Computers Math

The Fourier transform is an important mathematical tool that decomposes a function or dataset into a its constituting frequencies, much like one could decompose a musical chord into a combination of its notes. It is used across all fields of engineering in some form or another and, accordingly, algorithms to compute it efficiently have been developed — that is, at least for conventional computers. But what about quantum computers?
Though quantum computing remains an enormous technical and intellectual challenge, it has the potential to speed up many programs and algorithms immensely provided that appropriate quantum circuits are designed. In particular, the Fourier transform already has a quantum version called the quantum Fourier transform (QFT), but its applicability is quite limited because its results cannot be used in subsequent quantum arithmetic operations.
To address this issue, in a recent study published in Quantum Information Processing, scientists from Tokyo University of Science developed a new quantum circuit that executes the “quantum fast Fourier transform (QFFT)” and fully benefits from the peculiarities of the quantum world. The idea for the study came to Mr. Ryo Asaka, first-year Master’s student and one of the scientists on the study, when he first learned about the QFT and its limitations. He thought it would be useful to create a better alternative based on a variant of the standard Fourier transform called the “fast Fourier transform (FFT),” an indispensable algorithm in conventional computing that greatly speeds things up if the input data meets some basic conditions.
To design the quantum circuit for the QFFT, the scientists had to first devise quantum arithmetic circuits to perform the basic operations of the FFT, such as addition, subtraction, and digit shifting. A notable advantage of their algorithm is that no “garbage bits” are generated; the calculation process does not waste any qubits, the basic unit of quantum information. Considering that increasing the number of qubits of quantum computers has been an uphill battle over the last few years, the fact that this novel quantum circuit for the QFFT can use qubits efficiently is very promising.
Another merit of their quantum circuit over the traditional QFT is that their implementation exploits a unique property of the quantum world to greatly increase computational speed. Associate Professor Kazumitsu Sakai, who led the study, explains: “In quantum computing, we can process a large amount of information at the same time by taking advantage of a phenomenon known as ‘superposition of states.’ This allows us to convert a lot of data, such as multiple images and sounds, into the frequency domain in one go.” Processing speed is regularly cited as the main advantage of quantum computing, and this novel QFFT circuit represents a step in the right direction.
Moreover, the QFFT circuit is much more versatile than the QFT, as Assistant Professor Ryoko Yahagi, who also participated in the study, remarks: “One of the main advantages of the QFFT is that it is applicable to any problem that can be solved by the conventional FFT, such as the filtering of digital images in the medical field or analyzing sounds for engineering applications.” With quantum computers (hopefully) right around the corner, the outcomes of this study will make it easier to adopt quantum algorithms to solve the many engineering problems that rely on the FFT.

Story Source:
Materials provided by Tokyo University of Science. Note: Content may be edited for style and length. More

