Computers Math

Anime, the Japanese art of animation, comprises hand-drawn sketches in an abstract form with unique characteristics and exaggerations of real-life subjects. While generative artificial intelligence (AI) has found use in the content creation such as anime portraits, its use to augment human creativity, and guide freehand drawings proves challenging. The primary challenge lies with the generation of suitable reference images corresponding with the incomplete and abstract strokes made during the freehand drawing process. This is particularly true when the strokes created during the drawing process are incomplete and offer insufficient information for generative AI to predict the final shape of the drawing.
To tackle this problem, a research team from Japan Advanced Institute of Science and Technology (JAIST) and Waseda University in Japan, sought to develop a novel generative AI tool that offers progressive drawing assistance and helps generate anime portraits from freehand sketches. The tool is based on a sketch-to-image (S2I) deep learning framework that matches raw sketches with latent vectors of the generative model. It employs a two-stage training strategy through the pre-trained Style Generative Adversarial Network (StyleGAN) — a state-of-the-art generative model that uses adversarial networks to generate new images.
The team, led by Dr. Zhengyu Huang from JAIST, including Associate Professor Haoran Xie and Professor Kazunori Miyata, and Lecturer Tsukasa Fukusato from Waseda University proposed a novel “stroke-level disentanglement,” a strategy that associates input strokes of a freehand sketch with edge-related attributes, in the latent structural code of StyleGAN. This approach allows users to manipulate the attribute parameters, thereby having greater autonomy over the properties of generated images. Dr. Huang says, “We introduced an unsupervised training strategy for stroke-level disentanglement in StyleGAN, which enables the automatic matching of rough sketches with sparse strokes to the corresponding local parts in anime portraits, all without the need for semantic labels.”
This study will be presented at ACM SIGGRAPH 2023, the premier conference for computer graphics and interactive techniques and the only CORE ranking A* conference in the research fields worldwide.
Regarding the development of the tool, Prof. Xie adds, “We first trained an image encoder using a pre-trained StyleGAN model as a teacher encoder. In the second stage, we simulated the drawing process of generated images without additional data to train the sketch encoder for incomplete progressive sketches. This helped us generate high-quality portrait images that align with the disentangled representations of teacher encoder.”
To further highlight the effectiveness and usability of AniFaceDrawing in aiding users with anime portrait creation, the team conducted a user study. They invited 15 graduate students to draw digital freehand anime-style portraits using the AniFaceDrawing tool, with the option to switch between rough and detailed guidance modes for line art. While the former provided prompts for specific facial parts, the latter provided prompts for the full-face portrait based on the user’s drawing progress. Participants could pin the generated guidance once it matched their expectations, and further refine their input sketch. This tool also allowed participants to select a reference image to generate a color portrait of their input sketch. Next, they evaluated the tool for user satisfaction and guidance matching through a survey.
The team noted that the system consistently provided high-quality facial guidance and effectively supported the creation of anime-style portraits, by not only enhancing user sketches, but also by generating desirable corresponding colored images. Prof. Fukusato remarks, “Our system could successfully transform the user’s rough sketches into high-quality anime portraits. The user study indicated that even novices could make reasonable sketches with the help of the system and end up with high-quality color art drawings.”
“Our generative AI framework enables users, regardless of their skill level and experience, to create professional anime portraits even from incomplete drawings. Our approach consistently produces high-quality image generation results throughout the creation process, regardless of the drawing order or how poor the initial sketches are,” summarizes Prof. Miyata.
In the long run, these findings can help democratize AI technology and assist users with creative tasks, thereby augmenting their creative capacity without technological barriers. More

