Showing papers in "Communications of The ACM in 2017"
TL;DR: A large, deep convolutional neural network was trained to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes and employed a recently developed regularization method called "dropout" that proved to be very effective.
Abstract: We trained a large, deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes. On the test data, we achieved top-1 and top-5 error rates of 37.5% and 17.0%, respectively, which is considerably better than the previous state-of-the-art. The neural network, which has 60 million parameters and 650,000 neurons, consists of five convolutional layers, some of which are followed by max-pooling layers, and three fully connected layers with a final 1000-way softmax. To make training faster, we used non-saturating neurons and a very efficient GPU implementation of the convolution operation. To reduce overfitting in the fully connected layers we employed a recently developed regularization method called "dropout" that proved to be very effective. We also entered a variant of this model in the ILSVRC-2012 competition and achieved a winning top-5 test error rate of 15.3%, compared to 26.2% achieved by the second-best entry.
33,301 citations
TL;DR: When bottleneck buffers are large, loss-based congestion control keeps them full, causing bufferbloat, leading to low throughput, which requires an alternative to loss- based congestion control.
Abstract: When bottleneck buffers are large, loss-based congestion control keeps them full, causing bufferbloat When bottleneck buffers are small, loss-based congestion control misinterprets loss as a signa
850 citations
TL;DR: Addressing unresolved questions concerning computational thinking with a focus on the role of reinforcement learning.
Abstract: Addressing unresolved questions concerning computational thinking.
271 citations
TL;DR: While it may not be possible to build a data brain identical to a human, data science can still aspire to imaginative machine thinking.
Abstract: While it may not be possible to build a data brain identical to a human, data science can still aspire to imaginative machine thinking.
262 citations
TL;DR: In a decentralized marketplace, buyers and sellers transact directly, without manipulation by intermediary platforms.
Abstract: In a decentralized marketplace, buyers and sellers transact directly, without manipulation by intermediary platforms.
201 citations
[...]
TL;DR: Model learning emerges as an effective method for black-box state machine models of hardware and software components that combine simulation and physical measurements to solve the challenge of integrating smart phones to manage complex black- box systems.
Abstract: Model learning emerges as an effective method for black-box state machine models of hardware and software components.
200 citations
University of Sydney1, North Carolina State University2, University of South Carolina3, Macquarie University4, RMIT University5, Rochester Institute of Technology6, University of New South Wales7, University of Michigan8, Qatar Computing Research Institute9, University of Trento10, Southeast University11, Swinburne University of Technology12, Sun Yat-sen University13, Commonwealth Scientific and Industrial Research Organisation14, Texas A&M University15, Qatar University16, University of Miami17, Vienna University of Technology18, University of Stuttgart19, Tilburg University20
TL;DR: Mapping out the challenges and strategies for the widespread adoption of service computing shows clear trends in adoption and a clear road map for the future direction is proposed.
Abstract: Mapping out the challenges and strategies for the widespread adoption of service computing.
194 citations
TL;DR: Both practitioners and academics would do well to revisit old ideas to glean insights for present systems, as Bitcoin was unusual and successful not because it was on the cutting edge of research on any of its components, but because it combined old ideas from many previously unrelated fields.
Abstract: We’ve seen repeatedly that ideas in the research literature can be gradually forgotten or lie unappreciated, especially if they are ahead of their time, even in popular areas of research. Both practitioners and academics would do well to revisit old ideas to glean insights for present systems. Bitcoin was unusual and successful not because it was on the cutting edge of research on any of its components, but because it combined old ideas from many previously unrelated fields. This is not easy to do, as it requires bridging disparate terminology, assumptions, etc., but it is a valuable blueprint for innovation.
191 citations
TL;DR: Microsecond-scale I/O means tension between performance and productivity that will need new latency-mitigating ideas, including in hardware, according to researchers at the Massachusetts Institute of Technology.
Abstract: Microsecond-scale I/O means tension between performance and productivity that will need new latency-mitigating ideas, including in hardware.
190 citations
[...]
TL;DR: Healthcare robotics can provide health and wellness support to billions of people and help improve the quality of life in the developing world.
Abstract: Robots have the potential to be a game changer in healthcare: improving health and well-being, filling care gaps, supporting care givers, and aiding health care workers. However, before robots are able to be widely deployed, it is crucial that both the research and industrial communities work together to establish a strong evidence-base for healthcare robotics, and surmount likely adoption barriers. This article presents a broad contextualization of robots in healthcare by identifying key stakeholders, care settings, and tasks; reviewing recent advances in healthcare robotics; and outlining major challenges and opportunities to their adoption.
