kastnerkyle

91 ms·

181.
by kastnerkyle 12y ago
Excellent discussion - and the application area (cross-correlation of sensor network data) is actually quite a bit more popular than one might think! Avalanche/landslide detection, earthquakes, bridge monitoring, industrial settings...
182.
by kastnerkyle 12y ago
Google search is the ultimate example (IMO) of an effective, simple, and powerful Markov chain algorithm. For anyone interested in this stuff, check out "The $25 Billion Dollar Eigenvector"[0]. I did a simple page rank experiment
183.
by kastnerkyle 12y ago
A. Krizhevsky put out a pretty interesting paper recently on this same topic. Also, it has the best research paper title I have seen [1] By changing the type of parallelism depending on where in the network you are (fully connected vs. conv
184.
by kastnerkyle 12y ago
For anyone who is interested, there is pretty good writeup here [1] on how to do kernel builds for this card. I also have a WIP of crosscompiling for this card as part of a wifi connectable (using waypoints) automated quadcopter, though I p
185.
by kastnerkyle 13y ago
I strongly disagree! Text to speech needs the same data as speech to text - a well annotated collection of raw, single speaker speech data from a variety of speakers and accompanying text labels. It is very, very difficult to find a large
186.
by kastnerkyle 13y ago
From the source itself: http://people.csail.mit.edu/mrub/vidmag/ Patent pending... Fun to play with, but I am disappointed that they patented it.
187.
by kastnerkyle 13y ago
Nice, simple explanation of this concept. +1000 for using IPython notebook, though the left-right layout of the cells is a little hard to follow. Shamelessly, I prefer my own layout based on pelican-bootstrap: http://kastnerkyle.
188.
by kastnerkyle 13y ago
I found some related papers for others who are interested: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.56.... http://bmi.osu.edu/resources/techreports/ahmedBokhariV21.pdf Also a
189.
by kastnerkyle 13y ago
So... duh. You would need > 1.2 TB to fit all of ImageNet on the card :). Thanks for clarifying, and pardon my brain lapse! Also, thanks for putting this out there - if I get some time I may send some pull requests your way. Awesome stuf
190.
by kastnerkyle 13y ago
So you are storing the entire dataset on the GPU then, to speed up processing? Or is there support for a "minibatch" mode that sends chunks at a time to be processed? Unless this more an issue with the model size of the "Kriz
191.
by kastnerkyle 13y ago
I think this is a great thing to put out into the community, and I hope it gains widespread adoption. I do have a few things to comment on here: The architectures themselves have never been closed source (hence publication). Frankly, I thin
192.
by kastnerkyle 13y ago
What is the advantage of using this instead of OverFeat for natural images? I have been operating on the assumption that the strange license of OverFeat is not a problem, since you are using someone's software to feed in input, and gen
193.
by kastnerkyle 13y ago
It depends - some algorithms exploit correlations between pixels to reduce the computational load. This is bad if the Z channel is not strongly correlated with the others (RGB are very strongly correlated in natural images). Since depth sho
194.
by kastnerkyle 13y ago
Open data and data sharing is one of the key components of the reproducible science movement. Some teams have even gone so far as to share their entire virtual image with data, code, and associated libraries (when data size is small). Docke
195.
by kastnerkyle 13y ago
I agree 100%. One of my favorite ways to learn a new language is by doing problems from https://projecteuler.net . The mathematical basis of Project Euler seems to map into core language concepts better than many other "get
196.
by kastnerkyle 13y ago
Can someone who knows crypto shed some light on the "gravity" of this algorithm? Is this going to out and out replace MD5, or will it fall among the many "awesome, but not widely used" hashes out there? It sounds pretty
197.
by kastnerkyle 13y ago
Great article. I am looking forward to using Julia once the data-munging (and vectorization) packages start catching up to things like Pandas and numpy. Did you have any experiences with vectorized Julia being slower than the "loopy&
198.
by kastnerkyle 13y ago
Correlation does not imply causation... that said, congratulations on your success! I would be interested to see more about the parameters of your A/B tests.
199.
Video Background Subtraction in Python(kastnerkyle.github.io)
1 points by kastnerkyle 13y ago|0 comments
200.
by kastnerkyle 13y ago
This will not magically be overcome by constructive interference - each beam will still lose a significant amount of power as function of distance. Higher frequency RF signals are fairly easy (especially V/UHF) to beamform into a tight