highd

19 ms·

1. 151.▲
   by highd 10y ago

   Thank you!
2. 152.▲
   by highd 10y ago

   Thanks for describing MS-SIM - I've always been concerned about using L2 or similar metrics for image similarity. Is there any work on learning an image similarity metric? I would think if you Mechanical Turk'd a bunch of A/B
3. 153.▲
   by highd 10y ago

   I wouldn't say "teens using snapchat" is "the industry", I would say it is "a industry". If you look at the S&P 500, a tiny fraction of the companies there are "social media", and even softwa
4. 154.▲
   by highd 10y ago

   This is a brilliant idea, IMO, and a real game changer (in principle, not necessarily tied to the implementation). Without this, CV problems are largely only limited by how much data you can get. If you can simulate arbitrary scene configur
5. 155.▲
   by highd 10y ago

   Deconvolution is pretty hard since it's basically inverting a very poorly-conditioned matrix - that's also assuming you know the kernel. The solution is generally to use some kind of prior on the recovered image. While a normal co
6. 156.▲
   by highd 10y ago

   I highly recommend considering consulting/freelancing when looking at starting a company. Hourly rates generally price in an expectation of not having work full-time, so you work less than 40 hours/week to begin with. Drop an hour
7. 157.▲
   by highd 10y ago

   I love detailed, niche technical blogs like this. Another one I really like is Image Sensors World: https://image-sensors-world.blogspot.com/
8. 158.▲
   Show HN: Explore the Patent Citation Graph(patgraph.com)

   9 points by highd 10y ago|0 comments
9. 159.▲
   by highd 10y ago

   That is good to know. I have enjoyed writing C++ in the past - I like the clarity and forced obviousness of specifying whether you're talking about a variable value or a pointer reference to it. When switching languages I always have t
10. 160.▲
    by highd 10y ago

    I have an algorithm that can guess that sequence right 62% of the time then ;)
11. 161.▲
    by highd 10y ago

    SEEKING WORK Boston / Remote Primary expertise in algorithms R&D for machine learning and data science applications. Particular expertise in incorporating physics based models of sensing phenomena into signals processing problems,
12. 162.▲
    by highd 10y ago

    It's not stated very directly, but it appears the article is referring to hybrid propulsion systems much like in hybrid cars. That certainly makes sense for reducing fuel costs, as well as reducing the volume of fuel you'd need, t
13. 163.▲
    by highd 10y ago

    I was toying with a little android tablet (Samsung) and a bluetooth keyboard for a while. I basically just ssh'd somewhere and used vim mostly. I had some keybinding issues, but the principle was sound. With the case it's all abou
14. 164.▲
    by highd 10y ago

    Results should be the same as the convex optimization technique in the "well-defined" entries, though numerical stability is a real concern that I haven't looked at yet. Empirically I haven't had any issues, but if you h
15. 165.▲
    by highd 10y ago

    Wow, that's a very elegant connection! I'll have to do some more research on Secant varieties.
16. 166.▲
    A Different Approach to Low-Rank Matrix Completion(highdimensionality.com)

    78 points by highd 10y ago|5 comments