« first day (3193 days earlier)      last day (451 days later) » 

1:04 AM
@JohnRennie have you ever work as a Scientific American Chief Editor?
worked*
 
1:54 AM
Hey there!
Is an affine parametrization only defined for geodesics? or also for any arbitrary curve in spacetime?
 
@eigenvalue usually affine parametrization means a parameter such that the geodesic equation holds
and it's a theorem that for spacelike and timelike geodesics, it's always a linear function of the "arclength"
 
@RyanUnger I know why it makes sense to use affine parameters for geodesic curves. But I am wondering if there is an "affine parametrization" for random curves, too. The reason I am asking is the statement by J. Been in his book saying "the concept of an affine parameter is extended from geodesics to all C^1 curves".
PS: this is done regarding the introduction/definition for "generalized affine parameters"
 
2:09 AM
If the plates of a capacitor are joined together by a conducting wire, then its capacitance will become infinity as distance between the plates reach 0 or will it become 0 because it can no longer hold charge?
 
2:37 AM
@eigenvalue well unless he says otherwise I'd guess he just means any parameter that's a linear function of arc length
 
 
2 hours later…
4:12 AM
@M.N.Raia No, that's a different and unrelated John Rennie.
 
4:32 AM
@JohnRennie Oh, I see, beg your pardon then. The reason for that questions is that I was looking for better explanations of Warp Drive spacetimes and then your name poped up kkkk. But the expecific question (if you want to look) about Warp Drive spacetime is:
2
Q: Question about the travel time of a ship "using" a warp drive metric

M.N.RaiaI - The Warp Drive metric: The Warp Drive is a geometry in a spacetime $(\mathcal{M},g)$ given (in geometrized coordinates $c=G=1$) by the following metric tensor: $$ ds^{2} = -dt^{2}+ (dx-v_{s}f(r_{s})dt)^{2}+dx^{2}+dy^{2}+dz^{2} \tag{1}$$ The basic interpretation is: suppose that we have a ...

But another point of physics which I struggling is given by the paper wrote by M J Pfenning and L H Ford "unphysical nature of warp drives". It seems that the generation of the warp bubble is much like to "enter a black hole". Concerning this another question is: to use a warp drive an spaceship which generates the curved metric must to pass the outer wall<inner wall and the reach the center of the bubble?
 
 
1 hour later…
5:53 AM
Where does Ampere's law actually derive from? Wikipedia briefly mentions that Maxwell derived it using principles of fluid dynamics, but that's so vague that it doesn't even help me find it in his paper.
Does anyone know if Griffiths or Jackson (or any other book) has an explanation of this? Or is it just an entirely empirically justified law like Faraday's?
 
Ampere's law was originally an experimental observation
But of course it is a consequence of Maxwell's equations.
So ultimately I guess you're asking how Maxwell's equations were derived?
Actually someone just asked a question about that ...
 
I can't find anything about Ampere's experiment then, if that's who did it
 
2
Q: Is there any particular reason why Faraday's and Ampère's laws are valid?

DarudeSamstormI know that the Maxwell equations are usually the explanation for all electromagnetic phenomena, but I would like to know why those are valid, if there is any reason for them.

 
I guess I'm kind of asking two questions, but primarily the one you just mentioned
@JohnRennie OK, that helps.
 
Maxwell's equations are remarkably simple when you simply require that they respect special relativity.
They more or less just fall into your hands.
 
6:54 AM
user image
4
 
 
2 hours later…
8:50 AM
I have a question of which I'm not sure it warrants opening one on the main page. It's about the Josephson effect. I've learned two interpretations of it; one is that the superconductors on either side of the junction have a finite overlap between their wavefunctions, allowing Cooper pairs to tunnel. The other is that Andreev bound states allow transferring Cooper pairs across the junction depending on the transparency of the junction and the number of channels.
Is the ABS picture simply a more refined one of the first one, or are they separate processes?
Direct tunneling, vs ABS mediated
 
 
2 hours later…
10:25 AM
If I connect two capacitor with capacitance C1 and C2 with charge Q1 and Q2, with opposite plates joined together will total capacitence be $ C_1+C_2 $ or $ \dfrac{1}{\dfrac{1}{C_1} + \dfrac{1}{c_2} } $
@JohnRennie
I think it should be the latter as they are in series but the answer is the former.
 