International scientists are challenging their colleagues to make Artificial Intelligence (AI) research more transparent and reproducible to accelerate the impact of their findings for cancer patients.
In an article published in Nature on October 14, 2020, scientists at Princess Margaret Cancer Centre, University of Toronto, Stanford University, Johns Hopkins, Harvard School of Public Health, Massachusetts Institute of Technology, and others, challenge scientific journals to hold computational researchers to higher standards of transparency, and call for their colleagues to share their code, models and computational environments in publications.
“Scientific progress depends on the ability of researchers to scrutinize the results of a study and reproduce the main finding to learn from,” says Dr. Benjamin Haibe-Kains, Senior Scientist at Princess Margaret Cancer Centre and first author of the article. “But in computational research, it’s not yet a widespread criterion for the details of an AI study to be fully accessible. This is detrimental to our progress.”
The authors voiced their concern about the lack of transparency and reproducibility in AI research after a Google Health study by McKinney et al., published in a prominent scientific journal in January 2020, claimed an artificial intelligence (AI) system could outperform human radiologists in both robustness and speed for breast cancer screening. The study made waves in the scientific community and created a buzz with the public, with headlines appearing in BBC News, CBC, CNBC.
A closer examination raised some concerns: the study lacked a sufficient description of the methods used, including their code and models. The lack of transparency prohibited researchers from learning exactly how the model works and how they could apply it to their own institutions.
“On paper and in theory, the McKinney et al. study is beautiful,” says Dr. Haibe-Kains, “But if we can’t learn from it then it has little to no scientific value.”
According to Dr. Haibe-Kains, who is jointly appointed as Associate Professor in Medical Biophysics at the University of Toronto and affiliate at the Vector Institute for Artificial Intelligence, this is just one example of a problematic pattern in computational research.
“Researchers are more incentivized to publish their finding rather than spend time and resources ensuring their study can be replicated,” explains Dr. Haibe-Kains. “Journals are vulnerable to the ‘hype’ of AI and may lower the standards for accepting papers that don’t include all the materials required to make the study reproducible — often in contradiction to their own guidelines.”
This can actually slow down the translation of AI models into clinical settings. Researchers are not able to learn how the model works and replicate it in a thoughtful way. In some cases, it could lead to unwarranted clinical trials, because a model that works on one group of patients or in one institution, may not be appropriate for another.
In the article titled Transparency and reproducibility in artificial intelligence, the authors offer numerous frameworks and platforms that allow safe and effective sharing to uphold the three pillars of open science to make AI research more transparent and reproducible: sharing data, sharing computer code and sharing predictive models.
“We have high hopes for the utility of AI for our cancer patients,” says Dr. Haibe-Kains. “Sharing and building upon our discoveries — that’s real scientific impact.”

Story Source:
Materials provided by University Health Network. Note: Content may be edited for style and length. More

The closer people are physically to one another, the higher the chance for exchange, of things like ideas, information, and even infection. Now researchers at MIT and Boston Children’s Hospital have found that, even in the microscopic environment within a single cell, physical crowding increases the chance for interactions, in a way that can significantly alter a cell’s health and development.
In a paper published today in the journal Cell Stem Cell, the researchers have shown that physically squeezing cells, and crowding their contents, can trigger cells to grow and divide faster than they normally would.
While squeezing something to make it grow may sound counterintuitive, the team has an explanation: Squeezing acts to wring water out of a cell. With less water to swim in, proteins and other cell constituents are packed closer together. And when certain proteins are brought in close proximity, they can trigger cell signaling and activate genes within the cell.
In their new study, the scientists found that squeezing intestinal cells triggered proteins to cluster along a specific signaling pathway, which can help cells maintain their stem-cell state, an undifferentiated state in which can quickly grow and divide into more specialized cells. Ming Guo, associate professor of mechanical engineering at MIT, says that if cells can simply be squeezed to promote their “stemness,” they can then be directed to quickly build up miniature organs, such as artificial intestines or colons, which could then be used as platforms to understand organ function and test drug candidates for various diseases, and even as transplants for regenerative medicine.
Guo’s co-authors are lead author Yiwei Li, Jiliang Hu, and Qirong Lin from MIT, and Maorong Chen, Ren Sheng, and Xi He of Boston Children’s Hospital.
Packed in
To study squeezing’s effect on cells, the researchers mixed various cell types in solutions that solidified as rubbery slabs of hydrogel. To squeeze the cells, they placed weights on the hydrogel’s surface, in the form of either a quarter or a dime.

advertisement

“We wanted to achieve a significant amount of cell size change, and those two weights can compress the cell by something like 10 to 30 percent of their total volume,” Guo explains.
The team used a confocal microscope to measure in 3D how individual cells’ shapes changed as each sample was compressed. As they expected, the cells shrank with pressure. But did squeezing also affect the cell’s contents? To answer this, the researchers first looked to see whether a cell’s water content changed. If squeezing acts to wring water out of a cell, the researchers reasoned that the cells should be less hydrated, and stiffer as a result.
They measured the stiffness of cells before and after weights were applied, using optical tweezers, a laser-based technique that Guo’s lab has employed for years to study interactions within cells, and found that indeed, cells stiffened with pressure. They also saw that there was less movement within cells that were squeezed, suggesting that their contents were more packed than usual.
Next, they looked at whether there were changes in the interactions between certain proteins in the cells, in response to cells being squeezed. They focused on several proteins that are known to trigger Wnt/?-catenin signaling, which is involved in cell growth and maintenance of “stemness.”
“In general, this pathway is known to make a cell more like a stem cell,” Guo says. “If you change this pathway’s activity, how cancer progresses and how embryos develop have been shown to be very different. So we thought we could use this pathway to demonstrate how cell crowding is important.”
A “refreshing” path

