvis-Kimura.
We consider a general equilibrium model in asset markets with a countable set of states and expec... more We consider a general equilibrium model in asset markets with a countable set of states and expected risk averse utilities. The agents do not have the same beliefs. We use the methods in Le Van Truong Xuan (JME, 2001) but one of their assumption which is crucial for obtaining their result cannot be accepted in our model when the number of states is countable. We give a proof of existence of equilibrium when the number of states is infinite or finite.
We combine graph neural networks with Gaussian Process regression through deep graph kernel learn... more We combine graph neural networks with Gaussian Process regression through deep graph kernel learning and demonstrate its robustness on quantitative structure-activity relationship (QSAR) modeling tasks. Equipped with such a model, a Bayesian optimization experiment on chemical space is conducted and compared against the time-stamped acquisition records of a real-world, time-sensitive molecular optimization mission: the identification of potent inhibitors of the main protease of SARS-CoV-2, the viral pathogen responsible for the COVID pandemic.
2019 Computing in Cardiology Conference (CinC), 2019
As part of the PhysioNet/Computing in Cardiology Challenge 2019, we propose a neural network call... more As part of the PhysioNet/Computing in Cardiology Challenge 2019, we propose a neural network called AEC-Net to early detect sepsis based on physiological data. AEC-Net consists of two main components: 1) an Auto Encoder for dimension reduction and feature extraction, and 2) a Fully Connected Neural Network (FCNN) taking the extracted features by the Auto Encoder as the input and generating prediction of sepsis as output. The losses of both the Auto Encoder and FCNN are minimized concurrently. This concurrent optimization helps AEC-Net to have a better generalization and the extracted features by Auto Encoder to be more relevant to the classification problem. Finally, we propose an ensemble method of AEC-Net, Random Forest and Gradient Boosting Decision Trees to achieve a better prediction. We train our proposed models using data from 40336 patients with 40 physiological features ranging from 8 to 336 hours. Our team Infolab USC evaluated Ensemble with the hidden full test set of the Physionet Challenge 2019, and achieved a Utility score of 0.284 and 24 th place in the challenge.
The goal of this work was to study the effects of cyclic close die forging on the microstructure ... more The goal of this work was to study the effects of cyclic close die forging on the microstructure and mechanical properties of Ti–5Al–3Mo–1.5V alloy, which was produced in Vietnam. The factors considered include the deformation temperature (Td), at 850 °C, 900 °C, and 950 °C, and the number of cycles performed while forging in closed die (n)— 3, 6, and 9 times. The responses measured were average grain diameter (dtb) and tensile stress (σb). The results indicate that the smallest average grain size of 1 μm could be obtained at Td = 900 °C, n = 9 times and the tensile stresses were enhanced. The experimental results we obtained also suggest that the microstructure of Ti–5Al–3Mo–1.5V alloy is accordant for superplastic deformation. The superplastic forming of this alloy can show maximum elongation of 1000% or more.
Medical Image Computing and Computer Assisted Intervention – MICCAI 2020, 2020
Image synthesis from corrupted contrasts increases the diversity of diagnostic information availa... more Image synthesis from corrupted contrasts increases the diversity of diagnostic information available for many neurological diseases. Recently the image-to-image translation has experienced significant levels of interest within medical research, beginning with the successful use of the Generative Adversarial Network (GAN) to the introduction of cyclic constraint extended to multiple domains. However, in current approaches, there is no guarantee that the mapping between the two image domains would be unique or one-to-one. In this paper, we introduce a novel approach to unpaired image-to-image translation based on the invertible architecture. The invertible property of the flow-based architecture assures a cycle-consistency of image-to-image translation without additional loss functions. We utilize the temporal information between consecutive slices to provide more constraints to the optimization for transforming one domain to another in unpaired volumetric medical images. To capture temporal structures in the medical images, we explore the displacement between the consecutive slices using a deformation field. In our approach, the deformation field is used as a guidance to keep the translated slides realistic and consistent across the translation. The experimental results have shown that the synthesized images using our proposed approach are able to archive a competitive performance in terms of mean squared error, peak signalto-noise ratio, and structural similarity index when compared with the existing deep learning-based methods on three standard datasets, i.e. HCP, MRBrainS13 and Brats2019.
