Hasil untuk "General Works"

Erico Tjoa, Cuntai Guan

Recently, artificial intelligence and machine learning in general have demonstrated remarkable performances in many tasks, from image processing to natural language processing, especially with the advent of deep learning (DL). Along with research progress, they have encroached upon many different fields and disciplines. Some of them require high level of accountability and thus transparency, for example, the medical sector. Explanations for machine decisions and predictions are thus needed to justify their reliability. This requires greater interpretability, which often means we need to understand the mechanism underlying the algorithms. Unfortunately, the blackbox nature of the DL is still unresolved, and many machine decisions are still poorly understood. We provide a review on interpretabilities suggested by different research works and categorize them. The different categories show different dimensions in interpretability research, from approaches that provide “obviously” interpretable information to the studies of complex patterns. By applying the same categorization to interpretability in medical research, it is hoped that: 1) clinicians and practitioners can subsequently approach these methods with caution; 2) insight into interpretability will be born with more considerations for medical practices; and 3) initiatives to push forward data-based, mathematically grounded, and technically grounded medical education are encouraged.

Xin Wang, Yudong Chen, Wenwu Zhu

Curriculum learning (CL) is a training strategy that trains a machine learning model from easier data to harder data, which imitates the meaningful learning order in human curricula. As an easy-to-use plug-in, the CL strategy has demonstrated its power in improving the generalization capacity and convergence rate of various models in a wide range of scenarios such as computer vision and natural language processing etc. In this survey article, we comprehensively review CL from various aspects including motivations, definitions, theories, and applications. We discuss works on curriculum learning within a general CL framework, elaborating on how to design a manually predefined curriculum or an automatic curriculum. In particular, we summarize existing CL designs based on the general framework of Difficulty Measurer $+$+ Training Scheduler and further categorize the methodologies for automatic CL into four groups, i.e., Self-paced Learning, Transfer Teacher, RL Teacher, and Other Automatic CL. We also analyze principles to select different CL designs that may benefit practical applications. Finally, we present our insights on the relationships connecting CL and other machine learning concepts including transfer learning, meta-learning, continual learning and active learning, etc., then point out challenges in CL as well as potential future research directions deserving further investigations.

I. Sirés, E. Brillas, M. Oturan et al.

D. Alexander

Abstract. This paper examines the development over historical time of the meaning and uses of the term resilience. The objective is to deepen our understanding of how the term came to be adopted in disaster risk reduction and resolve some of the conflicts and controversies that have arisen when it has been used. The paper traces the development of resilience through the sciences, humanities, and legal and political spheres. It considers how mechanics passed the word to ecology and psychology, and how from there it was adopted by social research and sustainability science. As other authors have noted, as a concept, resilience involves some potentially serious conflicts or contradictions, for example between stability and dynamism, or between dynamic equilibrium (homeostasis) and evolution. Moreover, although the resilience concept works quite well within the confines of general systems theory, in situations in which a systems formulation inhibits rather than fosters explanation, a different interpretation of the term is warranted. This may be the case for disaster risk reduction, which involves transformation rather than preservation of the "state of the system". The article concludes that the modern conception of resilience derives benefit from a rich history of meanings and applications, but that it is dangerous – or at least potentially disappointing – to read to much into the term as a model and a paradigm.

Emil Björnson, J. Hoydis, M. Kountouris et al.

The use of large-scale antenna arrays can bring substantial improvements in energy and/or spectral efficiency to wireless systems due to the greatly improved spatial resolution and array gain. Recent works in the field of massive multiple-input multiple-output (MIMO) show that the user channels decorrelate when the number of antennas at the base stations (BSs) increases, thus strong signal gains are achievable with little interuser interference. Since these results rely on asymptotics, it is important to investigate whether the conventional system models are reasonable in this asymptotic regime. This paper considers a new system model that incorporates general transceiver hardware impairments at both the BSs (equipped with large antenna arrays) and the single-antenna user equipments (UEs). As opposed to the conventional case of ideal hardware, we show that hardware impairments create finite ceilings on the channel estimation accuracy and on the downlink/uplink capacity of each UE. Surprisingly, the capacity is mainly limited by the hardware at the UE, while the impact of impairments in the large-scale arrays vanishes asymptotically and interuser interference (in particular, pilot contamination) becomes negligible. Furthermore, we prove that the huge degrees of freedom offered by massive MIMO can be used to reduce the transmit power and/or to tolerate larger hardware impairments, which allows for the use of inexpensive and energy-efficient antenna elements.