advertisement

To see whether cell squeezing affects the Wnt pathway, and how fast a cell grows, the researchers grew small organoids — miniature organs, and in this case, clusters of cells that were collected from the intestines of mice.
“The Wnt pathway is particularly important in the colon,” Guo says, pointing out that the cells that line the human intestine are constantly being replenished. The Wnt pathway, he says, is essential for maintaining intestinal stem cells, generating new cells, and “refreshing” the intestinal lining.
He and his colleagues grew intestinal organoids, each measuring about half a millimeter, in several Petri dishes, then “squeezed” the organoids by infusing the dishes with polymers. This influx of polymers increased the osmotic pressure surrounding each organoid and forced water out of their cells. The team observed that as a result, specific proteins involved in activating the Wnt pathway were packed closer together, and were more likely to cluster to turn on the pathway and its growth-regulating genes.
The upshot: Those organoids that were squeezed actually grew larger and more quickly, with more stem cells on their surface than those that were not squeezed.
“The difference was very obvious,” Guo says. “Whenever you apply pressure, the organoids grow even bigger, with a lot more stem cells.”
He says the results demonstrate how squeezing can affect a organoid’s growth. The findings also show that a cell’s behavior can change depending on the amount of water that it contains.
“This is very general and broad, and the potential impact is profound, that cells can simply tune how much water they have to tune their biological consequences,” Guo says.
Going forward, he and his colleagues plan to explore cell squeezing as a way to speed up the growth of artificial organs that scientists may use to test new, personalized drugs.
“I could take my own cells and transfect them to make stem cells that can then be developed into a lung or intestinal organoid that would mimic my own organs,” Guo says. “I could then apply different pressures to make organoids of different size, then try different drugs. I imagine there would be a lot of possibilities.”
This research is supported, in part, by the National Cancer Institute and the Alfred P. Sloan Foundation. More

In the era of electric cars, machine learning and ultra-efficient vehicles for space travel, computers and hardware are operating faster and more efficiently. But this increase in power comes with a trade-off: They get superhot.
To counter this, University of Central Florida researchers are developing a way for large machines to “breathe” in and out cooling blasts of water to keep their systems from overheating.
The findings are detailed in a recent study in the journal Physical Review Fluids.
The process is much like how humans and some animals breath in air to cool their bodies down, except in this case, the machines would be breathing in cool blasts of water, says Khan Rabbi, a doctoral candidate in UCF’s Department of Mechanical and Aerospace Engineering and lead author of the study.
“Our technique used a pulsed water-jet to cool a hot titanium surface,” Rabbi says. “The more water we pumped out of the spray jet nozzles, the greater the amount of heat that transferred between the solid titanium surface and the water droplets, thus cooling the titanium. Fundamentally, an idea of optimum jet-pulsation needs to be generated to ensure maximum heat transfer performance.”
“It is essentially like exhaling the heat from the surface,” he says.

