检索结果-内蒙古大学图书馆

Learning to Write with Coherence From Negative Examples

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Son, Seonil Lim, Jaeseo Jang, Youwon Lee, Jaeyoung Zhang, Byoung-Tak Language AI Lab NCSOFT Gyeonggi-Do Korea Republic of Interdisciplinary Program in Cognitive Science Korea Republic of School of Computer Science and Engineering Seoul National University Seoul Korea Republic of Artificial Intelligence Institute Seoul National University Seoul Korea Republic of

Coherence is one of the critical factors that determine the quality of writing. We propose writing relevance (WR) training method for neural encoder-decoder natural language generation (NLG) models which improves coherence of the continuation by leveraging negative examples. WR loss regresses the vector representation of the context and generated sentence toward positive continuation by contrasting it with the negatives. We compare our approach with Unlikelihood (UL) training in a text continuation task on commonsense natural language inference (NLI) corpora to show which method better models the coherence by avoiding unlikely continuations. The preference of our approach in human evaluation shows the efficacy of our method in improving coherence. © 2022, CC BY.

关键词： Natural language processing systems

A model for building and sustaining communities of engineering education research scholars

学校读者我要写书评

暂无评论

A model for building and sustaining communities of engineeri...

113th Annual ASEE Conference and Exposition, 2006

作者： Adams, Robin Allendoerfer, Cheryl Bell, Philip Chen, Helen Fleming, Lorraine Leifer, Larry Maring, Bayta Williams, Dawn Purdue University United States Department of Engineering Education Purdue University United States University of Washington United States Center for the Advancement of Engineering Education University of Washington United States Cognitive Studies in Education University of Washington United States Stanford University United States United States Center for the Advancement Engineering Education Academic Pathways Study Research Team Howard University United States Department of Civil Engineering Howard University United States National Science Foundation HBCU Undergraduate Program United States Department of Mechanical Engineering Design Stanford Center for Design Research United States United States United States Center for the Advancement of Engineering Education Department of Educational Administration and Policy Howard University United States Institute for Scholarship on Engineering Education

Building and sustaining communities of new engineering education researchers is crucial for bringing people into an area of scholarship, supporting them working in this area, providing opportunities for long term professional development within a community of practice, and building capacity in the engineering education research infrastructure. The Institute for Scholarship on engineering Education (ISEE), an element of the NSF-funded Center for the Advancement of engineering Education (CAEE), is designed to build and sustain communities of engineering education scholars who can investigate student learning issues and transform findings into actionable improvements. The Institute has a revolving host structure across three schools - University of Washington, Stanford University, and Howard University. Over three consecutive year-long cycles, ISEE builds and extends a cohort of scholars at host and partner schools. In this paper we explore the question "how do we build capacity in engineering education research?" We pursue this question by examining the ISEE model, the successes and challenges experienced as we near completion of the second of three cycles, and implications for future community building efforts. © American Society for engineering Education, 2006.

关键词： engineering research

A low-power CMOS neural amplifier with amplitude measurements for spike sorting

学校读者我要写书评

暂无评论

A low-power CMOS neural amplifier with amplitude measurement...

IEEE International Symposium on Circuits and Systems (ISCAS)

作者： T. Horiuchi T. Swindell D. Sander P. Abshier Institute for Systems Research Neuroscience and Cognitive Science Program University of Maryland College Park MD USA Electrical and Computer Engineering Department University of Maryland College Park MD USA Electrical and Computer Engineering Department Electrical Engineering Department Pennsylvania State University State College PA USA

Integrated, low-power, low-noise CMOS neural amplifiers have recently grown in importance as large microelectrode arrays have begun to be practical. With an eye to a future where thousands of signals must be transmitted over a limited bandwidth link or be processed in situ, we are developing low-power neural amplifiers with integrated pre-filtering and measurements of the spike signal to facilitate spike-sorting and data reduction prior to transmission to a data-acquisition system. We have fabricated a prototype circuit in a commercially-available 1.5 /spl mu/m, 2-metal, 2-poly CMOS process that occupies approximately 91,000 square /spl mu/m. We report circuit characteristics for a 1.5 V power supply, suitable for single cell battery operation. In one specific configuration, the circuit bandpass filters the incoming signal from 22 Hz to 6.7 kHz while providing a gain of 42.5 dB. With an amplifier power consumption of 0.8 /spl mu/W, the rms input-referred noise is 20.6 /spl mu/V.

关键词： Sorting Signal processing Low-noise amplifiers Microelectrodes Bandwidth Integrated circuit measurements Prototypes CMOS process Power supplies Batteries

Improving Candle Direction Classification in Forex Market Using Support Vector Machine with Hyperparameters Tuning

学校读者我要写书评

暂无评论

Improving Candle Direction Classification in Forex Market Us...