179 citations
TL;DR: New blocks frameworks open doors to greater experimentation for novices and professionals alike as discussed by the authors, which can be used for both novice and professional developers, and can be found in this paper.
Abstract: New blocks frameworks open doors to greater experimentation for novices and professionals alike
TL;DR: This work systematically test popular open-source TLS implementations for state machine bugs and discovers several critical security vulnerabilities that have lain hidden in these libraries for years, and have now finally been patched due to the disclosures.
Abstract: The Transport Layer Security (TLS) protocol supports various authentication modes, key exchange methods, and protocol extensions. Confusingly, each combination may prescribe a different message sequence between the client and the server, and thus a key challenge for TLS implementations is to define a composite state machine that correctly handles these combinations. If the state machine is too restrictive, the implementation may fail to interoperate with others; if it is too liberal, it may allow unexpected message sequences that break the security of the protocol. We systematically test popular TLS implementations and find unexpected transitions in many of their state machines that have stayed hidden for years. We show how some of these flaws lead to critical security vulnerabilities, such as FREAK. While testing can help find such bugs, formal verification can prevent them entirely. To this end, we implement and formally verify a new composite state machine for OpenSSL, a popular TLS library.
TL;DR: This framework for developing pre-service teachers' knowledge does not necessarily depend on computers or other educational technology.
Abstract: This framework for developing pre-service teachers' knowledge does not necessarily depend on computers or other educational technology.
TL;DR: Numerical evidence indicates that the heat method converges to the exact distance in the limit of refinement; the method can be applied in any dimension, and on any domain that admits a gradient and inner product---including regular grids, triangle meshes, and point clouds.
Abstract: We introduce the heat method for solving the single- or multiple-source shortest path problem on both flat and curved domains. A key insight is that distance computation can be split into two stages: first find the direction along which distance is increasing, then compute the distance itself. The heat method is robust, efficient, and simple to implement since it is based on solving a pair of standard sparse linear systems. These systems can be factored once and subsequently solved in near-linear time, substantially reducing amortized cost. Real-world performance is an order of magnitude faster than state-of-the-art methods, while maintaining a comparable level of accuracy. The method can be applied in any dimension, and on any domain that admits a gradient and inner product---including regular grids, triangle meshes, and point clouds. Numerical evidence indicates that the method converges to the exact distance in the limit of refinement; we also explore smoothed approximations of distance suitable for applications where greater regularity is desired.
TL;DR: The IoT can become ubiquitous worldwide---if the pursuit of systemic trustworthiness can overcome the potential risks.
Abstract: The IoT can become ubiquitous worldwide---if the pursuit of systemic trustworthiness can overcome the potential risks.
TL;DR: A new programming language for image processing pipelines, called Halide, that separates the algorithm from its schedule, and is expressive enough to describe organizations that match or outperform state-of-the-art hand-written implementations of many computational photography and computer vision algorithms.
Abstract: Writing high-performance code on modern machines requires not just locally optimizing inner loops, but globally reorganizing computations to exploit parallelism and locality---doing things such as tiling and blocking whole pipelines to fit in cache. This is especially true for image processing pipelines, where individual stages do much too little work to amortize the cost of loading and storing results to and from off-chip memory. As a result, the performance difference between a naive implementation of a pipeline and one globally optimized for parallelism and locality is often an order of magnitude. However, using existing programming tools, writing high-performance image processing code requires sacrificing simplicity, portability, and modularity. We argue that this is because traditional programming models conflate the computations defining the algorithm with decisions about intermediate storage and the order of computation, which we call the schedule.We propose a new programming language for image processing pipelines, called Halide, that separates the algorithm from its schedule. Programmers can change the schedule to express many possible organizations of a single algorithm. The Halide compiler then synthesizes a globally combined loop nest for an entire algorithm, given a schedule. Halide models a space of schedules which is expressive enough to describe organizations that match or outperform state-of-the-art hand-written implementations of many computational photography and computer vision algorithms. Its model is simple enough to do so often in only a few lines of code, and small changes generate efficient implementations for x86, ARM, Graphics Processors (GPUs), and specialized image processors, all from a single algorithm.Halide has been public and open source for over four years, during which it has been used by hundreds of programmers to deploy code to tens of thousands of servers and hundreds of millions of phones, processing billions of images every day.
TL;DR: New tools tackle an age-old practice as discussed by the authors, which is to tackle an old-fashioned practice, such as "self-congratulation" and self-criticism.
Abstract: New tools tackle an age-old practice.
TL;DR: Even when checked by fact checkers, facts are often still open to preexisting bias and doubt.
Abstract: Even when checked by fact checkers, facts are often still open to preexisting bias and doubt.
TL;DR: RRI requires doing the best science for the world, not only the bestScience in the world.