advertisement

The water is emitted from small water-jet nozzles, about 10 times the thickness of a human hair, that douse a hot surface of a large electronic system and the water is collected in a storage chamber, where it can be pumped out and circulated again to repeat the cooling process. The storage chamber in their study held about 10 ounces of water.
Using high-speed, infrared thermal imaging, the researchers were able to find the optimum amount of water for maximum cooling performance.
Rabbi says everyday applications for the system could include cooling large electronics, space vehicles, batteries in electric vehicles and gas turbines.
Shawn Putnam, an associate professor in UCF’s Department of Mechanical and Aerospace Engineering and study co-author, says that this research is part of an effort to explore different techniques to efficiently cool hot devices and surfaces.
“Most likely, the most versatile and efficient cooling technology will take advantage of several different cooling mechanisms, where pulsed jet cooling is expected to be one of these key contributors,” Putnam says.
The researcher says there are multiple ways to cool hot hardware, but water-jet cooling is a preferred method because it can be adjusted to different directions, has good heat-transfer ability, and uses minimum amounts of water or liquid coolant.
However, it has its drawbacks, namely either over or underwatering that results in floods or dry hotspots. The UCF method overcomes this problem by offering a system that is tunable to hardware needs so that the only water applied is the amount needed and in the right spot.
The technology is needed since once device temperatures surpass a threshold value, for example, 194 degrees Fahrenheit, the device’s performance decreases, Rabbi says.
“For this reason, we need better cooling technologies in place to keep the device temperature well within the maximum temperature for optimum operation,” he says. “We believe this study will provide engineers, scientists and researchers a unique understanding to develop future generation liquid cooling systems.”

Story Source:
Materials provided by University of Central Florida. Original written by Robert H Wells. Note: Content may be edited for style and length. More

Our understanding of quantum physics has involved the creation of a wide range of “quasiparticles.” These notional constructs describe emergent phenomena that appear to have the properties of multiple other particles mixed together.
An exciton, for example, is a quasiparticle that acts like an electron bound to an electron hole, or the empty space in a semiconducting material where an electron could be. A step further, an exciton-polariton combines the properties of an exciton with that of a photon, making it behave like a combination of matter and light. Achieving and actively controlling the right mixture of these properties — such as their mass, speed, direction of motion, and capability to strongly interact with one another — is the key to applying quantum phenomena to technology, like computers.
Now, researchers at the University of Pennsylvania’s School of Engineering and Applied Science are the first to create an even more exotic form of the exciton-polariton, one which has a defined quantum spin that is locked to its direction of motion. Depending on the direction of their spin, these helical topological exciton-polaritons move in opposite directions along the surface of an equally specialized type of topological insulator.
In a study published in the journal Science, they have demonstrated this phenomenon at temperatures much warmer than the near-absolute-zero usually required to maintain this sort of quantum phenomenon. The ability to route these quasiparticles based on their spin in more user-friendly conditions, and an environment where they do not back-scatter, opens up the possibility of using them to transmit information or perform computations at unprecedented speeds.
The study was led by Ritesh Agarwal, professor in the Department of Materials Science and Engineering, and Wenjing Liu, a postdoctoral researcher in his lab. They collaborated with researchers from Hunan University and George Washington University.
The study also demonstrates a new type of topological insulators, a class of material developed at Penn by Charles Kane and Eugene Mele that has a conductive surface and an insulating core. Topological insulators are prized for their ability to propagate electrons at their surface without scattering them, and the same idea can be extended to quasiparticles such as photons or polaritons.

advertisement

“Replacing electrons with photons would make for even faster computers and other technologies, but photons are very hard to modulate, route or switch. They cannot be transported around sharp turns and leak out of the waveguide,” Agarwal says. “This is where topological exciton-polaritons can be useful, but that means we need to make new types of topological insulators that can work with polaritons. If we could make this type of quantum material, we could route exciton-polaritons along certain channels without any scattering, as well as modulate or switch them via externally applied electric fields or by slight changes in temperature.”
Agarwal’s group has created several types of photonic topological insulators in the past. While the first “chiral” polariton topological insulator was reported by a group in Europe, it worked at extremely low temperatures while requiring strong magnetic fields The missing piece, and distinction between “chiral” and “helical” in this case, was the ability to control the direction of flow via the quasiparticles’ spin.
“To create this phase, we used an atomically thin semiconductor, tungsten disulfide, which forms very tightly bound excitons, and coupled it strongly to a properly designed photonic crystal via symmetry engineering. This induced nontrivial topology to the resulting polaritons,” Agarwal says. “At the interface between photonic crystals with different topology, we demonstrated the generation of helical topological polaritons that did not scatter at sharp corners or defects, as well as spin-dependent transport.”
Agarwal and his colleagues conducted the study at 200K, or roughly -100F without the need for applying any magnetic fields. While that seems cold, it is considerably warmer — and easier to achieve — than similar systems that operate at 4K, or roughly -450F.
They are confident that further research and improved fabrication techniques for their semiconductor material will easily allow their design to operate at room temperature.
“From an academic point of view, 200K is already almost room temperature, so small advances in material purity could easily push it to working in ambient conditions,” says Agarwal. “Atomically thin, ‘2D’ materials form very strong excitons that survive room temperature and beyond, so we think we need only small modifications to how our materials are assembled.”
Agarwal’s group is now working on studying how topological polaritons interact with one another, which would bring them a step closer to using them in practical photonic devices. More

