Group items tagged

love this one ... it seems to take physicist to explain to the AI crowd what they are actually doing ... Deep learning is a broad set of techniques that uses multiple layers of representation to automatically learn relevant features directly from structured data. Recently, such techniques have yielded record-breaking results on a diverse set of difficult machine learning tasks in computer vision, speech recognition, and natural language processing. Despite the enormous success of deep learning, relatively little is understood theoretically about why these techniques are so successful at feature learning and compression. Here, we show that deep learning is intimately related to one of the most important and successful techniques in theoretical physics, the renormalization group (RG). RG is an iterative coarse-graining scheme that allows for the extraction of relevant features (i.e. operators) as a physical system is examined at different length scales. We construct an exact mapping from the variational renormalization group, first introduced by Kadanoff, and deep learning architectures based on Restricted Boltzmann Machines (RBMs). We illustrate these ideas using the nearest-neighbor Ising Model in one and two-dimensions. Our results suggests that deep learning algorithms may be employing a generalized RG-like scheme to learn relevant features from data.

Astronomers using the Hubble Space Telescope have created the deepest multi-color* image of the Universe ever taken: the Hubble Extreme Deep Field, a mind-blowing glimpse into the vast stretches of our cosmos.

New experiments and supercomputer computations discovered that iron oxide undergoes a new kind of transition under deep Earth conditions. Iron oxide, FeO, is a component of the second most abundant mineral at Earth's lower mantle, ferropericlase. The finding, published in an upcoming issue of Physical Review Letters, could alter our understanding of deep Earth dynamics and the behavior of the protective magnetic field, which shields our planet from harmful cosmic rays.

Researchers from Northwestern University and the University of New Mexico report evidence for potentially oceans worth of water deep beneath the United States. Though not in the familiar liquid form-the ingredients for water are bound up in rock deep in the Earth's mantle-the discovery may represent the planet's largest water reservoir.

A good historical overview of the Deep Learning revolution. If you think the quote above is an exageration, here are some fresh news from Microsoft: Internal email: Microsoft forms new 5,000-person AI division

Winter is coming... and deep learning, too!

"Sergey Brin has said he wants to build a benign version of HAL in 2001: A Space Odyssey" ... didn't they try to do that in that movie called "2001: A Space Odyssey" ?

Physicists from MIT explain why deep learning is that much successful using physics.

Implementation of the deep learning-based image classifier (online). Try making a picture with your phone and upload it there. Pretty impressive results. EDIT: Okay, it works the best with well exposed simple objects (pen, mug).

NASA is setting a $2.25 million competition to design and build a 3-D printed habitat for deep space exploration - new task for the future architect of the team?

I kind of like Thijs's space mirror ablation method as a way of printing :))

In a world first, an artificial intelligence machine plays chess by evaluating the board rather than using brute force to work out every possible move. It's been almost 20 years since IBM's Deep Blue supercomputer beat the reigning world chess champion, Gary Kasparov, for the first time under standard tournament rules.

The disadvantage in this kind of engine lies exactly in its inability to extrapolate. You might actually be able to beat it if you play like an idiot.

Alvin, the only deep-diving manned submersible used for science has been upgraded, allowing scientists to explore for the 1st time the depth of the oceans down to 6500m bsl.

Genetic algorithms are a competitive alternative for training deep neural networks for reinforcement learning. For paper see: https://arxiv.org/pdf/1712.06567.pdf

Interesting pointers in this one! I would like to explore neuroevolution as well, although it seems extremely resource-demanding?

Not necessarily, I think it can be made to be much faster hybridizing it with backprop and Taylor maps. Its one ideas in the closet we still have not explored (Differential Intelligence: accelerating neuroevolution).

New optics based computer architecture

An intelligent (that is, self-adaptive) cloak driven by deep learning. Quite cool, eh? Something we can get inspired by?

Galactic Cosmic Radiation consisting of high-energy, high-charged (HZE) particles poses a significant threat to future astronauts in deep space. Aside from cancer, concerns have been raised about late degenerative risks, including effects on the brain. In this study we examined the effects of 56Fe particle irradiation in an APP/PS1 mouse model of Alzheimer's disease (AD). We demonstrated 6 months after exposure to 10 and 100 cGy 56Fe radiation at 1 GeV/µ, that APP/PS1 mice show decreased cognitive abilities measured by contextual fear conditioning and novel object recognition tests. Furthermore, in male mice we saw acceleration of Aβ plaque pathology using Congo red and 6E10 staining, which was further confirmed by ELISA measures of Aβ isoforms. Increases were not due to higher levels of amyloid precursor protein (APP) or increased cleavage as measured by levels of the β C-terminal fragment of APP. Additionally, we saw no change in microglial activation levels judging by CD68 and Iba-1 immunoreactivities in and around Aβ plaques or insulin degrading enzyme, which has been shown to degrade Aβ. However, immunohistochemical analysis of ICAM-1 showed evidence of endothelial activation after 100 cGy irradiation in male mice, suggesting possible alterations in Aβ trafficking through the blood brain barrier as a possible cause of plaque increase. Overall, our results show for the first time that HZE particle radiation can increase Aβ plaque pathology in an APP/PS1 mouse model of AD.

"A band of space hackers and engineers are trying to do something never done before - recover a 36 year old NASA spacecraft from the grips of deep space and time. With old NASA documents and Rockethub crowdfunding, a team led by Dennis Wingo and Keith Cowing is attempting to steer ISEE-3, later rechristened ICE, the International Cometary Explorer, back into an Earth orbit and return it to scientific operations. Dennis says, 'ISEE-3 can become a great teaching tool for future engineers and scientists helping with design and travel to Mars'. Only 40 days remain before the spacecraft will be out of range for recovery. A radio telescope is available, propulsion designs are in hand and the team is hoping for public support to provide the small amount needed to accomplish a very unique milestone in space exploration

Deep neural networks "dreaming" psychedelic images

Google released the code behind this, and applications of it are now popping up: - http://googleresearch.blogspot.nl/2015/07/deepdream-code-example-for-visualizing.html - https://imgur.com/a/nVlLe - https://twitter.com/hashtag/deepdream

Funny how representation errors in NNs suddenly become art. God.... neo-post-modernism.

strange that even if optical fibre this would not compromise the submarines location ...

Amazing "to look deep into space and time". What a nice pun!

Coooool!

Basics of Space Flight is a tutorial designed primarily to help operations people identify the range of concepts associated with deep space missions, and grasp the relationships among them.