Bachelor's in Applied Computing
 
Welcome to the April 7, 2021 edition of ACM TechNews, providing timely information for IT professionals three times a week.

ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).

To view "Headlines At A Glance," hit the link labeled "Click here to view this online" found at the top of the page in the html version. The online version now has a button at the top labeled "Show Headlines."

The digital yuan, as seen on a mobile phone. China Creates Its Own Digital Currency
The Wall Street Journal
James T. Areddy
April 5, 2021


China's digital yuan cryptocurrency is expected to give its government a vast economic and social monitoring tool, and strip users of their anonymity. Beijing is preparing the digital currency for international use, and designing it to be unconnected to the global financial system, to permit more centralized control. The cryptocurrency is accessible from the owner's cellphone or on a card, and it may be spent without an online connection. Analysts and economists say the digital yuan could gain a foothold on the fringes of the international financial system, allowing people in impoverished nations to transfer money internationally. With a trackable digital currency, China's government could impose and collect fines as soon as an infraction is detected, or enable parties sanctioned by the U.S. to exchange money outside of sanctions.

Full Article
*May Require Paid Registration
Why the Supreme Court's Ruling for Google Over Oracle Is a Win for Innovation
Fortune
Aaron Pressman
April 5, 2021


On Monday, the U.S. Supreme Court ended a decade-long legal battle in ruling that Google did not violate Oracle's copyrights associated with the Java programming language. The ruling largely maintains the use of application programming interfaces (APIs), which enable one company's hardware or software to interact with those from another. Microsoft, IBM's Red Hat, and Mozilla were among the technology companies that filed briefs contending new software development could be hampered if Oracle's demands were upheld. The Center for Democracy and Technology's Stan Adams said, "This decision is a huge win for developers and consumers. When software is interoperable—meaning it can talk to other software programs—it is easier to innovate and build new services."

Full Article

Using wide-field images and deep learning, researchers developed an analysis system for suspicious pigmented skin lesions. AI Tool Can Help Detect Melanoma
MIT News
Megan Lewis
April 2, 2021


Researchers at the Massachusetts Institute of Technology (MIT) have designed an artificial intelligence system that analyzes wide-field images of patients' skin in order to detect melanoma more efficiently. The process applies deep convolutional neural networks (DCNNs) to optimize the identification and classification of suspicious pigmented lesions (SPLs) in wide-field images. The MIT researchers trained the system on 20,388 wide-field images from 133 patients at Spain's Hospital Gregorio Marañón, and on publicly available images. Dermatologists visually classified lesions in the images for comparison, and the system achieved more than 90.3% sensitivity in differentiating SPLs from nonsuspicious lesions, skin, and complex backgrounds.

Full Article
Old Programming Language Suddenly Getting More Popular Again
ZDNet
Liam Tung
April 6, 2021


The latest edition of the Tiobe Programming Community index saw Objective-C fall off the list of the 20 most popular programming languages, while Fortran has risen from 34th place to 20th in the past year. Tiobe hypothesized that Objective-C maintained its popularity partly because the adoption of Swift decelerated as mobile application developers focused on languages that could be used for building apps on multiple platforms. Fortran, released by IBM in the 1950s, remains a popular language in scientific computing circles. Tiobe said, "Fortran was the first commercial programming language ever, and is gaining popularity thanks to the massive need for [scientific] number crunching."

Full Article
ML Approach Speeds Up Search for Molecular Conformers
Aalto University (Finland)
April 1, 2021


Researchers at Finland's Aalto University developed a molecular conformer search procedure that integrates an active learning Bayesian optimization algorithm with quantum chemistry techniques to accelerate the process. Searching for molecular conformers previously required the relaxation of thousands of structures, entailing a significant commitment of time and computational resources even when applied to small molecules. The Aalto team's algorithm samples the structures with low energies or high energy uncertainties, to minimize the required data points. The researchers tested the machine learning procedure on four amino acids, and found low-energy conformers in good correspondence with experimental measurements and reference calculations while using less than 10% of the computational cost of the current fastest method.

Full Article

A deep neural network taught to speak the answer demonstrates robust and efficient features of learning. Deep Learning Networks Prefer the Human Voice—Just Like Us
Columbia Engineering
Holly Evarts
April 6, 2021


Columbia University's Hod Lipson and Boyuan Chen demonstrated that artificial intelligence systems programmed with sound files of human language can outperform those coded with numerical data labels. The engineers created two neural networks and trained them to recognize 10 different types of objects in a set of 50,000 photos. One system was trained with binary inputs, while the other was fed a data table containing photos of animal or objects with corresponding audio files of a human voice speaking the names of those animals or objects. The Columbia researchers found that when presented with an image, the binary-programmed network answered with 1s and 0s, while the other network vocalized the name of the imaged object. When tested with ambiguous images, the voice-trained network was found to be 50% accurate, while the numerically trained network was only 20% accurate.

Full Article
Computing Algorithms Expand the Boundaries of Quantum Future
HPCwire
April 5, 2021


Prasanth Shyamsundar at the U.S. Department of Energy's Fermilab Quantum Institute announced two algorithms that build upon existing research to expand the types of problems quantum computers can solve. One algorithm can help search for a specific entry in an unsorted dataset (a record collection) by querying all the data at once via superposition. He said this non-Boolean quantum amplitude amplification algorithm is "open to more tasks; there are a lot of problems that can be solved more naturally in terms of a score rather than a yes-or-no output." Shyamsundar's second algorithm, a quantum mean estimation algorithm, lets researchers estimate the average rating of all the records in a stack. Both algorithms allow for a range of outputs, to characterize information more accurately in combination with a quantum speedup over classical computing methods.

