Collaborating Authors


Technical Perspective: Entity Matching with Magellan

Communications of the ACM

Ferdinand Magellan was a Portuguese explorer who launched a Spanish expedition that completed the first circumnavigation of the Earth. It is in this spirit that Magellan was used as the name of the end-to-end entity matching system that is developed at the University of Wisconsin. Entity matching (also known as entity resolution or reference reconciliation or deduplication) is a major task in the larger problem of data integration, a problem that is pervasive in many organizations. Despite being a subject of extensive research for many years, the entity matching problem is surprisingly simple to describe and understand. It is to determine whether two different representations refer to the same real-world entity. Doe, UWisc) and (John Doe, Univ. of Wisconsin)--refer to the same person.

3D hand-sensing wristband uses a Raspberry Pi for machine learning


Researchers from Cornell and the University of Wisconsin, Madison, have designed a wrist-mounted device that tracks the entire human hand in 3D. The device (pictured) uses the contours from the wearer's wrist to create an abstraction of 20 finger joint positions. The FingerTrak bracelet uses low-resolution thermal cameras that read the wrist contours and a tethered Raspberry Pi 4 and machine learning to teach itself what the hand is doing based on these readings. Cheng Zhang, assistant professor of information science and director of Cornell's new SciFi Lab, where FingerTrak was developed said: "The most novel technical finding in this work is discovering that the contours of the wrist are enough to accurately predict the entire hand posture," Zhang said. "This finding allows the reposition of the sensing system to the wrist, which is more practical for usability."

Wrist-mounted wearable tracks your hand in 3D using thermal sensors


Modern wearables like the Apple Watch use sensors like gyros and accelerometers to detect hand movements. Those components allow them to turn on their displays when you lift your wrist, as well as to ensure you've properly washed your hand. But thanks to the work of a joint team of researchers at Cornell Unversity and the University of Wisconsin-Madison, future wearables could offer more nuanced hand detection. The team has developed a device they call the FingerTrak. The wearable employs four low-resolution thermal sensors, each the size of about a pea, to monitor the contours of the wrist.

Researchers show FingerTrak, a hand tracking wristband for AR/VR input


As virtual and augmented reality steadily advance in both visual fidelity and headset comfort, researchers continue to work on input solutions that will feel more natural than holding controllers. Today, a group of researchers announced FingerTrak, a wristband-based solution that uses thermal cameras to track hand movements in 3D, abstracting 20 finger joint positions from contours on the wearer's wrist. Developed by Cornell University's SciFi Lab with assistance from University of Wisconsin, Madison researchers, FingerTrak uses a deep neural network to stitch together input from three or four miniature thermal cameras mounted around the wrist, collectively capturing an entire hand pose. Using silhouettes generated by the cameras, backbone and regression networks estimate fingertip and joint positions, and though the results aren't perfect, they could be used for some forms of VR and AR input. Other potential applications for FingerTrak include human-robot interaction and control, sign language translation, and mobile health, including early detection of degenerative diseases such as Parkinson's and Alzheimer's.

Apple acquires Inductiv, a machine learning startup


Apple recently acquired Inductiv Inc. Inductiv, a machine learning startup, is co-founded by Professor Ihab Francis Ilyas, University of Waterloo. It has other professor co-founders from the University of Wisconsin and Stanford University. Inductiv uses artificial intelligence to correct errors in data. Apple has so far acquired four companies in 2020– DarkSky, Voysis, and NextVR. Inductiv is the 5th acquisition in 2020.

Turn-Taking and Coordination in Human-Machine Interaction

AI Magazine

This issue of AI Magazine brings together a collection of articles on challenges, mechanisms, and research progress in turn-taking and coordination between humans and machines. The contributing authors work in interrelated fields of spoken dialog systems, intelligent virtual agents, human-computer interaction, human-robot interaction, and semiautonomous collaborative systems and explore core concepts in coordinating speech and actions with virtual agents, robots, and other autonomous systems. Several of the contributors participated in the AAAI Spring Symposium on Turn-Taking and Coordination in Human-Machine Interaction, held in March 2015, and several articles in this issue are extensions of work presented at that symposium. The articles in the collection address key modeling, methodological, and computational challenges in achieving effective coordination with machines, propose solutions that overcome these challenges under sensory, cognitive, and resource restrictions, and illustrate how such solutions can facilitate coordination across diverse and challenging domains.

Apple acquires machine learning startup Inductiv Inc. to improve Siri data – 9to5Mac – IAM Network


Apple has acquired the machine learning startup Inductiv Inc., according to a new report from Bloomberg. The startup had been developing technology that uses artificial intelligence to identify and correct errors in datasets. The report explains that the engineering team from Inductiv has joined Apple "in recent weeks" to work on several different projects including Siri, machine learning, and data science. Apple issued its standard statement regarding the acquisition, saying it "buys smaller technology companies from time to time and we generally do not discuss our purpose or plans." The startup was founded by professors from Stanford University, the University of Waterloo, and the University of Wisconsin.

Covid-19 Makes the Case for More Meatpacking Robots


On Memorial Day weekends past, you might have joined in the All-American ritual of firing up the grill, cracking a cold one, and feuding with your family over which hot dog condiment is correct. But this holiday, you might not have as many wieners to argue about. Across the US, the coronavirus that causes Covid-19 has rampaged through cold, cramped, meat processing facilities, sickening thousands of workers and killing at least 30 of them. With dozens of plants closed or cutting back operations, meat shortages have been forcing some grocery stores to ration grilling staples like ground beef and chicken breasts. At least one sausage factory, in Milwaukee, has had to hit pause on its hot dog production line.

Simple 'smart' glass reveals the future of artificial vision


From left to right, Zongfu Yu, Ang Chen and Efram Khoram developed the concept for a "smart" piece of glass that recognizes images without any external power or circuits. The sophisticated technology that powers face recognition in many modern smartphones someday could receive a high-tech upgrade that sounds -- and looks -- surprisingly low-tech. This window to the future is none other than a piece of glass. University of Wisconsin–Madison engineers have devised a method to create pieces of "smart" glass that can recognize images without requiring any sensors or circuits or power sources. "We're using optics to condense the normal setup of cameras, sensors and deep neural networks into a single piece of thin glass," says UW-Madison electrical and computer engineering professor Zongfu Yu.

Differential Privacy for Sequential Algorithms Machine Learning

We study the differential privacy of sequential statistical inference and learning algorithms that are characterized by random termination time. Using the two examples: sequential probability ratio test and sequential empirical risk minimization, we show that the number of steps such algorithms execute before termination can jeopardize the differential privacy of the input data in a similar fashion as their outputs, and it is impossible to use the usual Laplace mechanism to achieve standard differentially private in these examples. To remedy this, we propose a notion of weak differential privacy and demonstrate its equivalence to the standard case for large i.i.d. samples. We show that using the Laplace mechanism, weak differential privacy can be achieved for both the sequential probability ratio test and the sequential empirical risk minimization with proper performance guarantees. Finally, we provide preliminary experimental results on the Breast Cancer Wisconsin (Diagnostic) and Landsat Satellite Data Sets from the UCI repository.