In this short note, we prove a comparision theorem between Levine-Serpé's equivariant higher Chow... more In this short note, we prove a comparision theorem between Levine-Serpé's equivariant higher Chow groups of an algebraic variety equipped with an action of a finite group and ordinary higher Chow groups of its fixed points. As a consequence, we show that the equivariant motivic spectral sequence degenerates rationally. This yields a Riemann-Roch Theorem for equivariant algebraic K-theory.
This paper addresses a cornerstone of Automated Machine Learning: the problem of rapidly uncoveri... more This paper addresses a cornerstone of Automated Machine Learning: the problem of rapidly uncovering which machine learning algorithm performs best on a new dataset. Our approach leverages performances of such algorithms on datasets to which they have been previously exposed, i.e., implementing a form of meta-learning. More specifically, the problem is cast as a REVEAL Reinforcement Learning (RL) game: the meta-learning problem is wrapped into a RL environment in which an agent can start, pause, or resume training various machine learning algorithms to progressively “reveal” their learning curves. The learned policy is then applied to quickly uncover the best algorithm on a new dataset. While other similar approaches, such as Freeze-Thaw, were proposed in the past, using Bayesian optimization, our methodology is, to the best of our knowledge, the first that trains a RL agent to do this task on previous datasets. Using real and artificial data, we show that our new RL-based meta-learn...
There are many powerful techniques for automated termination analysis of term rewrite systems (TR... more There are many powerful techniques for automated termination analysis of term rewrite systems (TRSs). However, up to now they have hardly been used for real programming languages. In this talk, we describe recent results which permit the application of existing techniques from term rewriting in order to prove termination of programs. We discuss two possible approaches: 1. One could translate programs into TRSs and then use existing tools to verify termination of the resulting TRSs. 2. One could adapt TRS-techniques to the respective programming languages in order to analyze programs directly. We present such approaches for the functional language Haskell and the logic language Prolog. Our results have been implemented in the termination provers AProVE and Polytool. In order to handle termination problems resulting from real programs, these provers had to be coupled with modern SAT solvers, since the automation of the TRS-termination techniques had to improve significantly. Our resul...
The paper introduces the experimental research results of superplastic forming (SPF) of 15 AA7075... more The paper introduces the experimental research results of superplastic forming (SPF) of 15 AA7075 aluminum alloy sheet. The response surface methodology (RSM) based on a Box-Behnken 16 design (BBD) were used to study the influence of process parameters on the superplastic forming 17 ability. The analysis show the relationship between the relative height of the product and the main 18 process parameters: forming pressure of 0.7-0.9 MPa, deformation temperature of 500-5300C and 19 forming time of 20-40 minutes. The experimental results are consistent with the general trend of the 20 superplastic forming process: the relative height of the product increases with increasing pressure, 21 temperature, and forming time. However, there exist limit values of forming time, where the law of 22 the influence of temperature and forming pressure on relative height is reversed. Therefore, in each 23 specific machining case, it is necessary to select the range of appropriate process parameters to g...
In this work, the relative yields of aqueous secondary organic aerosols (aqSOAs) at the air–liqui... more In this work, the relative yields of aqueous secondary organic aerosols (aqSOAs) at the air–liquid (a–l) interface are investigated between photochemical and dark aging using in situ time-of-flight secondary ion mass spectrometry (ToF-SIMS). Our results show that dark aging is an important source of aqSOAs despite a lack of photochemical drivers. Photochemical reactions of glyoxal and hydroxyl radicals (•OH) produce oligomers and cluster ions at the aqueous surface. Interestingly, different oligomers and cluster ions form intensely in the dark at the a–l interface, contrary to the notion that oligomer formation mainly depends on light irradiation. Furthermore, cluster ions form readily during dark aging and have a higher water molecule adsorption ability. This finding is supported by the observation of more frequent organic water cluster ion formation. The relative yields of water clusters in the form of protonated and hydroxide ions are presented using van Krevelen diagrams to expl...
This paper analyzes volatility models and their risk forecasting abilities with the presence of j... more This paper analyzes volatility models and their risk forecasting abilities with the presence of jumps for the Vietnam Stock Exchange (VSE). We apply GARCH-type models, which capture short and long memory and the leverage effect, estimated from both raw and filtered returns. The data sample covers two VSE indexes, the VN index and HNX index, provided by the Ho Chi Minh City Stock Exchange (HOSE) and Hanoi Stock Exchange (HNX), respectively, during the period 2007 - 2015. The empirical results reveal that the FIAPARCH model is the most suitable model for the VN index and HNX index.