Engineers are constantly searching for materials with novel, desirable property combinations. For example, an ultra-strong, lightweight material could be used to make airplanes and cars more fuel-efficient, or a material that is porous and biomechanically friendly could be useful for bone implants.
Cellular metamaterials — artificial structures composed of units, or cells, that repeat in various patterns — can help achieve these goals. But it is difficult to know which cellular structure will lead to the desired properties. Even if one focuses on structures made of smaller building blocks like interconnected beams or thin plates, there are an infinite number of possible arrangements to consider. So, engineers can manually explore only a small fraction of all the cellular metamaterials that are hypothetically possible.
Researchers from MIT and the Institute of Science and Technology Austria have developed a computational technique that makes it easier for a user to quickly design a metamaterial cell from any of those smaller building blocks, and then evaluate the resulting metamaterial’s properties.
Their approach, like a specialized CAD (computer-aided design) system for metamaterials, allows an engineer to quickly model even very complex metamaterials and experiment with designs that may have otherwise taken days to develop. The user-friendly interface also enables the user to explore the entire space of potential metamaterial shapes, since all building blocks are at their disposal.
“We came up with a representation that can cover all of the different shapes engineers have traditionally shown interest in. Because you can build them all the same way, that means you can switch between them more fluidly,” says MIT electrical engineering and computer science graduate student Liane Makatura, co-lead author of a paper on this technique.
Makatura wrote the paper with co-lead author Bohan Wang, an MIT postdoc; Yi-Lu Chen, a graduate student at the Institute of Science and Technology Austria (ISTA); Bolei Deng, an MIT postdoc; Chris Wojtan and Bernd Bickel, professors at ISTA; and senior author Wojciech Matusik, a professor of electrical engineering and computer science at MIT who leads the Computational Design and Fabrication Group within the MIT Computer Science and Artificial Intelligence Laboratory. The research will be presented at SIGGRAPH.
A unified method
When a scientist develops a cellular metamaterial, she typically begins by choosing a representation that will be used to describe her potential designs. This choice determines the set of shapes that will be available for exploration. More

A collaboration of nuclear theorists at the U.S. Department of Energy’s (DOE) Brookhaven National Laboratory, Argonne National Laboratory, Temple University, Adam Mickiewicz University of Poland, and the University of Bonn, Germany, has used supercomputers to predict the spatial distributions of charges, momentum, and other properties of “up” and “down” quarks within protons. The results, just published in Physical Review D, revealed key differences in the characteristics of the up and down quarks.
“This work is the first to leverage a new theoretical approach to obtain a high-resolution map of quarks within a proton,” said Swagato Mukherjee of Brookhaven Lab’s nuclear theory group and a coauthor on the paper. “Our calculations show that the up quark is more symmetrically distributed and spread over a smaller distance than the down quark. These differences imply that up and down quarks may make different contributions to the fundamental properties and structure of the proton, including its internal energy and spin.”
Coauthor Martha Constantinou of Temple University noted, “Our calculations provide input for interpreting data from nuclear physics experiments exploring how quarks and the gluons that hold them together are distributed within the proton, giving rise to the proton’s overall properties.”
Such experiments are already taking place at the Continuous Electron Beam Accelerator Facility (CEBAF), a DOE Office of Science user facility at Thomas Jefferson National Accelerator Facility. Higher resolution versions are planned for the future Electron-Ion Collider (EIC) at Brookhaven Lab. In these experiments, high-energy electrons emit virtual particles of light that scatter off and change the overall momentum of a proton without breaking it apart. The way the momentum of the proton changes in response to these scatterings reveals details about the quarks and gluons — the inner components of the proton — sort of like an x-ray imaging technique for the building blocks of bulk matter.
New theoretical approach to GPD
Specifically, the scatterings give scientists access to the Generalized Parton Distribution (GPD) of the proton — parton being the collective name for quarks and gluons. If you picture the proton as a bag filled with marbles representing quarks and gluons, the GPD provides a description of how the energy-momentum and other characteristics of these marbles are distributed within the bag — for example, when the bag is shaken and the marbles move around. It can be compared to a map that indicates the likelihood of finding a marble with a specific energy-momentum at a particular position inside the bag. Knowing the distribution of these quark and gluon characteristics allows scientists to understand the inner workings of the proton, which may lead to new ways to apply that knowledge.
“To obtain a detailed map, we need to analyze many scattering interactions, involving various values of momentum change of the proton,” said Shohini Bhattacharya, a research associate in Brookhaven’s nuclear theory group and the RIKEN BNL Research Center (RBRC). More

