Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Solve Geometric Programs in the Browser

prandtl.design

Post reply on HN
Print view View original

1–10 of 23 posts

Re: Show HN: Solve Geometric Programs in the Browser

#2

Post by modalduality »

I'm not quite able to run the code (too much load on the server probably). But is this something that can solve LPs in the browser (possibly more than just LPs, I think)? This would be incredibly helpful for many people I'm sure--all the existing online ones have arcane input reqs (and unhelpful errors) and standalone applications either cost money (Gurobi, cplex) or are not easy to use at all (gnu glpk).

Re: Show HN: Solve Geometric Programs in the Browser

#3

Post by lochieferrier »

modalduality wrote: post #2

I'm not quite able to run the code (too much load on the server probably). But is this something that can solve LPs in the browser (possibly more than just LPs, I think)? This would be incredibly helpful for many people I'm sure--all the existing online ones have arcane input reqs (and unhelpful errors) and standalone applications either cost money (Gurobi, cplex) or are not easy to use at all (gnu glpk).

Yes, there was an issue with someone nuking the compiler box, and we're working on fixing the demo up.

This does more than LPs, it lets you formulate GPs, which are a way of doing non-linear analytical optimization. See here for more: https://gpkit.readthedocs.io/en/latest/gp101.html#what-is-a-...

Re: Show HN: Solve Geometric Programs in the Browser

#6

Post by kuon »

lochieferrier wrote: post #3
modalduality wrote: post #2

I'm not quite able to run the code (too much load on the server probably). But is this something that can solve LPs in the browser (possibly more than just LPs, I think)? This would be incredibly helpful for many people I'm sure--all the existing online ones have arcane input reqs (and unhelpful errors) and standalone applications either cost money (Gurobi, cplex) or are not easy to use at all (gnu glpk).

Yes, there was an issue with someone nuking the compiler box, and we're working on fixing the demo up. This does more than LPs, it lets you formulate GPs, which are a way of doing non-linear analytical optimization. See here for more: https://gpkit.readthedocs.io/en/latest/gp101.html#what-is-a-...

HN community should add an internet rule:

xx) If it's online, it will get nuked.

Re: Show HN: Solve Geometric Programs in the Browser

#9

Post by sleavey »

The problem of solving angular constraints has interested me for a while. I want to be able to programmatically lay out an optical setup with SVG. Here is my work so far, but I've run up against the problem of dealing with cyclic graphs. The "rewrite" branch uses a different approach based on an academic paper.

https://github.com/SeanDS/optivis/tree/rewrite

Re: Show HN: Solve Geometric Programs in the Browser

#10

Post by badminton1 »

aerioux wrote: post #5
badminton1 wrote: post #4

Doesn't scale

I dont think that's fair to them. Having a GP solver in the bg is very hard and very computationally expensive -- their costs are way more than similar Show HN and saying "Doesn't scale" I don't believe acknowledges this well

No, it is not unfair. It is valid, concise, actionable feedback, without personal attacks based on empirical evidence.

If you are not comfortable with constructive feedback then don't post an open invitation for feedback.

Then, it can be implemented in a way that scales. Since this is python + gpkit, maybe it could be executed on the client-side using emscripten. Then if the traffic is too high you can serve it from a CDN.

Post reply on HN