La classification (supervisee, non supervisee et semi-supervisee) est une thematique importante d... more La classification (supervisee, non supervisee et semi-supervisee) est une thematique importante de la fouille de donnees. Dans cette these, nous nous concentrons sur le developpement d'approches d'optimisation pour resoudre certains types des problemes issus de la classification de donnees. Premierement, nous avons examine et developpe des algorithmes pour resoudre deux problemes classiques en apprentissage non supervisee : la maximisation du critere de modularite pour la detection de communautes dans des reseaux complexes et les cartes auto-organisatrices. Deuxiemement, pour l'apprentissage semi-supervisee, nous proposons des algorithmes efficaces pour le probleme de selection de variables en semi-supervisee Machines a vecteurs de support. Finalement, dans la derniere partie de la these, nous considerons le probleme de selection de variables en Machines a vecteurs de support multi-classes. Tous ces problemes d'optimisation sont non convexe de tres grande dimension e...
Actuator fault estimation problem is tackled in this paper. The actuator faults are modeled in th... more Actuator fault estimation problem is tackled in this paper. The actuator faults are modeled in the form of multiplicative faults by using effectiveness factors representing the loss of efficiency of the actuators. The main contribution of this paper lies in the capability of dealing with the presented problem by using a switched LPV observer approach. The LTI system in the presence of faulty actuators is rewritten as a switched LPV system by considering the control inputs as scheduling parameters. Then, the actuator faults and the system states are estimated using a switched LPV extended observer. The observer gain is derived, based on the LMIs solution for the switched LPV systems. The presented actuator fault estimation approach is validated by two illustrative examples, the first one about a damper fault estimation of a semi-active suspension system, and the second one concerned to fault estimations on a multiple actuators system.
The aims of this study are to examine the effects of 24-form Tai Chi exercise in six months on ph... more The aims of this study are to examine the effects of 24-form Tai Chi exercise in six months on physical fitness, blood pressure and perceived health as well as sleep quality, visual attention, and balance ability of older people living in dwelling community. This study is also aimed to compare differences in physical fitness and subjectively rated health between German and Vietnamese older adults in terms cross-cultural study between Vietnamese and German samples. The subjects were divided randomly into two groups, Training group and Control group. The subjects were expected to consent or volunteer. Participants in Training group (forty eight subjects ranging in age from 60 years to 80 (69.02±5.16) were assigned 6-months Tai Chi training in Vinh city, Vietnam. Participants in the control group (forty eight subjects ranging in age from 60 years to 79 (68.72±4.94) were instructed to maintain their routine daily activities and not to begin any new exercise programs. The SFT were used i...
In the past, much effort of healthcare decision support systems were focused on the data acquisit... more In the past, much effort of healthcare decision support systems were focused on the data acquisition and storage, in order to allow the use of this data at some later point in time. Medical data was used in static manner, for analytical purposes, in order to verify the undertaken decisions. Due to the massive economical impact of today’s health system, great changes in medical treatments are notable. Apart of the humanitarian and healing nature of medicine, this industry is becoming more and more business like. The exploitation of evidence-based guidelines becomes a priority concern, as the awareness of the importance of knowledge management rises. Consequently, interoperability between medical information systems is becoming a necessity in modern health care. Under strong security measures, health care organisations are striking to unite and share their (partly very high sensitive) data assets in order to achieve a wider knowledge base and to provide a matured decision support serv...
Due to the work of many authors in the last decades, given an algebraic orbifold (smooth proper D... more Due to the work of many authors in the last decades, given an algebraic orbifold (smooth proper Deligne-Mumford stack with trivial generic stabilizer), one can construct its orbifold Chow ring and orbifold Grothendieck ring, and relate them by the orbifold Chern character map, generalizing the fundamental work of Chen-Ruan on orbifold cohomology. In this paper, we extend this theory naturally to higher Chow groups and higher algebraic K-theory, mainly following the work of Jarvis-Kaufmann-Kimura and Edidin-Jarvis-Kimura.