It isn't clear what you mean by opposite plates joined together. Have you got a diagram?
 
like opposite terminal of each capacitor connected.
 
So they are being joined in parallel?
 
No
@JohnRennie
 
I'll have to get back to you. I'm busy for a few minutes.
 
10:36 AM
Its okay
 
10:48 AM
@user541396 I'd have to see the question as it isn't clear to me what is being asked. It looks like two capacitors in series, in which case the answer is $1/C = 1/C_1 + 1/C_2$
 
 
2 hours later…
1:00 PM
@EmilioPisanty That's good news for Physics.SE, but I hope he doesn't revert to pestering us in Astronomy again.
 
@PM2Ring note that there is now a well established precedent that site members can be suspended for repeated low quality posts.
 
@RyanUnger In the dupe target there's a link to a good article by Greg Egan. He cheats a bit, by just examining a Rindler horizon in flat spacetime, but it's still worth reading, IMHO.
 
@PM2Ring the Rindler horizon is an excellent way to understand what a coordinate singularity like the event horizon looks like. One day I must get round to writing an article explained the relationshp between the two.
 
@JohnRennie Ok, but it's not just low quality. It's persistent deliberate posting of non-orthodox material, even after being told numerous times that his opinions clash with the mainstream understanding of general relativity.
 
@PM2Ring Then petition the mods to suspend him.
Refer them to us for a demonstration that it can be done.
 
1:09 PM
@JohnRennie Oh, I agree! And why bother with the complications of curved spacetime when it's easy to show the desired effects with a Rindler horizon?
 
ok my question seems rather naive now but I'm still having a hard time imagining what this actually looks like from my perspective
 
@JohnRennie wtf
what promotional content
 
@JohnRennie Excellent.
 
1:18 PM
@RyanUnger I don't know. I don't hang out in the Astronomy SE so I don't know what went down there.
 
@RyanUnger Probably something to do with his stupid idiosyncratic version of GR, or as I've previously called it, bozonic gravity theory.
 
@PM2Ring you mean Einstein's version of GR.
 
Einstein himself wasn't without his blind spots.
He refused to accept the validity of the Gullstrand-Painleve coordinates for a while, even though they are a simple coordinate transformation.
 
@RyanUnger JD's version of Einstein's version. :) But as John Rennie just said, even old Albert himself didn't have a flawless grasp of GR. But that's quite understandable. It takes a while for new paradigms to properly settle into the landscape.
 
1:33 PM
Ironically Einstein died just a few years before the big revival of interest in GR. By his death it had become a stagnant area.
 
And while Einstein surely had some great physics intuition he was by no means perfect. His antagonistic attitude towards QM may have been somewhat useful in the very early days, keeping guys like Bohr on their toes, but in later years it effectively sidelined Einstein. But I guess that didn't bother him much, since he was focusing on trying to unify gravity with the other forces...
 
My own theory, for which I have absolutely no proof, is that Einstein suffered a crisis of confidence as theoretical physics got too mathematical for him, and he deliberately isolated himself by working in an area where there was no competition.
 
2:00 PM
Perhaps. Let's not forget that he introduced some pretty far-out mathematics into physics with GR. Although he wouldn't have been able to do that if not for the help of Minkowski.
But I guess the kind of maths needed for QM & particle physics is of a different character. Eg, someone might find differential geometry quite natural, but have difficulties with group theory stuff.
 
@JohnRennie that's a bold claim
especially because physics didn't get particularly mathematical until much later
 
@RyanUnger I did say it's just a personal theory.
 
You know
Duffield is pretty old
do you think he was born before Einstien died
 
2:16 PM
I've read a number of biographies of Einstein and it's the impression I get.
 
Or maybe... Just after???
Perhaps he is his reincarnation
2
 
I think he managed to get a good grasp of differential geometry, mostly learned from Marcel Grossmann, but I think his maths pretty much stopped there.
 
