Computers Math

The Abbey Library of St. Gall in Switzerland is home to approximately 160,000 volumes of literary and historical manuscripts dating back to the eighth century — all of which are written by hand, on parchment, in languages rarely spoken in modern times.
To preserve these historical accounts of humanity, such texts, numbering in the millions, have been kept safely stored away in libraries and monasteries all over the world. A significant portion of these collections are available to the general public through digital imagery, but experts say there is an extraordinary amount of material that has never been read — a treasure trove of insight into the world’s history hidden within.
Now, researchers at University of Notre Dame are developing an artificial neural network to read complex ancient handwriting based on human perception to improve capabilities of deep learning transcription.
“We’re dealing with historical documents written in styles that have long fallen out of fashion, going back many centuries, and in languages like Latin, which are rarely ever used anymore,” said Walter Scheirer, the Dennis O. Doughty Collegiate Associate Professor in the Department of Computer Science and Engineering at Notre Dame. “You can get beautiful photos of these materials, but what we’ve set out to do is automate transcription in a way that mimics the perception of the page through the eyes of the expert reader and provides a quick, searchable reading of the text.”
In research published in the Institute of Electrical and Electronics Engineers journal Transactions on Pattern Analysis and Machine Intelligence, Scheirer outlines how his team combined traditional methods of machine learning with visual psychophysics — a method of measuring the connections between physical stimuli and mental phenomena, such as the amount of time it takes for an expert reader to recognize a specific character, gauge the quality of the handwriting or identify the use of certain abbreviations.
Scheirer’s team studied digitized Latin manuscripts that were written by scribes in the Cloister of St. Gall in the ninth century. Readers entered their manual transcriptions into a specially designed software interface. The team then measured reaction times during transcription for an understanding of which words, characters and passages were easy or difficult. Scheirer explained that including that kind of data created a network more consistent with human behavior, reduced errors and provided a more accurate, more realistic reading of the text.
“It’s a strategy not typically used in machine learning,” Scheirer said. “We’re labeling the data through these psychophysical measurements, which comes directly from psychological studies of perception — by taking behavioral measurements. We then inform the network of common difficulties in the perception of these characters and can make corrections based on those measurements.”
Using deep learning to transcribe ancient texts is something of great interest to scholars in the humanities.
“There’s a difference between just taking the photos and reading them, and having a program to provide a searchable reading,” said Hildegund Müller, associate professor in the Department of Classics at Notre Dame. “If you consider the texts used in this study — ninth-century manuscripts — that’s an early stage of the Middle Ages. It’s a long time before the printing press. That’s a time when an enormous amount of manuscripts was produced. There is all sorts of information hidden in these manuscripts — unidentified texts that nobody has seen before.”
Scheirer said challenges remain. His team is working on improving accuracy of transcriptions, especially in the case of damaged or incomplete documents, as well as how to account for illustrations or other aspects of a page that could be confusing to the network.
However, the team was able to adjust the program to transcribe Ethiopian texts, adapting it to a language with a completely different set of characters — a first step toward developing a program with the capability to transcribe and translate information for users.
“In the literary field, it could be really helpful. Every good literary work is surrounded by a vast amount of historical documents, but where it’s really going to be useful is in historical archival research,” said Müller. “There is a great need to advance the digital humanities. When you talk about the Middle Ages and early modern times, if you want to understand the details and consequences of historical events, you have to look through the written material, and these texts are the only thing we have. The problem may be even greater outside the Western world. Think of languages that are disappearing in cultures that are under threat. We must first of all preserve these works, make them accessible and, at some point, incorporate translations to make them a part of cultural processes that are still underway — and we are racing against time.”
Story Source:
Materials provided by University of Notre Dame. Original written by Jessica Sieff. Note: Content may be edited for style and length. More

Brains have evolved to do more with less. Take a tiny insect brain, which has less than a million neurons but shows a diversity of behaviors and is more energy-efficient than current AI systems. These tiny brains serve as models for computing systems that are becoming more sophisticated as billions of silicon neurons can be implemented on hardware.