A common metal paper clip will stick to a magnet. Scientists classify such iron-containing materials as ferromagnets. A little over a century ago, physicists Albert Einstein and Wander de Haas reported a surprising effect with a ferromagnet. If you suspend an iron cylinder from a wire and expose it to a magnetic field, it will start rotating if you simply reverse the direction of the magnetic field.
“Einstein and de Haas’s experiment is almost like a magic show,” said Haidan Wen, a physicist in the Materials Science and X-ray Science divisions of the U.S. Department of Energy’s (DOE) Argonne National Laboratory. ​”You can cause a cylinder to rotate without ever touching it.”
In Nature magazine, a team of researchers from Argonne and other U.S. national laboratories and universities now report an analogous yet different effect in an ​”anti”-ferromagnet. This could have important applications in devices requiring ultra-precise and ultrafast motion control. One example is high-speed nanomotors for biomedical applications, such as use in nanorobots for minimally invasive diagnosis and surgery.
The difference between a ferromagnet and antiferromagnet has to do with a property called electron spin. This spin has a direction. Scientists represent the direction with an arrow, which can point up or down or any direction in between. In the magnetized ferromagnet mentioned above, the arrows associated with all the electrons in the iron atoms can point in the same direction, say, up. Reversing the magnetic field reverses the direction of the electron spins. So, all arrows are pointing down. This reversal leads to the cylinder’s rotation.
“In this experiment, a microscopic property, electron spin, is exploited to elicit a mechanical response in a cylinder, a macroscopic object,” said Alfred Zong, a Miller Research Fellow at the University of California, Berkeley.
In antiferromagnets, instead of the electron spins all pointing up, for example, they alternate from up to down between adjacent electrons. These opposite spins cancel each other out, and antiferromagnets thus do not respond to changes in a magnetic field as ferromagnets do.
“The question we asked ourselves is, can electron spin elicit a response in an antiferromagnet that is different but similar in spirit to that from the cylinder rotation in the Einstein-de Hass experiment?” Wen said. More

A new statistical tool can help researchers get meaningful results when a randomized experiment, considered the gold standard, is not possible.
Randomized experiments split participants into groups by chance, with one undergoing an intervention and the other not. But in real-world situations, they can’t always be done. Companies might not want to use the method, or such experiments might be against the law.
Developed by a researcher at The University of Texas at Austin, the new tool called two-step synthetic control adapts an existing research workaround, known as the synthetic control method.
The traditional synthetic control method creates synthetic control groups from the data, in place of real ones. The groups are weighted statistically and compared with a group undergoing an intervention.
But the synthetic control method does not perfectly apply to all situations, especially ones in which the intervention group is different from control groups, according to Kathleen Li, an assistant professor of marketing at the McCombs School of Business. In these scenarios, the method’s lack of flexibility could lead to less accurate results.
“Our framework allows managers and policymakers to estimate effects they previously weren’t able to estimate accurately,” said Li, who developed the tool along with Venkatesh Shankar of Texas A&M University. “They get a more precise estimate that can help them make more informed decisions.”
The study, published in advance online in the journal Management Science, offers a two-step synthetic control approach: First, it determines whether the traditional synthetic control method applies to a given case. If it does not, the second step uses a more flexible framework that allows weighted controls to differ from 100% or to shift the control group up and down.The researchers tested the new method on a real-world situation by looking at sales of tampons: how they responded in 2016, when New York repealed a sales tax on them. More

An international research team from the Max Planck Institute of Microstructure Physics, Halle (Saale), Germany, the University of Cambridge, UK and the University of Pennsylvania, USA reported the first realization of single-crystalline T-Nb2O5 thin films having two-dimensional (2D) vertical ionic transport channels, which results in a fast and colossal insulator-metal transition via Li ion intercalation through the 2D channels.
Since the 1940s, scientists have been exploring the use of niobium oxide, specifically a form of niobium oxide known as T-Nb2O5, to create more efficient batteries. This unique material is known for its ability to allow lithium ions, the tiny charged particles that make batteries work, to move quickly within it. The faster these lithium ions can move, the faster a battery can be charged.
The challenge, however, has always been to grow this niobium oxide material into thin, flat layers, or ‘films’ that are of high enough quality to be used in practical applications. This problem stems from the complex structure of T-Nb2O5 and the existence of many similar forms, or polymorphs, of niobium oxide.
Now, in a paper published in Nature Materials, researchers from the Max Planck Institute of Microstructure Physics, University of Cambridge and the University of Pennsylvania have successfully demonstrated the growth of high-quality, single-crystal thin films of T-Nb2O5, aligned in such a way that the lithium ions can move even faster along vertical ionic transport channels.
The T-Nb2O5 films undergo a significant electrical change at an early stage of Li insertion into the initially insulating films. This is a dramatic shift — the resistivity of the material decreases by a factor of 100 billion. The research team further demonstrate tunable and low voltage operation of thin film devices by altering the chemical composition of the ‘gate’ electrode, a component that controls the flow of ions in a device, further extending the potential applications.
The Max Planck Institute of Microstructure Physics group realized the growth of the single-crystalline T-Nb2O5 thin films, and showed how Li-ion intercalation can dramatically increase their electrical conductivity. Together with the University of Cambridge group multiple previously unknown transitions in the material’s structure were discovered as the concentration of lithium ions was changed. These transitions change the electronic properties of the material, allowing it to switch from being an insulator to a metal, meaning that it goes from blocking electric current to conducting it. Researchers from the University of Pennsylvania rationalized the multiple phase transitions they observed, as well as, how these phases might be related to the concentration of lithium ions and their arrangement within the crystal structure.
These results could only have been successful through synergies between the three international groups with diverse specialties: thin films from the Max Planck Institute of Microstructure Physics, batteries from the University of Cambridge, and theory from the University of Pennsylvania. More