Yossi Adi, Carsten Baum, Moustapha Cissé et al.

Deep Neural Networks have recently gained lots of success after enabling several breakthroughs in notoriously challenging problems. Training these networks is computationally expensive and requires vast amounts of training data. Selling such pre-trained models can, therefore, be a lucrative business model. Unfortunately, once the models are sold they can be easily copied and redistributed. To avoid this, a tracking mechanism to identify models as the intellectual property of a particular vendor is necessary. In this work, we present an approach for watermarking Deep Neural Networks in a black-box way. Our scheme works for general classification tasks and can easily be combined with current learning algorithms. We show experimentally that such a watermark has no noticeable impact on the primary task that the model is designed for. Moreover, we evaluate the robustness of our proposal against a multitude of practical attacks.

G. Berkolaiko, P. Kuchment

Zhongkui Li, G. Wen, Z. Duan et al.

This technical note addresses the distributed consensus protocol design problem for multi-agent systems with general linear dynamics and directed communication graphs. Existing works usually design consensus protocols using the smallest real part of the nonzero eigenvalues of the Laplacian matrix associated with the communication graph, which however is global information. In this technical note, based on only the agent dynamics and the relative states of neighboring agents, a distributed adaptive consensus protocol is designed to achieve leader-follower consensus in the presence of a leader with a zero input for any communication graph containing a directed spanning tree with the leader as the root node. The proposed adaptive protocol is independent of any global information of the communication graph and thereby is fully distributed. Extensions to the case with multiple leaders are further studied.

Jean-Yves Franceschi, Aymeric Dieuleveut, Martin Jaggi

Time series constitute a challenging data type for machine learning algorithms, due to their highly variable lengths and sparse labeling in practice. In this paper, we tackle this challenge by proposing an unsupervised method to learn universal embeddings of time series. Unlike previous works, it is scalable with respect to their length and we demonstrate the quality, transferability and practicability of the learned representations with thorough experiments and comparisons. To this end, we combine an encoder based on causal dilated convolutions with a novel triplet loss employing time-based negative sampling, obtaining general-purpose representations for variable length and multivariate time series.

H. Briegel, W. Dür, J. Cirac et al.

J. Markusen

J. Gauthier, H. Hammouri, S. Othman

Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ Ⅵ , B. Reviews, B. Worthen et al.

Mário A. T. Figueiredo, R. Nowak

Ruo Zhang, P. Tsai, J. Cryer et al.

K. Lake

Rani Mohanraj, Shuba Kumar, Sylvia Jayakumar et al.

Polished white rice (WR), high in refined carbohydrates, the main staple in South India is associated with enhanced risk of diabetes. Brown Rice (BR), with lower glycemic load, high fibre content and micronutrients, is a healthier choice. Two hundred and twelve Persons with Diabetes (PwD) attending a tertiary diabetes care centre in a city in South India responded to a questionnaire documenting types, frequency and reasons for rice consumption, awareness and beliefs about BR. A sub-set of 10, participated in qualitative interviews, which additionally, explored the influence of traditional beliefs on and consumption patterns of rice, barriers to BR consumption and willingness to accept it in their diet. Ninety-three percent reported consuming WR with traditional usage (97 %) being the main reason for its preference. Brand image, grain size, texture and taste, of rice were other decisional considerations. Awareness about health benefits of BR was limited, with 69 % and 51 % believing it to be nutritious and helping to reduce blood sugar respectively. Appearance, texture, taste and cost were deterrents to its use. Over half agreed to switch to BR if they believed it would improve their health. Participants with a shorter duration of diabetes were more willing to change to BR. The study highlights the need to promote greater literacy regarding health benefits of BR and other forms of less polished rice. Larger trials examining the effectiveness of BR viz-a viz other types of less polished rice on blood glucose levels, metabolic factors and nutritional content among PwD are needed.

