摘要 :
While current neural networks (NNs) are becoming good at deriving single types of abstractions for a small set of phenomena, for example, using a single NN to predict a flow velocity field, NNs are not good at composing large syst...
展开
While current neural networks (NNs) are becoming good at deriving single types of abstractions for a small set of phenomena, for example, using a single NN to predict a flow velocity field, NNs are not good at composing large systems as compositions of small phenomena and reasoning about their interactions. We want to study how NNs build both the abstraction and composition of phenomena when a single NN model cannot suffice. Rather than a single NN that learns one physical or social phenomenon, we want a group of NNs that learn to abstract, compose, reason, and correct the behaviors of different parts in a system. In this paper, we investigate the joint use of Physics-Informed (Navier-Stokes equations) Deep Neural Networks (i.e., Deconvolutional Neural Networks) as well as Geometric Deep Learning (i.e., Graph Neural Networks) to learn and compose fluid component behavior. Our models successfully predict the fluid flows and their composition behaviors (i.e., velocity fields) with an accuracy of about 99%.
收起
摘要 :
While current neural networks (NNs) are becoming good at deriving single types of abstractions for a small set of phenomena, for example, using a single NN to predict a flow velocity field, NNs are not good at composing large syst...
展开
While current neural networks (NNs) are becoming good at deriving single types of abstractions for a small set of phenomena, for example, using a single NN to predict a flow velocity field, NNs are not good at composing large systems as compositions of small phenomena and reasoning about their interactions. We want to study how NNs build both the abstraction and composition of phenomena when a single NN model cannot suffice. Rather than a single NN that learns one physical or social phenomenon, we want a group of NNs that learn to abstract, compose, reason, and correct the behaviors of different parts in a system. In this paper, we investigate the joint use of Physics-Informed (Navier-Stokes equations) Deep Neural Networks (i.e., Deconvolutional Neural Networks) as well as Geometric Deep Learning (i.e., Graph Neural Networks) to learn and compose fluid component behavior. Our models successfully predict the fluid flows and their composition behaviors (i.e., velocity fields) with an accuracy of about 99%.
收起
摘要 :
Global optimization of aerodynamic shapes requires a large number of expensive CFD simulations because of the high dimensionality of the design space. One means to combat that problem is to reduce the dimension of the design space...
展开
Global optimization of aerodynamic shapes requires a large number of expensive CFD simulations because of the high dimensionality of the design space. One means to combat that problem is to reduce the dimension of the design space-for example, by constructing low dimensional parametric functions (such as PARSEC and others)-and then optimizing over those parameters instead. Such approaches require first a parametric function that compactly describes useful variation in airfoil shape-a non-trivial and error-prone task. In contrast, we propose to use a deep generative model of aerodynamic designs (specifically airfoils) that reduces the dimensionality of the optimization problem by learning from shape variations in the UIUC airfoil database. We show that our data-driven model both (1) learns realistic and compact airfoil shape representations and (2) empirically accelerates optimization convergence by over an order of magnitude.
收起
摘要 :
Global optimization of aerodynamic shapes requires a large number of expensive CFD simulations because of the high dimensionality of the design space. One means to combat that problem is to reduce the dimension of the design space...
展开
Global optimization of aerodynamic shapes requires a large number of expensive CFD simulations because of the high dimensionality of the design space. One means to combat that problem is to reduce the dimension of the design space-for example, by constructing low dimensional parametric functions (such as PARSEC and others)-and then optimizing over those parameters instead. Such approaches require first a parametric function that compactly describes useful variation in airfoil shape-a non-trivial and error-prone task. In contrast, we propose to use a deep generative model of aerodynamic designs (specifically airfoils) that reduces the dimensionality of the optimization problem by learning from shape variations in the UIUC airfoil database. We show that our data-driven model both (1) learns realistic and compact airfoil shape representations and (2) empirically accelerates optimization convergence by over an order of magnitude.
收起
摘要 :
Learning a new language is hard, but learning to use it confidently in conversations with native speakers is even harder. From our field research with language learners, with support from Cognitive Psychology and Second Language A...
展开
Learning a new language is hard, but learning to use it confidently in conversations with native speakers is even harder. From our field research with language learners, with support from Cognitive Psychology and Second Language Acquisition, we argue for the value of contextual microlearning in the many breaks spread across different places and throughout the day. We present a mobile application that supports such microlearning by leveraging the location-based service Foursquare to automatically provide contextually relevant content in the world's major cities. In an evaluation of Mandarin Chinese learning, a four-week, 23-user study spanning Beijing and Shanghai compared this contextual system to a system based on word frequency. Study sessions with the contextual version lasted half as long but occurred in twice as many places as sessions with the frequency version, suggesting a complementary relationship between the two approaches.
收起
摘要 :
Learning a new language is hard, but learning to use it confidently in conversations with native speakers is even harder. From our field research with language learners, with support from Cognitive Psychology and Second Language A...
