tgflynn

72 ms·

721.
by tgflynn 10y ago
You may be correct technically. I'm not familiar with the formal definition of "confidence intervals" in statistics. Informally we often speak of confidence as an estimate of the conditional probability P( C | x ) where C i
722.
by tgflynn 10y ago
Isn't "confidence" measuring the internal state of the net I don't quite follow your meaning but no, that's not quite how I think about it. If x is the input image and C is the class selected by the classifier we
723.
by tgflynn 10y ago
This is great work for someone who's been dabbling with deep learning . I used to work on OCR systems for postal address recognition and for the most part this is how I would have approached the problem. A couple of suggestions: You
724.
by tgflynn 10y ago
Typically you can treat the output value as a probability. This is supposed to be the case with a softmax output layer (as is currently recommended for multiclass classifiers) and I believe it should also work with multiple logistic output
725.
by tgflynn 10y ago
I don't think they use git. I believe I've heard that they use an internally developed SCM.
726.
by tgflynn 11y ago
Saying that "statistics is the ultimate language of the brain" doesn't imply that the brain always performs statistical calculations correctly. In fact finding the true correct Bayesian answer is probably computationally inf
727.
by tgflynn 11y ago
As far as I can tell LSTM's are still using a combination of logistic and tanh transfer functions, at least through 2014 and some of the most impressive deep learning results have been obtained with these. For example: http:/&#x2
728.
by tgflynn 11y ago
I never heard of the BB problem before but according to Wikipedia it requires an N-state Turing Machine so I don't see how that's relevant to the question I posed. Could you elaborate ?
729.
by tgflynn 11y ago
This is a great question. What do we lose by thinking of real computers as Turing machines when they are in fact finite ? For one thing I believe the halting-problem doesn't exist in reality because it only holds for infinite systems
730.
by tgflynn 11y ago
You can get (lossy) compression by using an autoencoder with a hidden layer that has fewer nodes than the number of inputs, since then you only need to store the hidden layer excitations for each example. The weights and network architectu
731.
by tgflynn 11y ago
It depends on who you read. LeCun certainly advocated tanh for a long time. ReLU seems to be currently favored, especially for deep nets but I'm not sure it's being used universally. In my own experiments I am seeing it train f
732.
by tgflynn 11y ago
Great article, I'm a little skeptical about the sourcing though. Most of the information seems to come from general literature (as opposed to something more objective, like say taxation records). What people choose to write about isn
733.
by tgflynn 11y ago
Do we have good information on what the crop distributions were in Ireland prior to 1492 (I assume that's how they define "pre-colonial") ? I wonder how this could really be determined anywhere. I wouldn't think there
734.
by tgflynn 11y ago
And why is most of Europe except Ireland and Spain left out of the maps ? That alone suggests they could be cherry-picking the data.
735.
by tgflynn 11y ago
If by consciousness you mean the property of having subjective experiences I don't think that is something that you will ever "understand" through a computer simulation. Personally I do not believe that a computer simulation
736.
by tgflynn 11y ago
I don't know about software designed specifically for doing this with legal documents but most of this you could probably do quite easily with some simple Unix tools and a little scripting. I'm guessing the documents aren't
737.
by tgflynn 11y ago
It seems to me that the question of how to finance a UBI is probably the largest stumbling block. Of course it could be done through taxes but a lot of people have legitimate concerns about government taxation. For taxation to work you nee
738.
by tgflynn 11y ago
I have no idea what he's on about here, but just by making this kind of grandiose claim he makes me skeptical. I think he's referring to the free software movement. EDIT: Or maybe the open source movement, I don't recall wh
739.
by tgflynn 11y ago
Do you have evidence that the claims he made here are inaccurate ? I don't necessarily think accurately presenting one's accomplishments for the purpose of explaining the life basis for one's views constitutes arrogance, and
740.
by tgflynn 11y ago
The 1 T+ field your thinking of is a static magnetic field. Static magnetic fields don't induce currents (except in rapidly moving objects) and have no known effects on the brain or biology (at least for the field strengths available
741.
by tgflynn 11y ago
The diagnosis of Asperger's was removed in the 2013 fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) and people with these symptoms are now included within the autism spectrum disorder along with auti
742.
by tgflynn 11y ago
Yes, I agree with what esr says here but I would add along the same lines that it seems like a necessary condition for success in difficult fields is devoting a great deal of time and energy to that field and the problems one is working on.
743.
by tgflynn 11y ago
I have no concerns about cell phone radiation but I'd just like to point out that the high carrier frequencies don't necessarily preclude effects on low frequency systems. The modulation envelope probably extends down quite close
744.
by tgflynn 11y ago
I'm amazed by the improbability of this guy's life. He had an abusive father and a schizophrenic mother dropped out of school at 15 then worked as a sound engineer for a famous rock band, built a successful high-end auto repair b
745.
by tgflynn 11y ago
I think it should be clear from the context that this discussion was about macroscopic objects. Cosmic rays don't carry the kinetic energy equivalent of a medium size nuke.
746.
by tgflynn 11y ago
I don't think you'll find 0.2c relative velocities except near very energetic phenomena, especially not of things that are too small to easily see coming.
747.
by tgflynn 11y ago
There are some pretty far out concepts here 300 GW for 30 minutes, that's the output of 300 nuclear power reactors or the energy equivalent of a 100 kT nuclear explosion. Focussed onto a sail that's 100's of atoms thick to g
748.
Breakthrough Starshot: Mission to Alpha Centauri(centauri-dreams.org)
2 points by tgflynn 11y ago|1 comments
749.
by tgflynn 11y ago
In the day-to-day life you experience a boundary between that which we consider 'self' and that which we consider 'other'. That boundary gives the environment shape, defined by what you know to be you. To say the same
750.
by tgflynn 11y ago
Several years ago I worked for a very large company in a rather small group (about 40 people) that worked on what were then challenging image recognition problems. I think that most work in these areas is done either in large companies lik

More ›