Group items tagged

Francesco pointed out this research one year ago, we dropped it as noone was really considering it ... but in space a low CPU power consumption is crucial!! Maybe we should look back into this?

I don't remember what is the speed factor, but I guess this might do it! Although, I remember when using an IMU that you cannot have the data above a given rate (e.g. 20Hz even though the ADC samples the sensor at a little faster rate), so somehow it is not just the CPU that must be re-thought. When I say qualification I also imply the "hardened" phase.

I don't know if the (promised) one-order-of-magnitude improvements in power efficiency and performance are enough to justify looking into this. For once, it is not clear to me what embracing this technology would mean from an engineering point of view: does this technology need an entirely new software/hardware stack? If that were the case, in my opinion any potential benefit would be nullified. Also, is it realistic to build an entire self-sufficient chip on this technology? While the precision of floating point computations may be degraded and still be useful, how does all this play with integer arithmetic? Keep in mind that, e.g., in the Linux kernel code floating-point calculations are not even allowed/available... It is probably possible to integrate an "accelerated" low-accuracy floating-point unit together with a traditional CPU, but then again you have more implementation overhead creeping in. Finally, recent processors by Intel (e.g., the Atom) and especially ARM boast really low power-consumption levels, at the same time offering performance-boosting features such as multi-core and vectorization capabilities. Don't such efforts have more potential, if anything because of economical/industrial inertia?

from one of the reddit threads discussing this: "bit fishy, crazy if real". "Incredible claims: - Train only using about 10% of imagenet-12, i.e. around 120k images (i.e. they use 6k images per arm) - get to the same or better accuracy as the equivalent VGG net - Training is not via backprop but more simpler PCA + Sparsity regime (see section 4.1), shouldn't take more than 10 hours just on CPU probably "

clicking the link says the manuscript was withdrawn :))

This "one-shot learning" paper by Googe Deepmind also claims to be able to learn from very few training data. Thought it might be interesting for you guys: https://arxiv.org/pdf/1605.06065v1.pdf

and some more info (in German) http://www.pcgameshardware.de/aid,701186/Neuer-DKRZ-Supercomputer-Blizzard-berechnet-Klimamodelle-fuer-Kopenhagener-Weltklimakonferenz/CPU/News/

ZetaGrid is a platform independent grid system that uses idle CPU cycles from participating computers. Problem: The ZetaGrid activities must come to a final end! Now all services are down and this domain will be closed soon. The official last update note

NASA has it ... I WANT IT TOO!!!! 240 threads on 60 cores ... Imagine the possibilities of this new toy!! Francesco also has it in his new "kill the seals" job

true?

Isn't that why systems have a "wait for 15 minutes before trying again" after 3 or 5 wrong guesses? All the brute force in the word can't save you from real-life latency.

Oh, so you haven't heard about diceware yet? http://world.std.com/~reinhold/diceware.html And, of course, a related XKCD...

There you go, the latest-and-greatest chip is there. Now the only remaining tiny detail - program it.

Let's buy it first and we'll figure the rest out later :P

TensorFlow is an open source software library for numerical computation using data flow graphs. Nodes in the graph represent mathematical operations, while the graph edges represent the multidimensional data arrays (tensors) communicated between them. The flexible architecture allows you to deploy computation to one or more CPUs or GPUs in a desktop, server, or mobile device with a single API. TensorFlow was originally developed by researchers and engineers working on the Google Brain Team within Google's Machine Intelligence research organization for the purposes of conducting machine learning and deep neural networks research, but the system is general enough to be applicable in a wide variety of other domains as well.

And the interface even looks a bit less retarded than theano