Goto

Collaborating Authors

 Country


Gaussian-Spherical Restricted Boltzmann Machines

arXiv.org Artificial Intelligence

We consider a special type of Restricted Boltzmann machine (RBM), namely a Gaussian-spherical RBM where the visible units have Gaussian priors while the vector of hidden variables is constrained to stay on an ${\mathbbm L}_2$ sphere. The spherical constraint having the advantage to admit exact asymptotic treatments, various scaling regimes are explicitly identified based solely on the spectral properties of the coupling matrix (also called weight matrix of the RBM). Incidentally these happen to be formally related to similar scaling behaviours obtained in a different context dealing with spatial condensation of zero range processes. More specifically, when the spectrum of the coupling matrix is doubly degenerated an exact treatment can be proposed to deal with finite size effects. Interestingly the known parallel between the ferromagnetic transition of the spherical model and the Bose-Einstein condensation can be made explicit in that case. More importantly this gives us the ability to extract all needed response functions with arbitrary precision for the training algorithm of the RBM. This allows us then to numerically integrate the dynamics of the spectrum of the weight matrix during learning in a precise way. This dynamics reveals in particular a sequential emergence of modes from the Marchenko-Pastur bulk of singular vectors of the coupling matrix.


Integrating Machine Learning With Microsimulation to Classify Hypothet POR

#artificialintelligence

Purpose: Variability in patient treatment responses can be a barrier to effective care. Utilization of available patient databases may improve the prediction of treatment responses. We evaluated machine learning methods to predict novel, individual patient responses to pregabalin for painful diabetic peripheral neuropathy, utilizing an agent-based modeling and simulation platform that integrates real-world observational study (OS) data and randomized clinical trial (RCT) data. Patients and methods: The best supervised machine learning methods were selected (through literature review) and combined in a novel way for aligning patients with relevant subgroups that best enable prediction of pregabalin responses. Data were derived from a German OS of pregabalin (N 2642) and nine international RCTs (N 1320).


How machine learning is unlocking the secrets of human movement – and reshaping pro sports Transform

#artificialintelligence

Twelve of the best basketball players on the planet are each a blur of action and a bundle of hope. As hip-hop beats fill the lab, the group trains without touching a single basketball, individually bouncing side to side on an indoor track, soaring over a three-foot-tall box, and slinging weighted balls against a wall. One athlete, however, is producing both sweat and data – jumping, lifting and sprinting as cloud-connected cameras record his every movement at Peak Performance Project (P3) in Santa Barbara, California. All 12 of these college athletes expect to enjoy a long NBA career. And far away from P3, many sports analysts are equally certain they already know which of these NBA Draft hopefuls will become a legend, a valuable starter or a key contributor off the bench.


Detroit plant now producing self-driving vehicles with Waymo

#artificialintelligence

Not only is Detroit building vehicles people can drive, but now it is producing vehicles that can drive themselves. John Krafcik, CEO of Google self-driving affiliate Waymo LLC, said Monday that its Detroit plant is operating and outfitting fleets of vehicles with its autonomous driving hardware and software. The milestone allows the Alphabet Inc. subsidiary to put its automated "driver" into vehicles at mass scale. Doing so will help Waymo, an acknowledged leader in the self-driving space, to test its technology and expand its robotaxi service. Google self-driving affiliate Waymo LLC's Detroit plant already has outfitted 30 Jaguar I-PACE SUVs with the company's autonomous driving technology.


Volkswagen sets up autonomous driving subsidiary, plans Silicon Valley site next year

#artificialintelligence

Volkswagen Group announced the creation of a subsidiary called Volkswagen Autonomy (VWAT) on Monday, with the German car giant saying it planned to "make autonomous driving market-ready." With offices in Munich and Wolfsburg, Volkswagen said that VWAT would aim to "bring a self-driving system… to market maturity." As well as its sites in Germany, Volkswagen said it also planned to establish companies in Silicon Valley and China in 2020 and 2021 respectively. Alexander Hitzinger, the Volkswagen Group's senior vice president for autonomous driving, will manage the new company. "We want to establish Volkswagen Autonomy as a global technology company where we bundle expertise from the automotive and technology industries, combining the agility and creativity of a high-performance culture with process orientation and scalability," Hitzinger said in a statement.


Gasoline Or Batteries For Robo-taxis? It Depends.

#artificialintelligence

It seems inevitable that whenever automated cars are referenced, they're assumed to be electric vehicles, i.e. This fits the image of a utopian future of ubiquitous mobility which doesn't pump out tailpipe emissions and hasten the end of human civilization. From an engineering perspective, there are some specific advantages offered by electric drive, particularly the ability to directly control individual wheel motors for acceleration and braking. If you're starting from scratch and building your own vehicle, basic electric platforms are readily available; you don't have to work with a traditional car manufacturer. This is why we've seen the rise of low speed automated shuttles based on simple but effective electric platforms, as well as custom-built parcel delivery bots.


Microsoft Reveals Project HAMS, an AI-Powered Tool for Automating Driving Tests - WinBuzzer

#artificialintelligence

Microsoft has announced a new artificial intelligence-based project that aims to boost driver licensing by automating tests. Called Project HAMS, the new tool is being tested in India by the Regional Transport Office (RTO) in the city of Dehradun in the northern state of Uttarakhand. . Project HAMS (Harnessing Automobile for Safety) automates licensing by removing a trainer from the vehicle. Instead, a smartphone is docked to the windshield of a car. Microsoft says HAMS will allow Indian authorities to better police licensing discrepancies.


7 AI Cancer Diagnostics Startups Digitizing Healthcare

#artificialintelligence

The world seems more divided today than ever, whether we're talking about politics or the questionable art form of twerking. However, there's one thing we can all agree on: cancer sucks. Nearly 40% of us will receive the dreaded diagnosis at some point in our lives, according to the National Cancer Institute. That's one reason why we've spent quite a bit of time writing about the topic, particularly the different technologies being developed to detect various forms of the disease. It's really a no-brainer: Data from Cancer Research UK suggests 80% of patients survive for at least 10 years after being diagnosed in the early stages of eight of the most common cancers.


Pilots can't spot drones 70 per cent of the time - shock experiment reveals

Daily Mail - Science & tech

Pilots can't spot drones as they approach a runway, warns a shock new study. They fail to catch sight of the flying gadgets 70 per cent of the time - even when they are in their airspace, according to the findings. And they almost never identify the machines if they are hovering motionless above the ground. The disturbing findings uncover a'real and present danger' to safety, warn US aviation experts. Study co author Dr Ryan Wallace, of Embry-Riddle Aeronautical University in the United States, said: 'Dangerous close encounters between aircraft and drones are becoming an increasingly common problem.


Scientists say they can predict memory fading in your 70s by looking at tests you took at eight

Daily Mail - Science & tech

Scientists say they can tell how badly your memory will fade six decades before it happens. Participants who scored poorly on mental tests at the age of eight were more likely to have thinking and memory problems by the time they reached 70. Researchers said their findings suggest subtle cognitive differences could be a marker for dementia before the symptoms appear. However, the results did not investigate whether cognitive skills in childhood are linked to the risk of developing dementia. Education and income, assessed by occupation at the age of 53, was also shown to be an indicator of how brain power declines in your 70s.