International Conference on Informatics and Computing (ICIC)

作者： Raymond Sunardi Oetama Yaya Heryadi Lukas Wayan Suparta Doctor of Computer Science Bina Nusantara University Jakarta Indonesia Computer Science Department Binus Graduate Program Doctor of Computer Science Bina Nusantara University Jakarta Indonesia Cognitive Engineering Research Group Universitas Katolik Indonesia Atma Jaya Jakarta Indonesia Department of Electrical Engineering Faculty of Industrial Technology Institut Teknologi Nasional Yogyakarta Indonesia

ISBN: (纸本)9798350345728

Foreign Exchange market is the world's largest daily currency turnover. Two of the popular currencies Euro and Pound sterling traded against the US Dollar. Since the Russia and Ukraine war started in February 2022, their exchange rates decrease to the lowest rate ever. Even though the general trend is bearish, several daily candles increase for some days making challenges for forex analysts. To solve this problem, classification is applied. The data is labeled downward and upward. By utilizing Linear Kernel and Radial Basis Function (RBF) Kernel-based Support Vector Machines (SVM), the candle direction can be classified and optimized by tuning the Hyperparameters. The accuracy of candle direction classifications are highly improved. After tuning, in general, classification using Linear Models can outperform RBF Models. The best accuracy found on the Pound sterling against US Dollar by using the Linear model is 98.11% and the accuracy becomes 100% on data testing at a ratio of 70:30. Whilst for the Euro against the US Dollar, the best accuracy found the same for both Linear and RBF models on a ratio of 80:20 at 97.53%. However, on data testing, it decreases to 94.51% for Linear Model and 93.41% using RBF Model. The implication of this study is SVM can successfully classify candle direction on pairs in the Forex Market that are affected by a big event that comes for such a long period as long as the hyperparameter is tuned.

关键词： Support vector machines Training Exchange rates Market research Data models Kernel Informatics

Paradigm shift in electron-based crystallography via machine learning

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Kaufmann, Kevin Zhu, Chaoyi Rosengarten, Alexander S. Maryanovsky, Daniel Harrington, Tyler J. Marin, Eduardo Vecchio, Kenneth S. Department of NanoEngineering UC San Diego La JollaCA92093 United States Materials Science and Engineering Program UC San Diego La JollaCA92093 United States Department of Cognitive Science UC San Diego La JollaCA92093 United States

Accurately determining the crystallographic structure of a material, organic or inorganic, is a critical primary step in material development and analysis. The most common practices involve analysis of diffraction patterns produced in laboratory X-ray diffractometers, transmission electron microscopes, and synchrotron X-ray sources. However, these techniques are slow, require careful sample preparation, can be difficult to access, and are prone to human error during analysis. This paper presents a newly developed methodology that represents a paradigm change in electron diffraction-based structure analysis techniques, with the potential to revolutionize multiple crystallography-related fields. A machine learning-based approach for rapid and autonomous identification of the crystal structure of metals and alloys, ceramics, and geological specimens, without any prior knowledge of the sample, is presented and demonstrated utilizing the electron backscatter diffraction technique. Electron backscatter diffraction patterns are collected from materials with well-known crystal structures, then a deep neural network model is constructed for classification to a specific Bravais lattice or point group. The applicability of this approach is evaluated on diffraction patterns from samples unknown to the computer without any human input or data filtering. This is in comparison to traditional Hough transform electron backscatter diffraction, which requires that you have already determined the phases present in your sample. The internal operations of the neural network are elucidated through visualizing the symmetry features learned by the convolutional neural network. It is determined that the model looks for the same features a crystallographer would use, even though it is not explicitly programmed to do so. This study opens the door to fully automated, high-throughput determination of crystal structures via several electron-based diffraction techniques. Copyright © 2019, The Auth

关键词： Crystal structure

Decoder remapping to counteract neuron loss in brain-machine interfaces

学校读者我要写书评

暂无评论

Decoder remapping to counteract neuron loss in brain-machine...

Annual International Conference of the IEEE engineering in Medicine and Biology Society (EMBC)

作者： Rodolphe Heliot Subramaniam Venkatraman Jose M. Carmena CEA LETI MINATEC France Department of Electrical Engineering and Computer Sciences University of California Berkeley CA USA QUALCOMM Inc.orporated USA Department of Electrical Engineering and Computer Sciences Helen Wills Neuroscience Institute and Program in Cognitive Science University of California Berkeley CA USA

Variability of single-unit neural recordings can significantly affect the overall performance achieved by brain machine interfaces (BMI). In this paper, we present a novel technique to adapt a linear filter commonly used in BMI to compensate for loss of neurons from the recorded neural ensemble, thus minimizing loss in performance. We simulate the gains achieved by this technique using a model of the learning process during closed-loop BMI operation. This simulation suggests that we can adapt to the loss of 24% of the neurons controlling a BMI with only 13% drop in performance.