Smartphone apps that tell commuters when a bus will arrive at a stop don’t result in less time waiting than reliance on an official bus route schedule, a new study suggests.
In fact, people who followed the suggestions of transit apps to time their arrival for when the bus pulls up to the stop were likely to miss the bus about three-fourths of the time, results showed.
“Following what transit apps tell you about when to leave your home or office for the bus stop is a risky strategy,” said Luyu Liu, lead author of the study and a doctoral student in geography at The Ohio State University.
“The app may tell you the bus will be five minutes late, but drivers can make up time after you start walking, and you end up missing the bus.”
The best choice on average for bus commuters is to refer to the official schedule, or at least build in extra time when using the app’s suggestions, according to the researchers.
Liu conducted the study with Harvey Miller, professor of geography and director of Ohio State’s Center for Urban and Regional Analysis. The study was published recently online in the journal Transportation Research Part A.

advertisement

“We’re not saying that real-time bus information is bad. It is reassuring to know that a bus is coming,” Miller said.
“But if you’re going to use these apps, you have to know how to use them and realize it still won’t be better on average than following the schedule.”
For the study, the researchers analyzed bus traffic for one year (May 2018 to May 2019) on one route of the Central Ohio Transit Authority (COTA), the public bus system in Columbus.
Liu and Miller used the same real-time data that publicly available apps use to tell riders where buses are and when they are likely to reach individual stops. They compared the real-time data predictions of when buses would arrive at stops to when buses actually arrived for a popular bus route that traverses a large part of the city. The researchers then calculated the average time commuters would wait at a stop if they used different tactics to time their arrival, including just following the bus schedule.
The absolute worst way to catch the bus was using what the researchers called the “greedy tactic” — the one used by many transit apps — in which commuters timed their arrival at the stop to when the app said the bus would pull up.

advertisement

The average wait using the greedy tactic was about 12½ minutes — about three times longer than simply following the schedule. That’s because riders using this tactic are at high risk of missing the bus, researchers found.
The app tells riders when the bus will arrive based on where it is and how fast it is traveling when a commuter checks it, Miller said.
But there are two problems with that method, he said. For one, drivers can make up lost time.
“COTA wants to deliver on-time service, so bus operators understandably will try to get back on schedule,” Miller said.
Plus, the apps don’t check the bus location often enough to get accurate real-time information.
Slightly better was the “arbitrary tactic” when a person just randomly walked up to a stop and caught the next bus that arrived. Commuters using this tactic would wait on average about 8½ minutes for the next bus.
The second-best tactic was what the researchers called the “prudent tactic,” which was using the app to plan for arrival at the stop but adding some time as an “insurance buffer.” Here the average wait time was four minutes and 42 seconds, with a 10 percent risk of missing the bus.
The prudent tactic waiting time was similar to the “schedule tactic,” which is just using the public schedule to determine when to arrive at the stop. These commuters waited an average of four minutes and 12 seconds, with only a 6 percent risk of missing the bus.
There is some variation on waiting time within these averages, especially with the two tactics that use real-time information from apps. One of the most important factors is the length of a commuter’s walk to the bus stop.
Those who have longer walks take more risks when they rely on real-time information. If the app tells commuters their bus is running late, a long walk gives the bus more time to speed up to get back on schedule.
Another important factor is the length of time between buses arriving at a stop. A longer time between buses means more risk if you miss a bus, and results in more time waiting.
While on average the schedule tactic worked best, there were minor exceptions.
Results showed that it was generally better for work commuters to follow the schedule tactic in the morning when going to work and follow the prudent tactic using an app in the afternoon.
But one thing was certain, the researchers said: It was never a good idea to be greedy and try to achieve no waiting at the bus stop.
Waiting time for buses is an important issue, Miller said. For one, long waiting times is one of the top issues cited by people for not using public transportation.
It is also a safety concern for people to not have to wait for long periods at stops, especially at night, or for those rushing around busy streets because they are late for a bus. And for many people, missing buses can jeopardize their jobs or important health care appointments, Miller said.
Miller said the apps themselves could be more helpful by taking advantage of the data used in this study to make better recommendations.
“These apps shouldn’t be pushing risky strategies on users for eliminating waiting time. They should be more sophisticated,” he said. More