Mammography screening supported by artificial intelligence (AI) is a safe alternative to today’s conventional double reading by radiologists and can reduce heavy workloads for doctors. This has now been shown in an interim analysis of a prospective, randomised controlled trial, which addressed the clinical safety of using AI in mammography screening. The trial, led by researchers from Lund University in Sweden, has been published in The Lancet Oncology.
Each year around one million women in Sweden are called to mammography screening. Each screening examination is reviewed by two breast radiologists to ensure a high sensitivity, so called double reading. There is however a workforce shortage of breast radiologists, in Sweden and elsewhere, which can put the screening service at risk. Lately, the potential of AI to support mammography screening has attracted much attention, but how this is to be optimally conducted and what the clinical consequences will be, remains unclear.
To know with certainty what happens when radiologists work with the support of AI requires studies in which women are randomly allocated to AI-supported screening or to standard screening. The Mammography Screening with Artificial Intelligence (MASAI) trial is the first randomised controlled trial evaluating the effect of AI-supported screening.
“In our trial, we used AI to identify screening examinations with a high risk of breast cancer, which underwent double reading by radiologists. The remaining examinations were classified as low risk and were read only by one radiologist. In the screen reading, radiologists used AI as detection support, in which it highlighted suspicious findings on the images,” says Kristina Lång, researcher and associate professor in diagnostic radiology at Lund University and consultant at Skåne University Hospital, who led the study.
The 80,033 women included in the safety analysis were randomly allocated into two groups: 40,003 women in the intervention group that underwent AI-supported screening and 40,030 in the control group that underwent standard double reading without AI support.
“We found that using AI resulted in the detection of 20 % (41) more cancers compared with standard screening, without affecting false positives. A false positive in screening occurs when a woman is recalled but cleared of suspicion of cancer after workup,” says Kristina Lång.
At the same time, the screen-reading workload for radiologists was reduced by 44 %. The number of screen readings with AI-supported screening was 46,345 compared with 83,231 with standard screening. More

A proposed machine learning framework and expanded use of blockchain technology could help counter the spread of fake news by allowing content creators to focus on areas where the misinformation is likely to do the most public harm, according to new research from Binghamton University, State University of New York.
The research led by Thi Tran, assistant professor of management information systems at Binghamton University’s School of Management, expands on existing studies by offering tools for recognizing patterns in misinformation and helping content creators zero in the worst offenders.
“I hope this research helps us educate more people about being aware of the patterns,” Tran said, “so they know when to verify something before sharing it and are more alert to mismatches between the headline and the content itself, which would keep the misinformation from spreading unintentionally.”
Tran’s research proposed machine learning systems — a branch of artificial intelligence (AI) and computer science that uses data and algorithms to imitate the way humans learn while gradually improving its accuracy — to help determine the scale to which content could cause the most harm to its audience.
Examples could include stories that circulated during the height of the COVID-19 pandemic touting false alternate treatments to the vaccine.
The framework would use data and algorithms to spot indicators of misinformation and use those examples to inform and improve the detection process. It would also consider user characteristics from people with prior experience or knowledge about fake news to help piece together a harm index. The index would reflect the severity of possible harm to a person in certain contexts if they were exposed and victimized by the misinformation.
“We’re most likely to care about fake news if it causes a harm that impacts readers or audiences. If people perceive there’s no harm, they’re more likely to share the misinformation,” Tran said. “The harms come from whether audiences act according to claims from the misinformation, or if they refuse the proper action because of it. If we have a systematic way of identifying where misinformation will do the most harm, that will help us know where to focus on mitigation.”
Based on the information gathered, Tran said, the machine learning system could help fake news mitigators discern which messages are likely to be the most damaging if allowed to spread unchallenged. More