@JohnRennie well not to sound too elitist (or maybe this is just ignorance), but what more was there to physics in the 40s
 
His post GR work seems to be trying to apply the same maths in various different ways without any clear strategy.
 
quantum mechanics was very rough mathematically speaking
the big breakthroughs in mathematical physics started in the 60s
 
2:20 PM
Quantum field theory started very early on. Remarkably early really. OK it was in a crude state before renormalisation was understood, but the basic ideas were there.
 
cool
I was very confused because I was reading [18] first
@JohnRennie but that's not mathematically sophisticated is it
 
@RyanUnger I think you underestimate the sophistication because you have a good grasp of the maths involved. It seems hard to me. I'm not claiming to be ona par with Einstein, but I suspect it would have seemed hard to him as well.
 
@JohnRennie @PM2Ring or, more precisely, document the full pattern of low-quality contributions, via flags.
 
@EmilioPisanty he's banned for a year anyway so it doesn't matter for now.
 
The text here is largely recycled from your website, meaning you're still looking for ways to use this site to get more eyes on your content. That is not a good faith use of this site. — called2voyage ♦ Jul 1 at 19:57
seems to be this
 
2:28 PM
well if his point is on his website why should he not just copy paste
the answer is not just a straight link
 
@RyanUnger I'm documenting, not endorsing
 
I suspect the reason posted isn't that significant. The real reason is his gift for making himself loved wherever he goes.
 
@EmilioPisanty I know
 
... but, that said, if the point is on External Website A and External Website A is just pseudoscience, then copy-pasting from it is still not appropriate, and it does have the promotional-content aggravant
 
If you're feeling unusually bored try Googling for his name along with his alter ego farsight. You'll find complaints about him going back years on all sorts of different forums.
 
2:30 PM
uhhh
I'll stick to figuring out this boundary condition thanks
 
:-)
I wonder how many SEs he's managed to get himself banned on.
 
I know ACM is in Wacken; however, that doesn't mean you should discuss people who aren't here to defend themselves.
 
@JohnRennie holy coincidences, Batman, look at what the earliest result turns up
 
loooooooooooooooool
 
another PSE favourite
 
2:32 PM
parents are away
@Loong is the babysitter
2
 
@Loong but Loooong is right
 
freezes the room until Wacken is finished
 
> Sometimes it feels like I'm witnessing the growth of cargo-cult science.
yeah it's him alright
 
vzn
re the einstein topic... think he had very deep intuition alongside his math prowess that few physicists had. he seemed to anticipate some kuhnian ideas even as living them out/ epitomizing them himself. yes he became an isolated figure at end of his life. there are rare cases where scientists are vindicated long after they die & think einstein will eventually be put in this category along with eg Bohm. there are further surprises in store in 21st century physics, its not "finished" yet...
 
Einstein's physical insight was remarkable, and I think it's that which singles him out as one of the all time great scientists.
The irony is that he abandoned physical insight as a means to make progress after GR.
 
vzn
2:41 PM
also, einstein was on the trail of some particular realist-flavored ideas that are starting to actually play out for those attuned to subtle shifts in overall physics directions. it seems that physics moved on without him without examining his ideas that carefully, and suspect mainstream physicists are not exactly aware of the "leads" he was pursuing. it looks like some has been overlooked/ forgotten in the sands of time...
 
Have you read, for example, Abraham Pais' bio of Einstein? It's hard to believe Einstein was going anywhere interesting in his later years.
But I think we've had this conversation before.
 
vzn
have read some books on the topic. einstein had a close personal relationship with Godel at the end of his life. at the very end it looks like einstein wasnt working on much at all. his groundbreaking appointment to the IAS was something of a bust wrt "productive output/ discoveries" so to speak.
 
As it happens Einstein did important work in fluid mechanics, though amazingly he failed to realise GR was just a hydrodynamic theory.
I recall learning Einstein's model for the viscosity of colloidal suspension.
 
vzn
it seems that few people consider GR a "hydrodynamic theory" but have pointed that out at length/ repeatedly myself to great pushback around here.
 
oh my god
stop talking about fluid mechanics vzn
 
vzn
2:47 PM
lol was triggered cant help it :P (speaking of "pushback"...) :P
 
The point is that a mathematical model is just a mathematical model. It stands or falls by its predictions, and if it makes the correct predictions it's a good model.
It's tempting to take the model as illustrating some fundamental reality, but that isn't the job of a model.
 
vzn
einstein said differently in ways than what you are asserting. some of this cuts to philosophy of science which he had strong expressed opinions on. yes, some of his ideas are now regarded as minority-to-fringe pov(s).
 
