Group items tagged

In this paper published on Nature, AI researchers used deep Q-network with very good adaptability and obtained performances comparable to those of a human games tester.

Bastards! And that was to be my next idea. Still no recurrency as I see it so far, so this is just some fancy way to do a markov model. Not sure if this is that particular paper or an earlier version but here it is for those interested: http://www.cs.toronto.edu/~vmnih/docs/dqn.pdf

According to their Nature article, they can detect "a tunnel less than 1m across, buried 2m underground" just from its gravitational difference. Using a device that they predict could cost ~100 € in mass production. UK researchers have built a small device that measures tiny fluctuations in gravity, and could be used to monitor volcanoes or search for oil. Such gravimeters already exist but compared to this postage stamp-sized gadget, they are bulky and pricy.

Impressive ... Should we have a closer look at this?

Just out - I wish I could understand this :(

ignorance is a bliss :)

RIP Mars One, you were stupid to begin with.

Who would have thought?

while the whole project was obviously a scam and i'm surprised how long it took courts to figure that out, it managed to get an enormous amount of people interested in space and ignited a conversation (even if that one was just about bashing the project :D)

Did we just solve overpopulation and climate change? With 40% more efficient crops we could easily sustain 10+ billion people on Earth. And 40% more efficient plants would absorb much more CO2 than we are emitting (currently: artificial CO2 emission ~29 GT/y, photosynthesis CO2 capture through plants ~450 GT/y) I am usually very worried about the risks of climate change, but this could be a real game changer!

I love the car animation!

This best-in-class streaming algorithm works by remembering just enough of what it's seen to tell you what it's seen most frequently. It suggests that compromises that seemed intrinsic to the analysis of streaming data are not actually necessary.

The paper I talked about with the proof of a gigantic lava tube discovered in 2017 from data of the Japanese lunar mission

Given the size of this thing we may consider a project on city urban planning inside. I would see a rocket-port outside the crater, some lift systems to get in, a closed dome with artificial atmosphere where to develop the city possible trekking routes on the lunar surface, a mirror system to get sunlight in, an energy factory just outside the entry.

uups :)

Well, I wouldn't expect a reply: "all our past emails have been downloaded and sold to NASA" even if that was the case.

Of course Marek is right... What matters is the theatre of security, not security itself. Just like in airports :)

Some people answering the question 'What's cooler than blasting space debris with lasers...?'.

If it fires in both directions, can we align it such that it deorbits two debris with one shot?

:) the idea of having this method for debris removal is actually an ACT one from Claudio Bombardelli (ACT RF in MAD https://en.wikipedia.org/wiki/Ion-beam_shepherd). This is just a technological device to implement it so that the system on board is simplified (i.e. instead of two engines, you get away with one and a weird nozzle) Marcus, you cannot align it to get rid of two debris as you need to keep the spacecraft close to the debris as this is a long duration acion. One of the two would drift away (can only follow one!)

I know you like these... [Edit] And by the way, this website contains also more practical stuff, like this

of course, "they" being Polish Wikipedia contributors who geo-tag like mad...

Have you had a look on Japan? It looks like they just geo-tagged all their train stations.

"Publishing industry exerts too much influence over what scientists choose to study, which is ultimately bad for science itself"

On a related topic - a nice read written in 1939 from Abraham Flexner the founder of Institute for Advanced Study in Princeton, home of some great minds on the "Usefulness of Useless Knowledge". Enjoy https://library.ias.edu/files/UsefulnessHarpers.pdf

This article is fantastic - starts already well with : "r IT not a curious fact that in a world steeped in irrational hatreds which threaten civilization itself, men and women-old and young-detach them-selves wholly or partly from the angry current of daily life to devote themselves to the cultivation ofbeauty, to the exten-sion ofknowledge, to the cure ofdisease, to the amelioration of suffering, just as though fanatics were not simultaneously engaged in spreading pain, ugliness, and suffering?" Could almost be written now

"Just as Pierre de Fermat scribbled his famous last theorem in the margins, professional scientists, academics and citizen scientists can annotate equations, figures and ideas and also write in the margins." Interesting way of analyzing research in the 21st century

it works! No, of course it doesn't. I've wasted your time just posting this.

One of the best use cases for MS Excel so far.

Prince Rupert's drops (PRDs), also known as Batavian tears, have been in existence since the early 17th century. They are made of a silicate glass of a high thermal expansion coefficient and have the shape of a tadpole. Typically, the diameter of the head of a PRD is in the range of 5-15 mm and that of the tail is 0.5 to 3.0 mm. PRDs have exceptional strength properties: the head of a PRD can withstand impact with a small hammer, or compression between tungsten carbide platens to high loads of ∼15 000 N, but the tail can be broken with just finger pressure leading to catastrophic disintegration of the PRD. We show here that the high strength of a PRD comes from large surface compressive stresses in the range of 400-700 MPa, determined using techniques of integrated photoelasticity. The surface compressive stresses can suppress Hertzian cone cracking during impact with a small hammer or compression between platens. Finally, it is argued that when the compressive force on a PRD is very high, plasticity in the PRD occurs, which leads to its eventual destruction with increasing load.

The best paper on transformation optics written ever :-)

Concerning the fabrication... as usual, no idea. I agree that this is the main drawback of MM, and certainly difficult to overcome. I would double check that number, because its value is related with the beam shape of Fig. 1 A. I believe that the simulations are correct, it's just a detail.

wow ... still publishing despite babysitting and new job!!

This has been around for over a year. The current trend in deep learning is "deeper is better". But a consequence of this is that for a given network depth, we can only feasibly evaluate a tiny fraction of the "search space" of NN architectures. The current approach to choosing a network architecture is to iteratively add more layers/units and keeping the architecture which gives an increase in the accuracy on some held-out data set i.e. we have the following information: {NN, accuracy}. Clearly, this process can be automated by using the accuracy as a 'signal' to a learning algorithm. The novelty in this work is they use reinforcement learning with a recurrent neural network controller which is trained by a policy gradient - a gradient-based method. Previously, evolutionary algorithms would typically be used. In summary, yes, the results are impressive - BUT this was only possible because they had access to Google's resources. An evolutionary approach would probably end up with the same architecture - it would just take longer. This is part of a broader research area in deep learning called 'meta-learning' which seeks to automate all aspects of neural network training.

Btw that techxplore article was cringing to read - if interested read this article instead: https://research.googleblog.com/2017/05/using-machine-learning-to-explore.html

A very nice expo currently showcased in Berlin, well worth a visit. https://www.koeniggalerie.com/exhibitions/38455/machine-hallucinations-nature-dreams/ The concept is quite cool and teh result just stunning (I recommend giving the youtube videos a good luck on your TV if you have teh chance)