The secret to achieving energy-efficiency lies in the silicon neurons’ ability to learn to communicate and form networks, as shown by new research from the lab of Shantanu Chakrabartty, the Clifford W. Murphy Professor in the Preston M. Green Department of Electrical & Systems Engineering at Washington University in St. Louis’ McKelvey School of Engineering.
Their results were published July 28, 2021 in the journal Frontiers in Neuroscience.
For several years, his research group studied dynamical systems approaches to address the neuron-to-network performance gap and provide a blueprint for AI systems as energy efficient as biological ones.
Previous work from his group showed that in a computational system, spiking neurons create perturbations which allow each neuron to “know” which others are spiking and which are responding. It’s as if the neurons were all embedded in a rubber sheet formed by energy constraints; a single ripple, caused by a spike, would create a wave that affects them all. Like all physical processes, systems of silicon neurons tend to self-optimize to their least-energetic states, while also being affected by the other neurons in the network. These constraints come together to form a kind of secondary communication network, where additional information can be communicated through the dynamic but synchronized topology of spikes. It’s like the rubber sheet vibrating in a synchronized rhythm in response to multiple spikes.
In the latest research result, Chakrabartty and doctoral student Ahana Gangopadhyay showed how the neurons learn to pick the most energy-efficient perturbations and wave patterns in the rubber sheet. They show that if the learning is guided by sparsity (less energy), it’s like the electrical stiffness of the rubber sheet is adjusted by each neuron so that the entire network vibrates in a most energy-efficient way. The neuron does this using only local information which is communicated more efficiently. Communications between the neurons then become an emergent phenomenon guided by the need to optimize energy use. More

In a paper published in Nature Quantum Information, EPFL professor Giuseppe Carleo and Matija Medvidovi?, a graduate student at Columbia University and at the Flatiron Institute in New York, have found a way to execute a complex quantum computing algorithm on traditional computers instead of quantum ones.
The specific “quantum software” they are considering is known as Quantum Approximate Optimization Algorithm (QAOA) and is used to solve classical optimization problems in mathematics; it’s essentially a way of picking the best solution to a problem out of a set of possible solutions. “There is a lot of interest in understanding what problems can be solved efficiently by a quantum computer, and QAOA is one of the more prominent candidates,” says Carleo.
Ultimately, QAOA is meant to help us on the way to the famed “quantum speedup,” the predicted boost in processing speed that we can achieve with quantum computers instead of conventional ones. Understandably, QAOA has a number of proponents, including Google, who have their sights set on quantum technologies and computing in the near future: in 2019 they created Sycamore, a 53-qubit quantum processor, and used it to run a task it estimated it would take a state-of-the-art classical supercomputer around 10,000 years to complete. Sycamore ran the same task in 200 seconds.
“But the barrier of “quantum speedup” is all but rigid and it is being continuously reshaped by new research, also thanks to the progress in the development of more efficient classical algorithms,” says Carleo.
In their study, Carleo and Medvidovi? address a key open question in the field: can algorithms running on current and near-term quantum computers offer a significant advantage over classical algorithms for tasks of practical interest? “If we are to answer that question, we first need to understand the limits of classical computing in simulating quantum systems,” says Carleo. This is especially important since the current generation of quantum processors operate in a regime where they make errors when running quantum “software,” and can therefore only run algorithms of limited complexity.
Using conventional computers, the two researchers developed a method that can approximately simulate the behavior of a special class of algorithms known as variational quantum algorithms, which are ways of working out the lowest energy state, or “ground state” of a quantum system. QAOA is one important example of such family of quantum algorithms, that researchers believe are among the most promising candidates for “quantum advantage” in near-term quantum computers.