Full Article
Scientists Design First System to Help Film Scriptwriters Produce Storylines with Best Chance of Box-Office Success
Universidad de Granada (Spain)
April 5, 2021


Scientists at Spain's universities of Granada (UGR) and Cádiz analyzed tropes of movie storylines and designed the first computer system to help screenwriters compose storylines with optimal chances of scoring with audiences. The team analyzed over 25,000 tropes associated with 10,766 films, and devised a methodology to understand how tropes operate, visually represent their interrelationships and roles in different genres, and infer which combinations might be most successful. The researchers used UGR's TropeScraper software to scour the TVTropes database for tropes used in films, then mapped each film's user rating and popularity based on the IMDb website. Network analysis rated tropes' popularity, their status as transversal (general or basic) across all films or specific/specialized, and whether they were gaining or losing relevance. UGC's Pablo García-Sánchez said, "Depending on the combination and design of the actions based the tropes, we can now broadly ascertain the level of interest each kind of storyline is likely to generate."

Full Article
Computational Tool for Materials Physics Growing in Popularity
California Institute of Technology
Emily Velasco
April 1, 2021


The Perturbo software developed by California Institute of Technology (Caltech) researchers is increasingly popular among scientists for its ability to perform quantum calculations of electron dynamics. Perturbo can simulate how electrons moving through a material interact with the material's constituent atoms, and how electrical properties and light-excitation responses are shaped by those interactions. Caltech's Marco Bernardi said, "One could investigate the microscopic physics of a large number of compounds with this method and use that information to engineer better materials." Bernardi said Perturbo was crafted to run on modern supercomputers, and a recent paper demonstrated that the software can operate efficiently on a computer with thousands of processing cores.

Full Article

A moving base will reduce the need for printed supports, cutting waste and saving time. How a Moving Platform for 3D Printing Can Cut Waste, Costs
USC Viterbi School of Engineering
Greta Harrison
April 1, 2021


Researchers at the University of Southern California Viterbi School of Engineering (USC Viterbi) designed a low-cost movable surface for three-dimensional (3D) printers that reduces waste and accelerates production. The prototype platform has a programmable, dynamically controlled surface composed of movable metal pins that replace printed supports. Each individual support operates from a single motor that moves the platform. The pins elevate as the printer progressively constructs the product. USC Viterbi's Yong Chen said in tests, the device saved roughly 35% in materials usage, and was about 40% faster in printing than standard Fused Deposition Modeling 3D printers. Chen added that the system could be modified easily for large-scale manufacturing.

Full Article

The algorithm has been organized into layers to make its own intelligent decisions. Screening for Skin Disease on Your Laptop
University of Houston Cullen College of Engineering
Laurie Fickman
April 6, 2021


University of Houston (UH) researchers have developed a deep neural network architecture that facilitates early diagnosis of systemic sclerosis (SSc) by immediately differentiating between images of healthy and diseased skin. The network was trained using the parameters of the MobileNetV2 mobile vision application, pretrained on the 1.4-million-image ImageNet dataset. The UH team added layers to the UNet, a modified convolutional neural network (CNN) architecture, then devised a mobile training module. Results indicated the proposed architecture outperformed CNNs for SSc image classification. UH's Yasmin Akay said, "After fine-tuning, our results showed the proposed network reached 100% accuracy on the training image set, 96.8% accuracy on the validation image set, and 95.2% on the testing image set."

Full Article
Topological Data Analysis Can Help Predict Crashes
EPFL (Switzerland)
Anne-Muriel Brouet
April 6, 2021


A model for predicting major system shifts was developed by researchers at the Swiss Federal Institute of Technology, Lausanne, the School of Engineering and Management Vaud, and startup L2F. The open source giotto-tda model can help analysts identify when stock-market crashes, earthquakes, or other major systematic events are about to transpire, based on topological data analysis (TDA). The model's foundational principle is that as a system reaches a critical state, its representative data points start forming shapes that alter its overall structure. Scientists can closely track these point clouds to identify the system's normal state and when a sudden change is imminent, while TDA's resilience to noise prevents irrelevant data from distorting the signals.

Full Article
AI Method for Generating Proteins Will Speed Up Drug Development
Chalmers University of Technology (Sweden)
March 30, 2021


Researchers at Sweden's Chalmers University of Technology have developed artificial intelligence (AI) that can synthesize novel, functionally active proteins. Chalmers' Aleksej Zelezniak said the method can proceed from design to working protein in just a few weeks, much more quickly than current protein-engineering techniques. The ProteinGAN approach involves feeding the AI a large dataset of well-studied proteins, which it analyzes and attempts to generate new proteins; concurrently, another part of the AI tries to determine if the synthetic proteins are natural or not. Said Chalmers' Martin Engqvist, "Accelerating the rate at which we engineer proteins is very important for driving down development costs for enzyme catalysts. This is the key for realizing environmentally sustainable industrial processes and consumer products, and our AI model, as well as future models, will enable that."

Full Article
Master of Science in Biomedical Informatics
 
ACM Conferences
 

Association for Computing Machinery

1601 Broadway, 10th Floor
New York, NY 10019-7434
1-800-342-6626
(U.S./Canada)



ACM Media Sales

If you are interested in advertising in ACM TechNews or other ACM publications, please contact ACM Media Sales or (212) 626-0686, or visit ACM Media for more information.

To submit feedback about ACM TechNews, contact: [email protected]