highd

15 ms·

91.
by highd 10y ago
I think you're correct, barring significant nonlinearity in the bayer mask or object. Technically you can get 9 linearly independent points - each combination of light channels on with each combination of bayer mask channels. Ideally o
92.
by highd 10y ago
I'm addressing your point that it's a solved problem. The issues around making high bandwidth data interconnects are not just "straight bandwidth". I'm explaining that there are fundamental interconnections between
93.
by highd 10y ago
Higher bandwidth means more noise, which means you need either higher gain or more sensitive detection (until you hit thermal limits). They're very interrelated issues.
94.
by highd 10y ago
There's also issues with how fast the satellites move across the sky. At low orbits they fly by and you either need ridiculously sensitive wide angle antennas or solid-state steerable high-gain antennas that track them as they go by. S
95.
by highd 10y ago
You could potentially switch to a mesh or some kind of resonant structure with large enough gaps that cell signals could get through while blocking the rooms designed resonant frequency.
96.
by highd 10y ago
The company is called High Q Imaging. We have a bare bones website that we'll hopefully update in the next month or so: http://www.highqimaging.com/
97.
by highd 10y ago
Interesting, I'll have to get a quote and run some numbers.
98.
by highd 10y ago
Thank you!
99.
by highd 10y ago
I think you're exactly right - more frequent, less expensive testing means we have to change the criteria that mandates further investigation. One dilemma I've been thinking about recently is that there's an unfortunate burde
100.
by highd 10y ago
I can't go too much into details, but suffice it to say that we're not using super conductors at all. High T super conductors will be very interesting in the next decade, I think, but right now there are a few issues that I'm
101.
by highd 10y ago
Regarding cancer, we think you're right. That and possibly maintenance - for example, most prostate cancers are asymptotic for life if untreated. For slower cancers that are acquired in old age it can make sense to use less aggressive
102.
by highd 10y ago
I've been working on a bootstrapped company building a low-cost MRI for about a year now. We're hoping to have a demonstrable prototype in another year or so. So we think that we can get to that future where people pick MRI based
103.
by highd 10y ago
One interesting thing going forward (over the next decade) regarding deep learning and high-performance computing will be if we can circumvent the big serial bottleneck - that is, depth. The best benchmarks have generally all been set by ne
104.
by highd 10y ago
The improvements to gradient descent techniques over vanilla gradient descent as well as ideas like batch normalization and residual nets have made even moderately large networks that were previously impossible to train now possible, let al
105.
by highd 10y ago
Sure! SHM is really cool, I just found out about it. It's an old Posix functionality, so people should use it more! I'm using shm system calls in Python. Basically I get a buffer of raw bytes of a fixed size that is referred to by
106.
by highd 10y ago
I generally use a semaphore when I'm reading and writing from my shm'd things. The data structure will also likely be append-only for the training process, as I want to see how things are changing over time. Also I meant shm'
107.
by highd 10y ago
Yes! I'm thinking of building something like this for my neural net training (1-2 days on AWS, 16 GPUs/processes on the job). In this case the "state" that I'd like to access is all the parameters of the model and t
108.
by highd 10y ago
SEEKING WORK | Boston | Remote or local Algorithms / Machine Learning / Optimization Problems / Statistics as well as the full stack infrastructure to build a product around it (generally Flask + HTML/CSS/JS/D3
109.
by highd 10y ago
You may be able to do that. That's sort of a matter of preference which you'd like to do. If the two datasets share more structure than it's more advantageous to share the network. There's also a bunch of hybrid approach
110.
by highd 10y ago
If you're comparing deltas (i.e. x_{n+x} - x_{n+x-1}) that might be sufficient - otherwise it's hard to tell if you're just capturing that x_{n+1} is close to x_{n}. The primary risk would be that you're putting strong s
111.
by highd 10y ago
Timeseries can 100% work like that. If you expect your timeseries data to be coming from a similar distribution that is what you do to train your LSTM. It's not just a magic box - you have to train it to encode useful features in the g
112.
by highd 10y ago
FMCW systems would be slowly changing around the central wavelength. It wouldn't work for pulsed systems, though - I'm not sure what's more common.
113.
by highd 10y ago
What you can say is if there are N videos and k types of transforms that are used to "hide" content this would be hypothetically perfect with O(n + k) train examples - naive "reporting only" would be O(nk). This means th
114.
by highd 10y ago
I'm unclear - aren't you showing results for training data anyway? The network might just be compressing the trends in the training data into its function. The question should be if the same network works on another time series, r
115.
by highd 10y ago
True! And one should always KISS to start, at least.
116.
by highd 10y ago
I would construct a list of invariances that a classifier should be robust to, then construct (either empirically, analytically or via a machine learning algorithm using a synthetic dataset) a series of features that are invariant under the
117.
by highd 10y ago
I've gone through the process of port forwarding my router and setting up dynamic DNS (who does that anymore? :) ) so I can hit my home desktop with an Ubuntu VM on it and I usually have a few AWS instances up at any given time for dif
118.
by highd 10y ago
I'm a vim+screen+mosh(+Jupyter) guy, so my environment is "cloud-based" as-is (in that I'm mosh-ing to a dev server). I can see why some people might want more fully featured IDEs, but this works for me. As I get better
119.
by highd 10y ago
Wouldn't you normally mix down on the receive side for better SNR? Doppler should be pretty negligible, so if your LO tracks the output the receive side could probably be an audio ADC. Power on the output side is the big one for sure,
120.
by highd 10y ago
The latency / motion sickness issues probably aren't as bad, since you still see the surrounding environment to get a stable bearing. The display tech sounds very hard, though.

More ›