The approach is based on the idea that modern machine-learning tools, e.g. the ones used in learning complex games like Go, can also be used to learn and emulate the inner workings of a quantum computer. The key tool for these simulations are Neural Network Quantum States, an artificial neural network that Carleo developed in 2016 with Matthias Troyer, and that was now used for the first time to simulate QAOA. The results are considered the province of quantum computing, and set a new benchmark for the future development of quantum hardware.
“Our work shows that the QAOA you can run on current and near-term quantum computers can be simulated, with good accuracy, on a classical computer too,” says Carleo. “However, this does not mean that alluseful quantum algorithms that can be run on near-term quantum processors can be emulated classically. In fact, we hope that our approach will serve as a guide to devise new quantum algorithms that are both useful and hard to simulate for classical computers.”
Story Source:
Materials provided by Ecole Polytechnique Fédérale de Lausanne. Original written by Nik Papageorgiou. Note: Content may be edited for style and length. More

Quantum information could be behind the next technological revolution. By analogy with the bit in classical computing, the qubit is the basic element of quantum computing. However, demonstrating the existence of this information storage unit and using it remains complex, and hence limited.
In a study published on 3 August 2021 in Physical Review X, an international research team consisting of CNRS researcher Fabio Pistolesi1 and two foreign researchers used theoretical calculations to show that it is possible to realize a new type of qubit, in which information is stored in the oscillation amplitude of a carbon nanotube. These nanotubes can perform a large number of oscillations without diminishing, which shows their low level of interaction with the environment, and makes them excellent potential qubits. This property would enable for greater reliability in quantum computation.
A problem nevertheless persists with regard to the reading and writing of information stored in the first two energy levels2 of these oscillators. Scientists successfully proved that this information could be read by using the coupling between electrons, a negatively charged particle, and the flexural mode of these nanotubes.
This changes the spacing between the first levels of energy enough to make them accessible independently from other levels, thereby making it possible to read the information they contain. These promising theoretical predictions have not yet been verified experimentally.
Notes
1 — Researcher at the Laboratoire ondes et matières d’Aquitaine (CNRS/Université de Bordeaux). He worked with a scientist from the University of Chicago (United States) and from the Institute of Photonic Sciences in Barcelona (Spain).
2 — The level of energy is a quantity that can describe systems in physics, with a level corresponding to a “state” of the system.
Story Source:
Materials provided by CNRS. Note: Content may be edited for style and length. More

Researchers from the Neuroimaging Center at NYU Abu Dhabi (NYUAD) and Wisconsin Institute for the Discovery at University Wisconsin-Madison have developed the UW Virtual Brain Project™, producing unique, interactive, 3D narrated diagrams to help students learn about the structure and function of perceptual systems in the human brain. A new study exploring how students responded to these lessons on desktop PCs and in virtual reality (VR) offers new insights into the benefits of VR as an educational tool.
Led by Associate Professor and Director of NYUAD’s Neuroimaging Center Bas Rokers and Assistant Professor of Psychology and a Principal Investigator in the Virtual Environments Group at the Wisconsin Institute for Discovery at University of Wisconsin-Madison Karen Schloss, the researchers have published the findings of their work in a new paper, “The UW Virtual Brain Project: An immersive approach to teaching functional neuroanatomy”  in the journal Translational Issues in Psychological Science from the American Psychological Association (APA). In their experiments, the researchers found that participants showed significant content-based learning for both devices, with no significant differences between PC and VR devices for content-based learning outcomes. However, VR far exceeded PC viewing for achieving experience-based learning outcomes — VR was, in other words, more enjoyable and easier to use.
“Students are enthusiastic about learning in VR,” said Rokers. “However, our findings indicate that learners can have similar access to learning about functional neuroanatomy through multiple platforms, which means that those who don’t have access to VR technology are not at an inherent disadvantage. The power of VR is its ability to transport learners to new environments they might not otherwise be able to explore. But, importantly, VR is not a substitute for real-world interactions with peers and instructors.”