展开
Learning a new language is hard, but learning to use it confidently in conversations with native speakers is even harder. From our field research with language learners, with support from Cognitive Psychology and Second Language Acquisition, we argue for the value of contextual microlearning in the many breaks spread across different places and throughout the day. We present a mobile application that supports such microlearning by leveraging the location-based service Foursquare to automatically provide contextually relevant content in the world's major cities. In an evaluation of Mandarin Chinese learning, a four-week, 23-user study spanning Beijing and Shanghai compared this contextual system to a system based on word frequency. Study sessions with the contextual version lasted half as long but occurred in twice as many places as sessions with the frequency version, suggesting a complementary relationship between the two approaches.
收起
摘要 :
Learning a new language is hard, but learning to use it confidently in conversations with native speakers is even harder. From our field research with language learners, with support from Cognitive Psychology and Second Language A...
展开
Learning a new language is hard, but learning to use it confidently in conversations with native speakers is even harder. From our field research with language learners, with support from Cognitive Psychology and Second Language Acquisition, we argue for the value of contextual microlearning in the many breaks spread across different places and throughout the day. We present a mobile application that supports such microlearning by leveraging the location-based service Foursquare to automatically provide contextually relevant content in the world's major cities. In an evaluation of Mandarin Chinese learning, a four-week, 23-user study spanning Beijing and Shanghai compared this contextual system to a system based on word frequency. Study sessions with the contextual version lasted half as long but occurred in twice as many places as sessions with the frequency version, suggesting a complementary relationship between the two approaches.
收起
摘要 :
This paper presents a reconfigurable built-in self-repair (ReBISR) scheme for multiple repairable RAM cores with different sizes and redundancy organizations (i.e., spare rows/spare columns or spare rows/spare IOs). We also propos...
展开
This paper presents a reconfigurable built-in self-repair (ReBISR) scheme for multiple repairable RAM cores with different sizes and redundancy organizations (i.e., spare rows/spare columns or spare rows/spare IOs). We also propose an efficient built-in redundancy-analysis (BIRA) algorithm for allocating redundancies for the ReBISR scheme. A reconfigurable BIRA (ReBIRA) circuit is realized to perform the proposed BIRA algorithm for the ReBISR scheme. Experimental results show that the ReBISR scheme can achieve high repair rate (i.e., the ratio of the number of repaired memories to the number of defective memories). The area cost of the reconfigurable BIRA is very small, e.g., the area cost is only about 1.5percent if 512X4X256 design parameters and four memory instances (64X2X32, 128X2X64, 256X4X128, and 512X4X256) are considered. Also, the ratio of the redundancy analysis time to the test time is very small, e.g., the ratio for a 512X4X256-bit memory tested by a March-14N algorithm with solid data backgrounds is only about 0.25percent.
收起
摘要 :
This paper presents a reconfigurable built-in self-repair (ReBISR) scheme for multiple repairable RAM cores with different sizes and redundancy organizations (i.e., spare rows/spare columns or spare rows/spare IOs). We also propos...
展开
This paper presents a reconfigurable built-in self-repair (ReBISR) scheme for multiple repairable RAM cores with different sizes and redundancy organizations (i.e., spare rows/spare columns or spare rows/spare IOs). We also propose an efficient built-in redundancy-analysis (BIRA) algorithm for allocating redundancies for the ReBISR scheme, A reconfigurable BIRA (ReBIRA) circuit is realized to perform the proposed BIRA algorithm for the ReBISR scheme. Experimental results show that the ReBISR scheme can achieve high repair rate (i.e., the ratio of the number of repaired memories to the number of defective memories). The area cost of the reconfigurable BIRA is very small, e.g., the area cost is only about 1.5% if 512×4×256 design parameters and four memory instances (64×2×32, 128×2×64, 256×4×128, and 512×4×256) are considered. Also, the ratio of the redundancy analysis time to the test time is very small, e.g., the ratio for a 512×4×256-bit memory tested by a March-I4N algorithm with solid data backgrounds is only about 0.25%.
收起
摘要 :
With the growing community of CubeSat developers has emerged a need for a technical approach that would provide a means of fast-tracking the lifecycle development process of CubeSats, including the need to minimize design and deve...
展开
With the growing community of CubeSat developers has emerged a need for a technical approach that would provide a means of fast-tracking the lifecycle development process of CubeSats, including the need to minimize design and development time repeating prior engineering efforts. In addition to a technical approach, there is also the need for a language and toolchain with a demand for both minimal training and minimal IT overhead to configure. A model-based approach has become a topic of interest in addressing much of these systems engineering painpoints. However, the current state-of-the-art tools demand either considerable investment in training and/or IT overhead, making it difficult for CubeSat developers of small startups or academia to participate. This article presents a model-based technical approach in conjunction with a docs-as-code approach, used to fascilitate flight software architecture development, intended to guide an implementation, for the SeaLion CubeSat mission - a joint CubeSat mission between the Old Dominion University and Coast Guard Academy. The technical approach discussed in this paper was also used to model and generate the paper as itself.
收起