Artificial intelligence has arrived in our everyday lives — from search engines to self-driving cars. This has to do with the enormous computing power that has become available in recent years. But new results from AI research now show that simpler, smaller neural networks can be used to solve certain tasks even better, more efficiently, and more reliably than ever before.
An international research team from TU Wien (Vienna), IST Austria and MIT (USA) has developed a new artificial intelligence system based on the brains of tiny animals, such as threadworms. This novel AI-system can control a vehicle with just a few artificial neurons. The team says that system has decisive advantages over previous deep learning models: It copes much better with noisy input, and, because of its simplicity, its mode of operation can be explained in detail. It does not have to be regarded as a complex “black box,” but it can be understood by humans. This new deep learning model has now been published in the journal Nature Machine Intelligence.
Learning from nature
Similar to living brains, artificial neural networks consist of many individual cells. When a cell is active, it sends a signal to other cells. All signals received by the next cell are combined to decide whether this cell will become active as well. The way in which one cell influences the activity of the next determines the behavior of the system — these parameters are adjusted in an automatic learning process until the neural network can solve a specific task.
“For years, we have been investigating what we can learn from nature to improve deep learning,” says Prof. Radu Grosu, head of the research group “Cyber-Physical Systems” at TU Wien. “The nematode C. elegans, for example, lives its life with an amazingly small number of neurons, and still shows interesting behavioral patterns. This is due to the efficient and harmonious way the nematode’s nervous system processes information.”
“Nature shows us that there is still lots of room for improvement,” says Prof. Daniela Rus, director of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). “Therefore, our goal was to massively reduce complexity and enhance interpretability of neural network models.”
“Inspired by nature, we developed new mathematical models of neurons and synapses,” says Prof. Thomas Henzinger, president of IST Austria.

advertisement

“The processing of the signals within the individual cells follows different mathematical principles than previous deep learning models,” says Dr. Ramin Hasani, postdoctoral associate at the Institute of Computer Engineering, TU Wien and MIT CSAIL. “Also, our networks are highly sparse — this means that not every cell is connected to every other cell. This also makes the network simpler.”
Autonomous Lane Keeping
To test the new ideas, the team chose a particularly important test task: self-driving cars staying in their lane. The neural network receives camera images of the road as input and is to decide automatically whether to steer to the right or left.
“Today, deep learning models with many millions of parameters are often used for learning complex tasks such as autonomous driving,” says Mathias Lechner, TU Wien alumnus and PhD student at IST Austria. “However, our new approach enables us to reduce the size of the networks by two orders of magnitude. Our systems only use 75,000 trainable parameters.”
Alexander Amini, PhD student at MIT CSAIL explains that the new system consists of two parts: The camera input is first processed by a so-called convolutional neural network, which only perceives the visual data to extract structural features from incoming pixels. This network decides which parts of the camera image are interesting and important, and then passes signals to the crucial part of the network — a “control system” that then steers the vehicle.