The 3D narrated videos are already in active use at classes that include neuro-anatomy instruction both at the University of Wisconsin-Madison and at NYUAD.
Story Source:
Materials provided by New York University. Note: Content may be edited for style and length. More

Researchers at Tohoku University and the Swiss Federal Institute of Technology in Lausanne, with the support of the Human Frontier Science Program, have decoded the flexible motor control mechanisms underlying salamander walking.
Their findings were published in the journal Frontiers in Neurorobotics on July 30, 2021.
Animals with four feet can navigate complex, unpredictable, and unstructured environments. The impressive ability is thanks to their body-limb coordination.
The salamander is an excellent specimen for studying body-limb coordination mechanisms. It is an amphibian that uses four legs and walks by swaying itself from left to right in a motion known as undulation.
Their nervous system is simpler than those of mammals, and they change their walking pattern according to the speed at which they are moving.
To decode the salamander’s movement, researchers led by Professor Akio Ishiguro of the Research Institute of Electrical Communication at Tohoku University modeled the salamander’s nervous system mathematically and physically simulated the model.
In making the model, the researchers hypothesized that the legs and the body are controlled to support other motions by sharing sensory information. They then reproduced the speed-dependent gait transitions of salamanders through computer simulations.
“We hope this finding provides insights into the essential mechanism behind the adaptive and versatile locomotion of animals,” said Ishiguro.
The researchers are confident their discovery will aid the development of robots that can move with high agility and adaptability by flexibly changing body-limb coordination patterns.
Story Source:
Materials provided by Tohoku University. Note: Content may be edited for style and length. More

Internet-based cognitive behavioral therapy (CBT) for depression is often just as effective as traditional CBT. This is clear from an international study involving scientists at the University of Gothenburg. However, some online treatments have components that can be harmful.
Internet CBT (iCBT) as a method of delivering treatment is on the increase. Nevertheless, it has been unclear to date which parts of the treatment are most helpful against depression, which are less efficacious and which are potentially detrimental to patients.
In an international study, researchers at the University of Gothenburg participated in a systematic literature review and meta-analysis. The study was based on 76 randomized controlled trials (RCTs) in Sweden and elsewhere. In total, the RCTs included 17,521 patients, 71% of whom were women.
The study, under the aegis of Kyoto University in Japan, is now published in The Lancet Psychiatry. One coauthor is Cecilia Björkelund, Senior Professor of Family Medicine at the University of Gothenburg’s Sahlgrenska Academy.
“In mild or moderate depression, the effect of iCBT is as good as that of conventional CBT. For many, it’s a superb way of getting access to therapy without having to go to a therapist. We also saw that it was especially good for the elderly — a finding we didn’t entirely expect,” she says.
Just as in traditional CBT, its online counterpart involves modifying patients’ thoughts, feelings and behaviors that are obstacles in their lives and impair their mood. During the treatment, which often lasts about ten weeks, they are given tasks and exercises to perform on their own.
The factor that proved most significant for the prognosis was the depth of depression at the start of treatment. In milder depression, better results were obtained. Therapist support and text-message reminders increased the proportion of patients who completed the therapy.
“If you’re going to use iCBT in health care, the programs have to be regulated just as well as drugs are, but that’s not the case today. With this study, we’re taking a real step forward. First, the study surveys what’s most effective. Second, it provides knowledge of how to design a program and adapt its composition to patients’ problems,” Björkelund says.
However, iCBT requires continuous therapeutic contact. One reason is the importance of the therapist being able to see an improvement within three to four weeks, ensuring that the trend is not in the wrong direction. Björkelund stresses the great potential danger of depression. In severe depression, internet-mediated therapy is inappropriate.
The study shows the danger of using iCBT with programs that include relaxation therapy. Rather than being beneficial, this may have negative effects, exacerbating depressive symptoms and causing “relaxation-induced anxiety.”