We consider a general equilibrium model in asset markets with a countable set of states and expec... more We consider a general equilibrium model in asset markets with a countable set of states and expected risk averse utilities. The agents do not have the same beliefs. We use the methods in Le Van Truong Xuan (JME, 2001) but one of their assumption which is crucial for obtaining their result cannot be accepted in our model when the number of states is countable. We give a proof of existence of equilibrium when the number of states is infinite or finite.
We combine graph neural networks with Gaussian Process regression through deep graph kernel learn... more We combine graph neural networks with Gaussian Process regression through deep graph kernel learning and demonstrate its robustness on quantitative structure-activity relationship (QSAR) modeling tasks. Equipped with such a model, a Bayesian optimization experiment on chemical space is conducted and compared against the time-stamped acquisition records of a real-world, time-sensitive molecular optimization mission: the identification of potent inhibitors of the main protease of SARS-CoV-2, the viral pathogen responsible for the COVID pandemic.
2019 Computing in Cardiology Conference (CinC), 2019
As part of the PhysioNet/Computing in Cardiology Challenge 2019, we propose a neural network call... more As part of the PhysioNet/Computing in Cardiology Challenge 2019, we propose a neural network called AEC-Net to early detect sepsis based on physiological data. AEC-Net consists of two main components: 1) an Auto Encoder for dimension reduction and feature extraction, and 2) a Fully Connected Neural Network (FCNN) taking the extracted features by the Auto Encoder as the input and generating prediction of sepsis as output. The losses of both the Auto Encoder and FCNN are minimized concurrently. This concurrent optimization helps AEC-Net to have a better generalization and the extracted features by Auto Encoder to be more relevant to the classification problem. Finally, we propose an ensemble method of AEC-Net, Random Forest and Gradient Boosting Decision Trees to achieve a better prediction. We train our proposed models using data from 40336 patients with 40 physiological features ranging from 8 to 336 hours. Our team Infolab USC evaluated Ensemble with the hidden full test set of the Physionet Challenge 2019, and achieved a Utility score of 0.284 and 24 th place in the challenge.
The goal of this work was to study the effects of cyclic close die forging on the microstructure ... more The goal of this work was to study the effects of cyclic close die forging on the microstructure and mechanical properties of Ti–5Al–3Mo–1.5V alloy, which was produced in Vietnam. The factors considered include the deformation temperature (Td), at 850 °C, 900 °C, and 950 °C, and the number of cycles performed while forging in closed die (n)— 3, 6, and 9 times. The responses measured were average grain diameter (dtb) and tensile stress (σb). The results indicate that the smallest average grain size of 1 μm could be obtained at Td = 900 °C, n = 9 times and the tensile stresses were enhanced. The experimental results we obtained also suggest that the microstructure of Ti–5Al–3Mo–1.5V alloy is accordant for superplastic deformation. The superplastic forming of this alloy can show maximum elongation of 1000% or more.
Medical Image Computing and Computer Assisted Intervention – MICCAI 2020, 2020
Image synthesis from corrupted contrasts increases the diversity of diagnostic information availa... more Image synthesis from corrupted contrasts increases the diversity of diagnostic information available for many neurological diseases. Recently the image-to-image translation has experienced significant levels of interest within medical research, beginning with the successful use of the Generative Adversarial Network (GAN) to the introduction of cyclic constraint extended to multiple domains. However, in current approaches, there is no guarantee that the mapping between the two image domains would be unique or one-to-one. In this paper, we introduce a novel approach to unpaired image-to-image translation based on the invertible architecture. The invertible property of the flow-based architecture assures a cycle-consistency of image-to-image translation without additional loss functions. We utilize the temporal information between consecutive slices to provide more constraints to the optimization for transforming one domain to another in unpaired volumetric medical images. To capture temporal structures in the medical images, we explore the displacement between the consecutive slices using a deformation field. In our approach, the deformation field is used as a guidance to keep the translated slides realistic and consistent across the translation. The experimental results have shown that the synthesized images using our proposed approach are able to archive a competitive performance in terms of mean squared error, peak signalto-noise ratio, and structural similarity index when compared with the existing deep learning-based methods on three standard datasets, i.e. HCP, MRBrainS13 and Brats2019.
Uploads
Papers by Manh Nguyen