advertisement

Both subsystems are stacked together and are trained simultaneously. Many hours of traffic videos of human driving in the greater Boston area were collected, and are fed into the network, together with information on how to steer the car in any given situation — until the system has learned to automatically connect images with the appropriate steering direction and can independently handle new situations.
The control part of the system (called neural circuit policy, or NCP), which translates the data from the perception module into a steering command, only consists of 19 neurons. Mathias Lechner explains that NCPs are up to 3 orders of magnitude smaller than what would have been possible with previous state-of-the-art models.
Causality and Interpretability
The new deep learning model was tested on a real autonomous vehicle. “Our model allows us to investigate what the network focuses its attention on while driving. Our networks focus on very specific parts of the camera picture: The curbside and the horizon. This behavior is highly desirable, and it is unique among artificial intelligence systems,” says Ramin Hasani. “Moreover, we saw that the role of every single cell at any driving decision can be identified. We can understand the function of individual cells and their behavior. Achieving this degree of interpretability is impossible for larger deep learning models.”
Robustness
“To test how robust NCPs are compared to previous deep models, we perturbed the input images and evaluated how well the agents can deal with the noise,” says Mathias Lechner. “While this became an insurmountable problem for other deep neural networks, our NCPs demonstrated strong resistance to input artifacts. This attribute is a direct consequence of the novel neural model and the architecture.”
“Interpretability and robustness are the two major advantages of our new model,” says Ramin Hasani. “But there is more: Using our new methods, we can also reduce training time and the possibility to implement AI in relatively simple systems. Our NCPs enable imitation learning in a wide range of possible applications, from automated work in warehouses to robot locomotion. The new findings open up important new perspectives for the AI community: The principles of computation in biological nervous systems can become a great resource for creating high-performance interpretable AI — as an alternative to the black-box machine learning systems we have used so far.”
Code Repository: https://github.com/mlech26l/keras-ncp
Video: https://ist.ac.at/en/news/new-deep-learning-models/ More

Hang bugs — when software gets stuck, but doesn’t crash — can frustrate both users and programmers, taking weeks for companies to identify and fix. Now researchers from North Carolina State University have developed software that can spot and fix the problems in seconds.
“Many of us have experience with hang bugs — think of a time when you were on website and the wheel just kept spinning and spinning,” says Helen Gu, co-author of a paper on the work and a professor of computer science at NC State. “Because these bugs don’t crash the program, they’re hard to detect. But they can frustrate or drive away customers and hurt a company’s bottom line.”
With that in mind, Gu and her collaborators developed an automated program, called HangFix, that can detect hang bugs, diagnose the relevant problem, and apply a patch that corrects the root cause of the error. Video of Gu discussing the program can be found here.
The researchers tested a prototype of HangFix against 42 real-world hang bugs in 10 commonly used cloud server applications. The bugs were drawn from a database of hang bugs that programmers discovered affecting various websites. HangFix fixed 40 of the bugs in seconds.
“The remaining two bugs were identified and partially fixed, but required additional input from programmers who had relevant domain knowledge of the application,” Gu says.
For comparison, it took weeks or months to detect, diagnose and fix those hang bugs when they were first discovered.
“We’re optimistic that this tool will make hang bugs less common — and websites less frustrating for many users,” Gu says. “We are working to integrate Hangfix into InsightFinder.” InsightFinder is the AI-based IT operations and analytics startup founded by Gu.
The paper, “HangFix: Automatically Fixing Software Hang Bugs for Production Cloud Systems,” is being presented at the ACM Symposium on Cloud Computing (SoCC’20), being held online Oct. 19-21. The paper was co-authored by Jingzhu He, a Ph.D. student at NC State who is nearing graduation; Ting Dai, a Ph.D. graduate of NC State who is now at IBM Research; and Guoliang Jin, an assistant professor of computer science at NC State.
The work was done with support from the National Science Foundation under grants 1513942 and 1149445.
HangFix is the latest in a long line of tools Gu’s team has developed to address cloud computing challenges. Her 2011 paper, “CloudScale: Elastic Resource Scaling for Multi-tenant Cloud Systems,” was selected as the winner of the 2020 SoCC 10-Year Award at this year’s conference.

Story Source:
Materials provided by North Carolina State University. Note: Content may be edited for style and length. More