Christian Truitt Lüddeke, Birger Hagemann, Leonhard Ganzer

The energy transition is a long-term strategy dedicated to achieving carbon neutrality in Germany by 2045. This process includes the development of renewably generated electricity from wind and solar power. As these energy sources are dependent on weather conditions, imbalances in availability and demand might result in a deficit or excess of renewable energy. In times of excessive availability, hydrogen can be generated and stored in subsurface storage sites from which it can be withdrawn when the demand increases. Using subsurface storage sites is a viable option as those currently store natural gas. Considering the properties of hydrogen, converting existing storage sites poses many challenges, including the total stored energy content and the amount of extractable working gas. Using an exemplary storage formation, a numerical simulation model is set up in an open-source software and used to calculate and compare the stored volumes and energy contents for natural gas and hydrogen. Storage withdrawal and pressure profiles for both cases are developed. A pseudo gas is defined with individually alternating compressibility, density and viscosity behavior to independently assess the influence of these properties on the withdrawal rates. A sensitivity analysis of various storage, bottom-hole flowing pressures and the skin factor on the extraction rates is also performed. The results show that a larger volume of natural gas and a resulting higher energy content can be stored in the storage site compared to hydrogen. Changes in extraction rates occur earlier and pressure decrease is greater for stored hydrogen. The compressibility factor has the largest influence on the extraction behavior of the gas, leading to a steeper decline in hydrogen withdrawal rates and a quicker pressure depletion. The levels of the storage and the bottom-hole flowing pressure impact the slope of the decline and the level of withdrawal rates.

Donghyun Suh, Samil Oh

Using a representative survey of Korean workers, we provide evidence on the adoption of Generative AI (GenAI) and how GenAI reallocates time at work. We find that 51.8\% of workers use GenAI for work and GenAI reduces working time by 3.8\%. However, these gains may not materialize in aggregate productivity statistics yet: the correlation between time savings and output changes is near zero. We show this disconnect arises because workers capture efficiency gains primarily as on-the-job leisure, rather than increasing their output. These findings suggest that standard productivity measures may understate AI's impact by missing non-pecuniary welfare channels.

Nursan Celik

<i>Dies Buch gehört dem König</i> (<i>This Book Belongs to the King</i>), written and published in 1843 by the German Romantic author Bettina von Arnim, is a quasi-open letter, presented as a series of fictional dialogues with traces of a novel. Dedicated to the newly crowned King of Prussia, Friedrich Wilhelm IV, the letter unfolds social grievances and aims to persuade Friedrich Wilhelm to act like a just king. Due to its delicate socio-critical impetus, the letter does so through strategies of obfuscation and by using a richly pictorial, seemingly naive and lavish way of speech rather than taking an openly reproachful stance. Crucially, von Arnim does not install herself as the letter’s speaker but instead fictionalizes the letter and presents Goethe’s mother, Catharina Elisabeth Goethe, as the letter’s primary voice (‘Frau Rat’). By using a well-respected figure of the ruling class as the letter’s main voice, von Arnim aimed at minimizing its scandalous potential. But even prior to publishing the letter, von Arnim had already managed to trick Friedrich Wilhelm and the Prussian censors herself: by fusing the book’s title and dedication, she paratextually outwitted both the censors and the King, whose permission she sought precisely to bypass Prussian censorship. This article shows how von Arnim managed to avoid a larger scandal both textually by implementing semi-fictional devices and paratextually by presenting the letter as an affirmation of Friedrich Wilhelm IV and his policies.