“For a depressed person, it isn’t advisable. Relaxation programs shouldn’t be used as part of depression treatment in health care,” Björkelund says.
Story Source:
Materials provided by University of Gothenburg. Note: Content may be edited for style and length. More

In the U.S., the place where one was born, one’s social and economic background, the neighborhoods in which one spends one’s formative years, and where one grows old are factors that account for a quarter to 60% of deaths in any given year, partly because these forces play a significant role in occurrence and outcomes for heart disease, cancer, unintentional injuries, chronic lower respiratory diseases, and cerebrovascular diseases — the five leading causes of death.
While data on such “macro” factors is critical to tracking and predicting health outcomes for individuals and communities, analysts who apply machine-learning tools to health outcomes tend to rely on “micro” data constrained to purely clinical settings and driven by healthcare data and processes inside the hospital, leaving factors that could shed light on healthcare disparities in the dark.
Researchers at the NYU Tandon School of Engineering and NYU School of Global Public Health (NYU GPH), in a new perspective, “Machine learning and algorithmic fairness in public and population health,” in Nature Machine Intelligence, aim to activate the machine learning community to account for “macro” factors and their impact on health. Thinking outside the clinical “box” and beyond the strict limits of individual factors, Rumi Chunara, associate professor of computer science and engineering at NYU Tandon and of biostatistics at the NYU GPH, found a new approach to incorporating the larger web of relevant data for predictive modeling for individual and community health outcomes.
“Research of what causes and reduces equity shows that to avoid creating more disparities it is essential to consider upstream factors as well,” explained Chunara. She noted, on the one hand, the large body of work on AI and machine learning implementation in healthcare in areas like image analysis, radiography, and pathology, and on the other the strong awareness and advocacy focused on such areas as structural racism, police brutality, and healthcare disparities that came to light around the COVID-19 pandemic.
“Our goal is to take that work and the explosion of data-rich machine learning in healthcare, and create a holistic view beyond the clinical setting, incorporating data about communities and the environment.”
Chunara, along with her doctoral students Vishwali Mhasawade and Yuan Zhao, at NYU Tandon and NYU GPH, respectively, leveraged the Social Ecological Model, a framework for understanding how the health, habits and behavior of an individual are affected by factors such as public policies at the national and international level and availability of health resources within a community and neighborhood. The team shows how principles of this model can be used in algorithm development to show how algorithms can be designed and used more equitably.
The researchers organized existing work into a taxonomy of the types of tasks for which machine learning and AI are used that span prediction, interventions, identifying effects and allocations, to show examples of how a multi-level perspective can be leveraged. In the piece, the authors also show how the same framework is applicable to considerations of data privacy, governance, and best practices to move the healthcare burden from individuals, toward improving equity.
As an example of such approaches, members of the same team recently presented at the AAAI/ACM Conference on Artificial Intelligence, Ethics and Society a new approach to using “causal multi-level fairness,” the larger web of relevant data for assessing fairness of algorithms. This work builds on the field of “algorithmic fairness,” which, to date, is limited by its exclusive focus on individual-level attributes such as gender and race.
In this work Mhasawade and Chunara formalized a novel approach to understanding fairness relationships using tools from causal inference, synthesizing a means by which an investigator could assess and account for effects of sensitive macro attributes and not merely individual factors. They developed the algorithm for their approach and provided the settings under which it is applicable. They also illustrated their method on data showing how predictions based merely on data points associated with labels like race, income and gender are of limited value if sensitive attributes are not accounted for, or are accounted for without proper context.
“As in healthcare, algorithmic fairness tends to be focused on labels — men and women, Black versus white, etc. — without considering multiple layers of influence from a causal perspective to decide what is fair and unfair in predictions,” said Chunara. “Our work presents a framework for thinking not only about equity in algorithms but also what types of data we use in them.” More