[QUOTE=Hentie;50254125][media]https://www.youtube.com/watch?v=L_JBy5kdCqQ[/media]
video of neural network training in process. this was for the nasa space apps 2016 competition. the data points on the video are taken from recent natural disasters ranging from floods, earthquakes, wildfires, and storms. this is an older video so the output shown isn't as refined as what we have now.
i can't take all the credit for the neural network code since i used a framework, but i thought the result looked cool.
here is the current one using historical data instead of current data, laid out on a 3D sphere:
[img]http://i.imgur.com/4Dhhdyg.png[/img][/QUOTE]
Does it predict future disasters or what?
Looks more like an amalgation of statistics -- a way to increase preparedness in certain areas.
[QUOTE=fauxpark;50254107]If your method is throwing that many exceptions, your method is either doing too much, or you're not dealing with them properly[/QUOTE]
Plenty of methods in .NET throw that many exceptions and more
[QUOTE=voodooattack;50254066]And here's how you implement an ES6 promise in C++14:
[IMG]https://dl.dropboxusercontent.com/u/27714141/Screenshot%20from%202016-05-04%2009-19-48.png[/IMG][/QUOTE]
Your color scheme is very colorful
[QUOTE=MuffinZerg;50254228]Does it predict future disasters or what?[/QUOTE]
I don't see it predicting future disasters very well, it just gives you a general idea of what disasters occur where. Originally we were just going to suggest disaster tips depending on how close you were to a historical natural disaster, but we thought that this method was cooler.
[QUOTE=Map in a box;50254339]Looks more like an amalgation of statistics -- a way to increase preparedness in certain areas.[/QUOTE]
That's exactly the idea! : )
So the company Anker that sells computer electronics and what not, held a contest a few weeks back to win some of their stuff.
I ended up entering in it, and told some of my Steam buddies about it, "even a few WAYWO'ers <3" long story short I won their main prize, which is this $500 battery bank.
[media]http://imgur.com/a/5qAjZ[/media]
I'm aware this isn't exactly WAYWO worthy, but hey, I'm sure its something us nerds can all enjoy. :v:
[QUOTE=proboardslol;50255345]Your color scheme is very colorful[/QUOTE]
Is that a bad thing?
[QUOTE=sarge997;50256154]So the company Anker that sells computer electronics and what not, held a contest a few weeks back to win some of their stuff.
I ended up entering in it, and told some of my Steam buddies about it, "even a few WAYWO'ers <3" long story short I won their main prize, which is this $500 battery bank.
[media]http://imgur.com/a/5qAjZ[/media]
I'm aware this isn't exactly WAYWO worthy, but hey, I'm sure its something us nerds can all enjoy. :v:[/QUOTE]
That is badass, fucking 500$ battery bank with power supply. Excellent for camping and science/research in the wild work.
How come I never find all these awesome contests :saddowns:
[QUOTE=Hentie;50255795]I don't see it predicting future disasters very well, it just gives you a general idea of what disasters occur where. Originally we were just going to suggest disaster tips depending on how close you were to a historical natural disaster, but we thought that this method was cooler.
[/QUOTE]
But why use a neural network then?
[QUOTE=MuffinZerg;50256492]But why use a neural network then?[/QUOTE]
Because its an optimal way of doing it.
You aren't going to be able to predict natural disasters.
[QUOTE=voodooattack;50256290]Is that a bad thing?[/QUOTE]
I mean it's personal preference so whatever works for you is good, I just think I would personally get lost all the time :v:
And it also could indicate that perhaps you could use some more whitespace or comments to make things look more organized. But once again, whatever works for you.
[QUOTE=Map in a box;50256651]Because its an optimal way of doing it.
You aren't going to be able to predict natural disasters.[/QUOTE]
You could predict them, but I wasn't talking about that.
I still don't get it. Given some points on the map that represent locations of natural disasters, do you really need a neural net to draw a big circle containing the points plus some "safety" margin?
