Search
1000 results for “sparse_array”
-
The #Lehrmann family is reportedly prominent in the Southern US @martintheg
In Australia, I've seen reference to the #Tapscott family, who are influential in northern NSW and southern Qld. Evidence of that connection is increasingly hard to find, but the record hasn't been completely scrubbed. Bruce's own online presence is remarkably sparse.
https://independentaustralia.net/politics/politics-display/uncovering-the-protection-racket-behind-bruce-lehrmann-does-he-know-too-much,18534
It does seem that there were big plans for Bruce.
@Bot4Sale -
Update on my #os (tentatively codenamed #tyros) that I'm building in my spare time.
My first task is booting, and it's really made me appreciate how terrible everyone says the PC (x86) architecture is. I'd done a bit of 386 assembly before, so I knew it was a little goofy, but I was unprepared for the horrors of the actual architecture.
I actually got my first kernel booting after just a few hours. But, it was 16-bit real mode and relied on BIOS, so that was obviously a dead duck.
I got my second kernel booting after another couple days, which was transitioned (manually) into 32-bit protected mode. But the documentation for "long" mode (64-bit) was more sparse and more daunting, so I looked for another approach....
-
Update on my #os (tentatively codenamed #tyros) that I'm building in my spare time.
My first task is booting, and it's really made me appreciate how terrible everyone says the PC (x86) architecture is. I'd done a bit of 386 assembly before, so I knew it was a little goofy, but I was unprepared for the horrors of the actual architecture.
I actually got my first kernel booting after just a few hours. But, it was 16-bit real mode and relied on BIOS, so that was obviously a dead duck.
I got my second kernel booting after another couple days, which was transitioned (manually) into 32-bit protected mode. But the documentation for "long" mode (64-bit) was more sparse and more daunting, so I looked for another approach....
-
Time for a "Hot Off the Press" session!
1. Maintaining diversity of solutions in multi-objective problems
2. Using concepts from theory of evolution to improve Siamese networks performance in multi-class optimization
3. Find optimal recombination strategy when Genetic Programming chromosome feature combinations are sparse
4. Bi-objective optimization tracking by using wavelets and superpositions
5. Replacing population sampling to Hermite-Gaussian quadrature
40/🧵 #GECCO2023 #GECCO
-
🎧 Most mornings I take the road in silence till the engine hums its hymn and the day makes its slow demands. The sounds I carry come from folk such as Off Land, Radboud Mens, and Steve Swartz, who sent them like offerings, sparse and spectral, fit to score the long night that is Operation Watchtower. You can find their work on your favorite music streaming service.
Listen to the trailer here: https://youtu.be/woR_YNiEh1I
-
@angelo There were some big egos involved, or more charitably, at some point a formerly spectacularly good scientist stopped receiving actual feedback on the relevance and impact of the work of his lab members, and got detached from reality.
Was all too clear when a in a big neuroscience conference the then mature Blue Brain Project (precursor of the #HBP) got a whole wing of posters, all about circuit anatomy and function of ... made up neurons. I kid you not. Neurons made up from sparse, partial anatomical reconstructions of mammalian cortical neurons, from which their "statistics" got derived to, a la Monte Carlo-like data augmentation, generate a whole bunch of neurons and cortical circuits and simulate them. They called it a cortical column.
At least today such efforts could be guided by the MICrONs project volume and precursor volumes from Clay Reid's lab, and the barrel cortex column from the Helmstaedter lab. Back then it was all very, very speculative.
-
What's so difficult about reinforcement learning? Why is it so data inefficient?
One main issue is reward assignment. Classic reinforcement learning algorithms assign rewards backwards over sequential frames of experience. This model came from toy games like tik-tac-toe, where the frames correspond to discrete board states, where this process works reasonably well because the games are simple.
It starts failing with long-term dependencies where an action in the deeper past caused the reward or the penalty in the present, and the agent needs to return this reward across a huge number of frames and decisions, and they all confound together. The agent needs an insane number of repetitions for correlations to emerge from this backwards-exponential soup.
That is the root issue in reinforcement learning. Well, one of them. Another one is that the reward signal is information-poor and sparse, is often the sole source of world relevance information, and so doesn't correspond well with how animals learn.
The reward assignment becomes problematic already in games like chess and go which have long-term dependencies, but especially difficult in games like Starcraft or Counterstrike, and almost totally infeasible in the real world robotics applications.
But how do humans do it? How can humans learn tasks like bowling, with many many millisecond frames between the act of throwing the ball, and the result?
Humans don't model the world as movies composed out of sequential frames at all. They don't return reward across all those frames once the pins fall. They work by association. A human sees the pins fall, and then by association goes back to past decisions and thoughts associated with this result event. The reward isn't neutral, it has associative links to past key episodes. And the reward/penalty is trivially associated to these stored episodic memories.
Why is it so hard to do this correctly in reinforcement learning systems? Because they are stuck in dogma. Students are taught the Bellmann equations and that Sutton and Barto defined these problems like so, and are incapable of questioning these.
Instead of sequential, synchronous world state frames, we need to frame this problem as an associative problem, where the rewards are returned to the associated episodic memories.
-
@rachelwilliams, yes, the #DeepNeuralNetworks exhibit true #intuition and #creativity. However, the large amount of #compute required is because we are using traditional #computers which are #synchronous, #dense and #sequential to emulate these #NeuralNetworkArchitectures which are #asynchronous, #sparse and massively #parallel.
With proper #cores they should take much less power than the human #brain, which is 12 W. -
@rachelwilliams, yes, the #DeepNeuralNetworks exhibit true #intuition and #creativity. However, the large amount of #compute required is because we are using traditional #computers which are #synchronous, #dense and #sequential to emulate these #NeuralNetworkArchitectures which are #asynchronous, #sparse and massively #parallel.
With proper #cores they should take much less power than the human #brain, which is 12 W. -
@rachelwilliams, yes, the #DeepNeuralNetworks exhibit true #intuition and #creativity. However, the large amount of #compute required is because we are using traditional #computers which are #synchronous, #dense and #sequential to emulate these #NeuralNetworkArchitectures which are #asynchronous, #sparse and massively #parallel.
With proper #cores they should take much less power than the human #brain, which is 12 W.