关键词： Neurons Decoding Firing Wiener filter Brain modeling Maximum likelihood detection Nonlinear filters

Hierarchical policy learning is sensitive to goal space design

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Dwiel, Zach Candadai, Madhavun Phielipp, Mariano Bansal, Arjun K. Intel AI Lab BloomingtonIN United States Program in Cognitive Science School of Informatics Computing and Engineering BloomingtonIN United States Intel AI Lab PhoenixAZ United States Intel AI Lab San DiegoCA United States

Hierarchy in reinforcement learning agents allows for control at multiple time scales yielding improved sample efficiency, the ability to deal with long time horizons and transferability of sub-policies to tasks outside the training distribution. It is often implemented as a master policy providing goals to a sub-policy. Ideally, we would like the goal-spaces to be learned, however, properties of optimal goal spaces still remain unknown and consequently there is no method yet to learn optimal goal spaces. Motivated by this, we systematically analyze how various modifications to the ground-truth goal-space affect learning in hierarchical models with the aim of identifying important properties of optimal goal spaces. Our results show that, while rotation of ground-truth goal spaces and noise had no effect, having additional unnecessary factors significantly impaired learning in hierarchical models. Copyright © 2019, The Authors. All rights reserved.

关键词： Hierarchical systems

Topic modeling the Hàn diăn ancient classics

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Allen, Colin Luo, Hongliang Murdock, Jaimie Pu, Jianghuai Wang, Xiaohong Zhai, Yanjie Zhao, Kun Department of Philosophy School of Humanities and Social Sciences Xi'an Jiaotong University Shaanxi China Cognitive Science Program Indiana University BloomingtonIN United States Institute of Computer Software and Theory School of Electronic and Information Engineering Xi'an Jiaotong University Shaanxi China

Ancient Chinese texts present an area of enormous challenge and opportunity for humanities scholars interested in exploiting computational methods to assist in the development of new insights and interpretations of culturally significant materials. In this paper we describe a collaborative effort between Indiana University and Xi'an Jiaotong University to support exploration and interpretation of a digital corpus of over 18,000 ancient Chinese documents, which we refer to as the "Handian" ancient classics corpus (汉典古籍 or Hàn diăn gŭ jí, i.e, the "Han canon" or "Chinese classics"). It contains classics of ancient Chinese philosophy, documents of historical and biographical significance, and literary works. We begin by describing the Digital Humanities context of this joint project, and the advances in humanities computing that made this project feasible. We describe the corpus and introduce our application of probabilistic topic modeling to this corpus, with attention to the particular challenges posed by modeling ancient Chinese documents. We give a specific example of how the software we have developed can be used to aid discovery and interpretation of themes in the corpus. We outline more advanced forms of computer-aided interpretation that are also made possible by the programming interface provided by our system, and the general implications of these methods for understanding the nature of meaning in these texts. Copyright © 2017, The Authors. All rights reserved.

关键词： Distributed computer systems

Modular boundaries in recurrent neural networks

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Tanner, Jacob C. Mansour, Sina L. Coletta, Ludovico Gozzi, Alessandro Betzel, Richard F. Cognitive Science Program 2 School of Informatics Computing and Engineering Indiana University BloomingtonIN47405 United States National University of Singapore Singapore Fondazione Bruno Kessler Trento Italy Functional Neuroimaging Lab Istituto Italiano di Tecnologia Center for Neuroscience and Cognitive Systems Rovereto Italy Department of Psychological and Brain Sciences Indiana University BloomingtonIN47405 United States

Recent theoretical and experimental work in neuroscience has focused on the representational and dynamical character of neural manifolds –subspaces in neural activity space wherein many neurons coactivate. Importantly, neural populations studied under this "neural manifold hypothesis" are continuous and not cleanly divided into separate neural populations. This perspective clashes with the "modular hypothesis" of brain organization, wherein neural elements maintain an "all- or-nothing" affiliation with modules. In line with this modular hypothesis, recent research on recurrent neural networks suggests that multi-task networks become modular across training, such that different modules specialize for task-general dynamical motifs [1, 2]. If the modular hypothesis is true, then it would be important to use a dimensionality reduction technique that captures modular structure. Here, we investigate the features of such a method. We leverage RNNs as a model system to study the character of modular neural populations, using a community detection method from network science known as modularity maximization to partition neurons into distinct modules. These partitions allow us to ask the following question: do these modular boundaries matter to the system? We find evidence that they do. First, we find that these boundaries neatly divide the representational content and role of neurons. Next, we find that these boundaries can be directly inferred from features of the weight matrix in feed-forward neural networks, and are related to clustering of the Jacobian matrix in recurrent neural networks. We also find that the weights of input neurons to recurrent neurons partially inform their modular structure, an observation that we corroborated using structural and functional imaging data from mice and humans. Finally, we find that the dynamics of these RNNs reflected the boundaries between modules. Collectively, our results suggest that neural populations in RNNs sometimes form modu

关键词： Recurrent neural networks