If you formulate a model based on hydrodynamics that gives the same predictions then it's a perfectly good model. Does that mean the underlying reality is hydrodynamic? Well maybe, but just because the model works doesn't mean it reflects any reality.
 
vzn
the hydrodynamic model is seeing a resurgence and there will be new interpretations/ interpretations, theyre already starting to appear/ emerge. new physics isnt built in a day. lets recall the kuhnian like quote of Planck re new theories, do you recall?
 
2:51 PM
If you formulated a model based on hydrodynamics and it worked better e.g. allowed calculations in regimes where GR is too complicated to calculate then it would be a better model. If it doesn't then it's just another model.
 
> This paper has been withdrawn by the author due to a crucial error in one of references
oh no
 
@vzn so what you're saying is that it's a model that currently doesn't do as well as GR but it might do one day?
That's not a great advertisment for it.
 
vzn
> !!! This paper has been withdrawn by the author due to a crucial error in one of references
 
not hard to figure out which one
 
vzn
new models are created all the time in physics to replace old models. but its a (usually) gradual process that some dont recognize much in their own lifetimes. it helps to study the basic history of the field. it helps to realize everything taught in college textbooks can and does change over time.
 
2:55 PM
@vzn it's exceedingly rare that a fundamentally new model displaces an established model. Generally new models are developments of existing models not complete reformulations.
 
vzn
its rare but recurring in the history of science/ physics and the process doesnt end as long as science doesnt end.
 
And the old models usually remain accurate within their domain of applicability. We don't stop teaching Newtonian mechanics because we know it fails at velocities near $c$.
Most of what is written in old textbooks is still perfectly applicable and indeed still taught to first year undergrads.
 
vzn
yet, revolutions happen. they are exciting when lived thru/ recognized :)
 
If there is a revolution in cosmology due my guess is that it will be when we realise GR is an emergent theory i.e. a limit of some deeper underlying theory.
But it's hard to see that a hydrodynamic approach is going to offer that.
More likely a hydrodynamic model would also be an emergent theory, just as regular hydrodynamics emerges from interactions between atoms and molecules.
 
vzn
re new models of hydrodynamics for GR: honestly am a bit shocked you bring up the analogy, dont recall seeing you do so anywhere before. lets agree there are deep mysteries in GR wrt eg black hole behavior and wouldnt it be a bit naive to think that standard modern theory will calculate the correct answers?
 
3:00 PM
> deep mysteries in GR wrt eg black hole behavior
What deep mysteries in black hole behaviour?
I am unaware of any such deep mysteries in the way GR describes black holes.
 
vzn
it seems that few physicists would argue we completely understand black hole dynamics and there is much current research on the topic to work out the exact details. yes the general outlines are well known. same with big bang. and dark matter + energy are widely recognized massive open questions/ problems of the field that current theoretical frameworks seem incapable of answering. add GR + QM unification. these are not controversial statements.
 
Heyo
 
@vzn Quantum gravity is certainly not understood at the moment. But in the classical regime I can't think of any deep mysteries?
If you can point to a specidic example I'd be interested.
 
vzn
lol by "classical regime" do you mean newtonian physics? :P
 
@vzn I mean non-quantum
:51236746 it depends exactly what you mean by boiling
 
vzn
3:07 PM
theres a ton of fascinating black hole research going on. why dont you ever cite any of it? not your thing apparently
 
They've not specified in atkins
Just written boil
How long will it take to start to boil.
@JohnRennie I thought there was a scientific convention as to what's boiling
 
@vzn There's a lot of numerical work, driven by applications like LIGO. Can you point to fascinating theoretical work in black hole research.
 
But I couldn't find it on googling
That's why came here
 
@AvnishKabaj I would say boiling starts when the temperature has reached 100°C.
 
@JohnRennie thanks
 
vzn
3:09 PM
@JohnRennie its widespread & feel you will just dismiss anything cited. have cited stuff in here on occasion.
 
At this point any further heat added to the liquid causes evaporation instead of an increase in temperature.
 
Yup
Thought the same
Thanks
 
@AvnishKabaj what I was being careful about is some people take boiling to mean bubbling i.e. nucleation of bubbles.
 
@JohnRennie Ah
I thought that bubbling was boiling
So the water doesn't bubble at 100°C
?
 
Generation of bubbles usually takes temperatures a few degrees above 100°C because there's a barrier to nucleation of bubbles.
 