Abstract: At a time when increasingly potent technologies are being developed with the potential to transform society, researchers in all technological fields, including information and communications technology (ICT), are under growing pressure to consider and reflect on the motivations, purposes, and possible consequences associated with their research. This pressure comes from the general public, civil society, and government institutions. In parallel is a growing recognition that current ethics review procedures within ICT may not address broader concerns (such as the potential societal consequences of innovation).
TL;DR: A new biometric based on the human body's response to an electric square pulse signal, called pulse-response, is proposed, which integrates well with other established methods and offers a reliable additional layer of security, either on a continuous basis or at login time.
Abstract: We propose a new biometric based on the human body's response to an electric square pulse signal, called pulse-response. We explore how this biometric can be used to enhance security in the context of two example applications: (1) an additional authentication mechanism in PIN entry systems, and (2) a means of continuous authentication on a secure terminal. The pulse-response biometric is effective because each human body exhibits a unique response to a signal pulse applied at the palm of one hand, and measured at the palm of the other. Using a prototype setup, we show that users can be correctly identified, with high probability, in a matter of seconds. This identification mechanism integrates well with other established methods and offers a reliable additional layer of security, either on a continuous basis or at login time. We build a proof-of-concept prototype and perform experiments to assess the feasibility of pulse-response as a practical biometric. The results are very encouraging, achieving accuracies of 100% over a static data set, and 88% over a data set with samples taken over several weeks.
TL;DR: Development of energy-efficient software is hindered by a lack of knowledge and a lacks of tools, according to a report by the International Energy Agency.
Abstract: Development of energy-efficient software is hindered by a lack of knowledge and a lack of tools.
TL;DR: Mathematics solves problems by pen and paper, but CS helps us to go far beyond that.
Abstract: Mathematics solves problems by pen and paper. CS helps us to go far beyond that.
TL;DR: Integrating trust and automation in finance is a major challenge for the future of finance and will require new ideas and strategies to be developed.
Abstract: A major consequence of the Internet era is the emergence of complex “platforms” that combine technology and process in new ways that often disrupt existing industry structures and blur industry boundaries. These platforms allow easy participation that often strengthens and extends network effects, while the vast amounts of data captured through such participation can increase the value of the platform to its participants, creating a virtuous cycle. While initially slow to penetrate the financial services sector, such platforms are now beginning to emerge. We provide a taxonomy of platforms in Finance and identify the feasible strategies that are available to incumbents in the industry, innovators, and the major Internet giants.
TL;DR: Seeking more common ground between data scientists and their critics is part of a series of articles exploring the relationship between data science and its critics.
Abstract: Seeking more common ground between data scientists and their critics.
TL;DR: Incidents from the early days of AI research are instructive in the current AI environment.
Abstract: Incidents from the early days of AI research are instructive in the current AI environment.
TL;DR: DeepDive is described, a system that combines database and machine learning ideas to help to develop KBC systems, to frame traditional extract-transform-load (ETL) style data management problems as a single large statistical inference task that is declaratively defined by the user.
Abstract: The dark data extraction or knowledge base construction (KBC) problem is to populate a relational database with information from unstructured data sources, such as emails, webpages, and PDFs. KBC is a long-standing problem in industry and research that encompasses problems of data extraction, cleaning, and integration. We describe DeepDive, a system that combines database and machine learning ideas to help to develop KBC systems. The key idea in DeepDive is to frame traditional extract-transform-load (ETL) style data management problems as a single large statistical inference task that is declaratively defined by the user. DeepDive leverages the effectiveness and efficiency of statistical inference and machine learning for difficult extraction tasks, whereas not requiring users to directly write any probabilistic inference algorithms. Instead, domain experts interact with DeepDive by defining features or rules about the domain. DeepDive has been successfully applied to domains such as pharmacogenomics, paleobiology, and antihuman trafficking enforcement, achieving human-caliber quality at machine-caliber scale. We present the applications, abstractions, and techniques used in DeepDive to accelerate the construction of such dark data extraction systems.
TL;DR: Answering questions correctly from standardized eighth-grade science tests is itself a test of machine intelligence as mentioned in this paper, and answering questions correctly is itself an evaluation of machine learning skills in general.
Abstract: Answering questions correctly from standardized eighth-grade science tests is itself a test of machine intelligence.
TL;DR: In risk assessment and predictive policing, biased data can yield biased results.
Abstract: In risk assessment and predictive policing, biased data can yield biased results.
TL;DR: Recent attacks exploiting a known vulnerability continue a downward spiral of ransomware-related incidents.
Abstract: Recent attacks exploiting a known vulnerability continue a downward spiral of ransomware-related incidents.