chewxy

44 ms·

481.
Show HN: I wrote a simple simulation of the economy to show how money is created(nbviewer.jupyter.org)
6 points by chewxy 10y ago|4 comments
482.
by chewxy 10y ago
For Gorgonia it's the other way around. You build the graph by manually unrolling it. It'll be a big goddamn graph, which is why you subgraph it and run it. source: actually have various LSTMs running. Some with attention, some wi
483.
by chewxy 10y ago
I just checked out BrainScript. Looks pretty cool. Hacking up a parser for that into Gorgonia should be fairly trivial. Could be a fun weekend hacking project.
484.
by chewxy 10y ago
The trick, IMO, is to manually unroll them. So what I do is something like this: func (m *model) costFn() (cost *Node, n int, err error) { var prev *recurrentOutput for ...{ // build your recurren
485.
by chewxy 10y ago
I agree with you in general. However, I contend that people are lazy. Part of the reason why I wrote Gorgonia was because I spent waay too long trying to deploy theano on the cloud (this was circa 2 years ago). Also, thanks for informing me
486.
by chewxy 10y ago
It's not released - simply because it failed to even build properly. No I have not reached out to nvidia. Don't have any contacts there
487.
by chewxy 10y ago
Hi, I wrote Gorgonia: https://github.com/chewxy/gorgonia If there are any questions I'm available to answer them
488.
by chewxy 10y ago
Microsoft has CNTK - https://github.com/Microsoft/CNTK Which surprisingly is one of the few I didn't refer to when implementing Gorgonia
489.
by chewxy 10y ago
Hi, So I wrote my own CUDA wrapper. Turns out there was a bug. I'm not sure where and how, but it worked for the earlier versions of Gorgonia but not this release. You may also note that a bunch of assembly for the float32 tensor are m
490.
by chewxy 10y ago
I also documented some of the quirks and outlined the release in my blog: http://blog.chewxy.com/2016/09/19/gorgonia/ I'd love some feedback on Gorgonia - did I just waste my time writing it when Th
491.
Show HN: Gorgonia – primitives for building neural networks in Go (like Theano)(github.com)
3 points by chewxy 10y ago|1 comments
492.
by chewxy 10y ago
Well, for one, you will have to implement the memory/system model for the language. It may or may not have access to the host language's "printf" function. Initially of course, a shortcut way is to leverage as much of th
493.
by chewxy 10y ago
I put in all my interests (and current meetup groups) and Brisbane was in the top 10 list. The funny thing is I moved FROM Brisbane to Sydney (for better jobs)
494.
by chewxy 10y ago
There was a linguistic shift to prefer 吃. I would say it probably happened in the Cultural Revolution. In Cantonese 食 is pronounced "sek", and is regularly used - "sek fan" as in "eat rice". Since HK was quite
495.
by chewxy 10y ago
Agreed. The ghosts bit is apocryphal. Not difficult to see where it comes from though. Mouth + Qi = spiritual nonsense.
496.
by chewxy 10y ago
Probably some stupid made up shit to scare kids into using "proper" chinese (for certain definitions of proper as defined by political agendas). I do find 文言文 to be quite elegant and terse though.
497.
by chewxy 10y ago
The etymology of 食 is a bit more complex than that. It's ancient chinese that combined an upside down mouth: 亼 (best approximation) and a bowl of rice. This is the 甲骨文 version: http://imgur.com/NRBoG7F . Cute eh? It loo
498.
by chewxy 10y ago
When I was younger and took chinese lessons, there was a push to use more "official" chinese. We'd be told that in official use "吃" is a verb used by ghosts (which is to say, it's impolite to use 吃, and "食
499.
by chewxy 10y ago
That's because you're thinking in English. "Kingly" is a very English way of thinking of things (as in king-like behaviour for example). In Chinese and Japanese, there simply isn't a way to describe it as elegantly
500.
by chewxy 10y ago
At about 21 mins in, the awesome happens.
501.
by chewxy 10y ago
Because you will end up with unwieldly graphs like these: https://twitter.com/chewxy/status/433807908875145216 https://twitter.com/chewxy/status/433811541373116416 This is from an old JI
502.
by chewxy 10y ago
FYI, the verb for "convolution" is "convolve", not "convolute". English is a weird arse language
503.
by chewxy 10y ago
Nothing is stopping you from doing something stupid like this though: func AtRuntime(writer io.Writer){ templ := template.Must(template.New("blah").Parse(templateText)) templ.Execute(writer, someStructhere)
504.
by chewxy 10y ago
Villager: "What happened, Danny?" PC Butterman: "Car accident" Villager: "Nasty way to go" Sgt Angel: "Official vocab guideline state we no longer refer to these incidents as 'accidents'. They&#x
505.
Habits of highly mathematical people(medium.com)
4 points by chewxy 10y ago|0 comments
506.
by chewxy 10y ago
Anyone who starts a project for #1 will get instantly targeted by Samaritan's agents, right?
507.
by chewxy 10y ago
Damn. That's neat. I wrote my own python script to translate whatever `as` spits out into BYTE $.... your tool'd have been neat when I wrote this http://blog.chewxy.com/2016/07/25/on-the-memory-align
508.
by chewxy 10y ago
Appnexus is a RTB platform and a DSP (and a whole lot of other things they do). OpenX is/was an exchange and is generally the #1 place for buying/selling dodgy af traffic. Neither does what Adsense does per se. What Adsense does i
509.
by chewxy 10y ago
It DOES use GIFT - http://github.com/disintegration/gift , but eh.. it works
510.
Parellagram – Dynamic Image Resizing with Go(parellagram.com)
68 points by chewxy 10y ago|16 comments

More ›