3:12 PM
Hmm
Nice
 
Suppose you heat water to 100°C and then add extra heat slowly enough that the rate of surface evaporation matches the rate of heat input. Isn't that boiling? Heat is being added and it's producing steam. But there aren't any bubbles.
 
So there is a definition
;)
 
What happens in a real pan when you turn the gas up is that the water at the bottom of the pan gets superheated and that causes the bubbling.
 
But I'd guess a thermodynamics textbook will mean boiling to be when the liquid reaches the boiling point and extra heat causes the liquid-vapour transition.
 
3:17 PM
Yeah
 
3:30 PM
@JohnRennie To be fair, I prefer using "boil" to refer to actual bubbling evaporation; and just use evaporation in the general sense. The word "boil" is pretty well associated with bubbles (like the noun form).
but nowadays it can just mean regular evaporation; which kinda kills off some of it's utility
 
@JMac that's exactly why I was concerned to find out exactly what was being asked!
 
It kinda sucks when terms like that become too general IMO
 
In thermodynamics I think boiling is normlly taken to mean vapour production at constant temperature.
 
I guess that's less wordy; but not very intuitive IMO
 
Vapour production happens at all temperatures of course. What is special about boiling is the fact that adding heat does not increase the temperature.
 
3:40 PM
Has anyone seen Novalium around since he last said he was going to start experimenting with that e-paper stuff?
 
@JMac perhaps we should have warned him the solution would go critical and produce a 10 megaton blast ...
 
@JohnRennie What's a little home experiment if you don't accidentally create a runaway fusion reaction in your basement?
 
I remember when the Pons/Fleischmann cold fusion fuss kicked off people were saying be careful when you do the experiment as the amount of energy released can be dangerous.
 
4:30 PM
@JohnRennie
Are you free
?
 
@AvnishKabaj hi
 
So in a reversible adiabatic expansion of a perfect gas
Can we comment on the sign of enthalpy
∆H= ∆U +∆(PV)
∆U will be negative
I'm using the chemistry conventions
U= Q+w
Commenting on ∆(PV) on the other hand
I thought about it since it's recersible
 
9
Q: Why is the enthalpy change not zero in an adiabatic process?

halcyon Which of the following must be true for adiabatic processes? $C_\mathrm{V} = C_\mathrm{p}$ $\Delta H = 0$ $\Delta U = 0$ $\Delta S = 0$ $q = 0$ (Source: Chemistry GRE) The answer is $q = 0$. From what I can find, an adiabatic process is when there is no transfer of heat,...

That gives a really nice description
 
Nice
That solves my problem
Forgot about googling it first
@JohnRennie
I think it's wrong as
We can't take work done as p∆v
In a reversible adiabatic expansion
Pressure is not constant
 
If you make the changes infinitesimal you're going to end up with $dH = VdP$, then $\Delta H = \int V dP$.
So in general the end result is going to be similar.
 
4:41 PM
Ah ok thanks
 
5:05 PM
@vzn deservedly great pushback ;)
 
are there like
Isometry groups for Hilbert spaces
I guess it's just the set of unitary operators
@RyanUnger there was plenty of QM math in the 30's and such
 
e.g. the gruppenpest
 
Weyl and Wigner and whatnot
 
The group theory in Weyl's QM book (from 1931) is too hard
 
there was that Einstein quote
About how Einstein didn't understand weyl
 
5:15 PM
group theory is in general too hard and completely avoidable
 
5:29 PM
To my shame
I still don't know what a central extension is
 
5:42 PM
reddit.com/r/IAmA/comments/cl4boe/… This is hilarious IMO. Guy is trying to promote his "algorithm" for sustainable energy; when it just looks like he did basic energy calculations like everyone else, and didn't even get it peer reviewed.
 
> Link to the fourth revision of paper I wrote on SEGSs. The paper was written for people with little to no knowledge of power systems.
I guess that should read "written by people".
 
