cshenton

10 ms·

61.
by cshenton 8y ago
If you look up gaming benchmarks. The performance losses make it more economical to build an entire separate computer, at least for that use case.
62.
by cshenton 8y ago
Likely because a new MacBook is no longer backwards compatible with the last 3-5 iterations of chargers.
63.
by cshenton 8y ago
Well if you're comparing it to Leopard / Snow Leopard no wonder. That's probably the most neat and polished a desktop operating system has ever been. edit: To clarify, I think current macOS is a better OS, its features are wo
64.
by cshenton 8y ago
I just had this yesterday. Tried to log in, password forgotten, so went through password reset. Logged in with new password ‘suspicious activity’ (my failed logins before reset) ‘you now must reset your password.’ WTF, I just did minutes ag
65.
by cshenton 8y ago
If this had been communicated as "Improving Autoplay Blocking in Chrome" I think it would have been better received. Because that's just what it is. Instead I had a sort of loss aversion based reaction where I think now some
66.
by cshenton 8y ago
The algorithm only communicates sequences of seeds so the communication overhead is very low. The master server doesn’t even break 1% cpu servicing a few 100 workers.
67.
by cshenton 8y ago
If anyone’s interested, training a single game is affordable on spot instances. I have a replication including cloud formation scripts here https://github.com/cshenton/neuroevolution
68.
Distributed Neuroevolution with gRPC(github.com)
1 points by cshenton 9y ago|0 comments
69.
Atari Reinforcement Learning Leaderboard(github.com)
7 points by cshenton 9y ago|0 comments
70.
by cshenton 9y ago
Last time this happened, sci-hub posted DNS servers that you can use. 80.82.77.83 and 80.82.77.84
71.
by cshenton 9y ago
Not quite, genetic algorithms represent the search space usually as bit sequences, so you can just mix subsequences from two parents, that's reproduction. In contrast, evolution strategies have a search distribution, and compute a grad
72.
by cshenton 9y ago
Actually, evolutionary strategies make use of either empirical or naturalised gradients on the search space.
73.
by cshenton 9y ago
Probably they mean megawatt hours. Which is a common unit of energy in the industry. Then adding per minute makes sense.
74.
by cshenton 9y ago
We had a similar issue at an apartment block in Australia. The most amusing sight was a bunch of residents, all wandering the corridors trying to find which room was the source, but as in your case it was fans on the roof.
75.
by cshenton 9y ago
It's probably to do with the fact that one is a "number". If I'm correct, that means in `x[1.0 << s]`. 1.0 is a "number", therefore the shift is valid, but in `var u = 1.0 << s`, it's a float64
76.
by cshenton 9y ago
If you have a common problem, like facial recognition, there are also a bunch of 3rd party services out there, which will save a lot of time over generic solutions and can even be cheaper. We just launched one such solution for time series
77.
by cshenton 9y ago
I picked up Bulma for my startup’s landing page and account admin (URL in my profile), and as a backend engineer who finds the front end ecosystem overwhelming, Bulma was refreshingly sensible. Default styling feels a lot like github with a
78.
Show HN: Chrono API – Real Time Forecasting(chronoapi.com)
3 points by cshenton 9y ago|0 comments
79.
by cshenton 9y ago
Not OP, but I'd assumed it was referring to removal of literal data centre racks.
80.
by cshenton 9y ago
I’ve been using drone, which is open source. It’s the only CI tool I’ve found writing pipelines in to not be painful. Need a test DB? The syntax is just docker compose so you already know how to get yourself one.
81.
by cshenton 9y ago
So I'm guessing it's the runtime itself that expects to interact with things like an OS scheduler. Does that mean if you attempted to write an OS in Go, you'd basically have to do it without many of the features of the runtim
82.
by cshenton 9y ago
What use cases need no runtime (genuine question)? I understand the benefit of being compiled, since you can target a cpu instruction set and run without dependencies.
83.
by cshenton 9y ago
Add to that: - documentation is all in PDF format - can only install packages from the interpreter - testing libraries not feature complete - weird namespacing - poor test coverage in popular packages - no mature webserver
84.
by cshenton 9y ago
What's so hard about `pip install tensorflow`?
85.
by cshenton 9y ago
Neural Networks are a very small part of TFs API surface area. I'm writing tensorflow code every day and not doing anything remotely close to a neural network.
86.
by cshenton 9y ago
I had the same question when I was first choosing libraries for my current project! So I put together this comparison of PyMC3, Edward, and Stan. It's jut doing bayesian inference on the location and scale of data sampled from a normal
87.
by cshenton 9y ago
Time Series forecasting as a service. It's an area where careful treatment of error distributions matters a lot. There's a link in my HN profile, though the page won't be up for another week.
88.
by cshenton 9y ago
I've had the pleasure of working with Edward for the last couple of months, and it's been the first PPL I've felt comfortable building a stable production codebase on top of. It strikes a great balance between feeling like yo
89.
by cshenton 9y ago
I had a similar experience. I got severe migraines all throughout my childhood. When I was studying my masters I couldn't handle all the reading so I went to the optometrist, who sent me to a binocular vision specialist who prescribed
90.
by cshenton 9y ago
Being able to sell one cryptocurrency for another doesn't mean the first cryptocurrency is liquid.

More ›