 
1 hour later…
6:50 PM
Good evening everybody. Is there somebody that can help me for this question? Thank you very much.
0
Q: Vectorial doubts from the Maxwell-Proca's equations

SebastianoStarting from two equations: $$ \mathbf{E}=-\boldsymbol{\nabla} \varphi-\frac{\partial\mathbf{A}}{\partial t}, \qquad \mathbf{B} = \boldsymbol{\nabla} \times \mathbf{A}$$ and are valid the Maxwell-Proca's equations: \begin{equation} \begin{cases} \boldsymbol{\nabla} \cdot \mathbf{B}=0\\ \bold...

 
what is $\nabla^2$
 
@RyanUnger I have not understand: it is $\nabla \cdot \nabla$.
 
do you mean the Laplacian?
 
that's pretty common notation, including for the vector Laplacian
 
@RyanUnger It is = laplacian. Yes of course.
 
6:55 PM
@Semiclassical maybe in physics
 
hi, semiclassical user
 
whoever wrote the wiki article was very confused
 
well, we're in the physics room talking about a question about maxwell-proca equations
 
they switch from line to line
 
sooooo
that, i can believe
 
6:55 PM
lol
 
hah
the notation on the Vector Laplacian page is more consistent, though it's a shorter article
 
oh hai schrodinger
 
Is it correct in Time-dependent Schrödinger wave equation?
 
well, the term in red brackets is the Laplacian of $\Psi$
so $\Delta \Psi$
 
7:00 PM
In fact.Correct!!!! :-) But my question is very ugly or not clear?
 
what?
 
0
Q: Vectorial doubts from the Maxwell-Proca's equations

SebastianoStarting from two equations: $$ \mathbf{E}=-\boldsymbol{\nabla} \varphi-\frac{\partial\mathbf{A}}{\partial t}, \qquad \mathbf{B} = \boldsymbol{\nabla} \times \mathbf{A}$$ and are valid the Maxwell-Proca's equations: \begin{equation} \begin{cases} \boldsymbol{\nabla} \cdot \mathbf{B}=0\\ \bold...

 
as Ted would say, this is probably easy if you write it out in terms of differential forms
 
eh, it's not hard regardless if you remember what $\nabla^2 \vec{E}$ actually means in terms of div/grad/curl
What seems odd in the question is that you've got $\varphi=0$
 
@RyanUnger I'm agree with you.
 
7:07 PM
also $\nabla^2 \vec{E}=-\mu_\gamma^2 \varphi$ is not sensible. The left-hand side is a vector and the right-hand side is a scalar
 
Is there any mistakes?
 
Well, given that that expression makes no sense (like I just said) I would have to say so
 
Fine. That seems plausible enough. That's not what I'm objecting to.
 
Are correct these equations? I think yes.
 
7:10 PM
Your question asks "Why $\nabla^2 \vec{E} ≡−\mu_\gamma^2 \varphi=0$, and $\mu^2_\gamma \nabla \varphi=0$?"
I don't know about the second one, but the first is certainly not true
At least the first equality is not
$\nabla^2 \vec{E}$ is a vector function. There's no way it's going to be equivalent to a scalar function.
 
Also, you don't list the Coulomb gauge condition in your question.
 
Peraphs the first div A =0 is wrong. The 2nd image is correct.
 
div A is the Couloumb gauge
 
^ Right.
Regardless, though, I am not objecting to your quotation of the Maxwell-Proca equations. Those seem fine.
What I do think is that the "Why" question you ask at the end seems highly dubious
For one, it'd require that $\varphi=0$. And it seems quite unlikely that the the scalar potential would vanish identically.
And if it's not zero, then you're proposing that a vector function be equivalent to a scalar function. That, quite simply, has no hope of being true.
 
7:18 PM
I add an screenshot of my notes with my steps.
 
That really adds nothing. My point, again, is that I don't believe your claim is true.
I don't think that the last line follows from the rest.
 
@Semiclassical I wrote down everything I did based on notes I'm rebuilding. If you have the possibility, could you add an answer to me? At the moment I have to leave the station and I hope you will excuse me. Thank you for your interest. See you later. Greetings.
 
 
1 hour later…
8:36 PM
@bdegnan is bypassing the Von-Neumann Bottleneck a major benefit of using Neuromorphic chips? I see some papers mention it while others don't...
 
 
2 hours later…
10:45 PM
@enumaris I would say yes. The neuromorphic architectures are inherently parallel. An example of this would memory retrieval in humans, it takes milliseconds to search, but then the retrieval is Petabytes of equivalent data.
Most CPUs are inherently serial instructions
 
I see...
GPUs are massively parallel too though...I don't really know if the shared memory bus architecture is applied there as well...I should look this up...
 
GPUs are massively parallel, but they take a lot of power. most GPUs have a "deep sleep" cycle as they are designed for graphics, and effectively produce data at 60Hz. If you want to keep them running, you need to take turns with the cores.
The brain works on about 4W.
 
I see..
fascinating stuff :P
 
A silicon neuron is somewhere between 20fA and 10nA at 1v
Biology is crazy
 
hmmm
I get to chat with a UCI researcher in neuromorphic computing next week, that should be interesting as well :D
Do you think the major challenge of neuromorphic computing is the lack of a "best" learning algorithm for spiking neural nets?
 
10:52 PM
I have a key word recognition circuit that works on 2nW. that's the measured current. "Hey Siri" takes... who knows how much power when you include the network and everything else.
 
o.o
 
Honestly, I do not know. In the analog networks we used, there was only one way to "learn". The STDP algorithms were tied to the physics of the injection of the electrons on the floating-gates.
 
I see
But STDP is unsupervised learning right? How did you include the supervised part - to actually tell the network "this is the word 'siri'" or some such?
 
The key word circuit works by saying "some word" into it, and then it uses Vector Matrix Multiplies and Winner Take All's that are matched to the word
It works for anyone with the same accent.
 
Ah...
 
10:55 PM
Ah, so, that is based on a simple FFT in my case
It's basically an energy detector.
I have a MATLAB program that takes an input and hten generates the initial weights.
VMMs are different than silicon neurons, but solve the same problem.
BTW, I have found that neural networks failed about 10% of the time for the "rat in a maze" scenario. There's a lot of open research. We know barely anything about the brain
 
o.o
So the basic idea is look at neuronal activations when different things are being said, and if activations for neurons associated with the keyword shows up you kinda infer that the keyword was said?
so basically the STDP gives you clustering of "utterances" and then you just check which "cluster" your utterance most likely belongs to at inference time?
 
yes. BTW, excitation neurons are often fewer than inhibition neurons in these systems; however, in biology, it seems to be opposite. I only mention that because although the systems seem to operate on the generally same premise, they do seem to be a tad bit different.
Prof. Hasler is currently working through this with the neuroFPAAs and rats. I don't know the details.
I hope to see her next week an catchup.
 
what's a excitation vs inhibition neuron? everything I've read seems to imply just 1 type of neuron
generally an integrate-and-fire (or some leaky version)
 
yeah, so, in the networks, there are some that apply excitation to the next dendrite input, and some that calm the input. It keeps your system from getting a feed-forward loop. biology seems to do that too
so, it's more than just charge leaking off.
 
Interesting...I haven't come across that mechanism in my readings so far
 
11:09 PM
Something that is very excited (like a sensor) will have the it's behavior nulled a bit
 
maybe that can be implemented as a negative weight in what I read...
 
I assure you, this behavior is there. :)
 
I think that's kinda equivalent to having a negative weight on that synapse...hmmm but I guess I just assumed the weights were positive due to my internal conception of the network
yeah I think everything I read just assumes you can have negative weights represent the inhibatory synapses and the positive weights are excitory...but I just assumed everything was positive lol
 
On of the ways to do "negative weights" was just have leak paths. The neuMOS work by SHIBATA and OHMI in the 80s/90s did that. I have no idea when the idea of a negative weight was added instead of just leakage. Sarpeshkar's work at CalTech and Bohem (currently active in this field at Standford) both used negative weights.
If you are geographically located in the USA, there's a few lectures you probably could sneak into
 
You think there's lectures on Neuromorphic computing happening at CalTech or something?
I am close enough to CalTech that I could probably attend a lecture or two there if necessary
I'm closest to UCI
maybe there are some seminars or something...I'll take a look XD
 
11:23 PM
Carver stopped doing lectures. Honestly, I'm not really up who's doing what at CalTech recently.
I'm pretty much east coast or asia
 
Ah I see
I enrolled in an online computational neuroscience course given by coursera lol
we'll see if there's interesting info there
next week I'll look to see if there's some seminars or something I can attend around here
 
11:53 PM
hm that course on coursera looks neat...I can halfway finish it just like I do with all my coursera attempts
 

« first day (3193 days earlier)      last day (451 days later) »