[QUOTE=Map in a box;50256439]How come I never find all these awesome contests :saddowns:[/QUOTE]
I joined a lego sweepstakes once when I was 9 and got a signed copy of Bioncle comic #14:
[t]http://vignette2.wikia.nocookie.net/bionicle/images/a/a6/14-Cover.jpg/revision/latest/scale-to-width-down/250?cb=20130823021029[/t]
[QUOTE=MuffinZerg;50256668]You could predict them, but I wasn't talking about that.
I still don't get it. Given some points on the map that represent locations of natural disasters, do you really need a neural net to draw a big circle containing the points plus some "safety" margin?[/QUOTE]
I understand you completely, just fit a ellipse to points. But I guess the dude is just learning about neural networks and showing off.
[QUOTE=MuffinZerg;50256668]You could predict them, but I wasn't talking about that.
I still don't get it. Given some points on the map that represent locations of natural disasters, do you really need a neural net to draw a big circle containing the points plus some "safety" margin?[/QUOTE]
A neural net is a bit more sophisticated than that. It can approximate any function, so with precautions to prevent overfitting a neural network can make a very accurate statistical model without requiring the user to know or guess what kind of function their training data is going to produce.
Without a neural network, you'd have to take an educated guess at what kind of curve your data follows in order to generate a statistical model. For example, you mentioned a "big circle". You very well could approximate the function as a circle, and find the circle that best fits your data, but as the data is not distributed in a circle, the model would be highly inaccurate. It would take a lot of trial and error to find what kind of shape the data fit well into, and a neural network is essentially an efficient abstraction for that trial and error.
[QUOTE=voodooattack;50256290]Is that a bad thing?[/QUOTE]
not at all, I'd actually like to know what color scheme it is so I can use it too. or is it a custom one?
[QUOTE=Map in a box;50256651]Because its an optimal way of doing it.
You aren't going to be able to predict natural disasters.[/QUOTE]
That's a common misconception. Neural networks are the least optimal solution, as ist takes a very long time to find a solution. And even if you find one, you don't really learn how it did it, as you can only observe the end result.
[QUOTE=DrDevil;50257733]That's a common misconception. Neural networks are the least optimal solution, as ist takes a very long time to find a solution. And even if you find one, you don't really learn how it did it, as you can only observe the end result.[/QUOTE]
You don't need to know how it did it, you only need the end result.
I know what NNs are bad for but this is a really good use for them imo
[editline]4th May 2016[/editline]
alternatively, a heatmap
[QUOTE=DrDevil;50257733]That's a common misconception. Neural networks are the least optimal solution, as ist takes a very long time to find a solution. And even if you find one, you don't really learn how it did it, as you can only observe the end result.[/QUOTE]
Neural networks are just advanced case of optimization. Really it's just derivatives and finding the minimum in the end.
[editline]5th May 2016[/editline]
Well atleast for normal NN, I don't know the case for convolutional NNs.
[QUOTE=proboardslol;50256652]I mean it's personal preference so whatever works for you is good, I just think I would personally get lost all the time :v:
And it also could indicate that perhaps you could use some more whitespace or comments to make things look more organized. But once again, whatever works for you.[/QUOTE]
That's a good point. I do plan on commenting the code at one point in the future.
[QUOTE=Lumaio;50256880]not at all, I'd actually like to know what color scheme it is so I can use it too. or is it a custom one?[/QUOTE]
Just the default KDevelop dark theme.
[QUOTE=voodooattack;50258451]That's a good point. I do plan on commenting the code at one point in the future.
[/QUOTE]
All those nested if statements scream code smell. A way you can make your code more expressive without comments is to extract that fat method into many smaller private functions.
[vid]https://files.catbox.moe/n1e710.webm[/vid]
Just some small stuff I've been working on. Animated items/ two hats at the same time :incredible:
Haven't had time to do much else lately
edit:
bonus crab hat + santa hat
[IMG]https://dl.dropboxusercontent.com/u/2276133/ShareXmod/05-2016/javaw_2016-05-04_23-02-32.png[/IMG]
hello friends, which perspective do you think works better
[media]http://www.youtube.com/watch?v=wMX143I2SU4[/media]
[media]http://www.youtube.com/watch?v=YtyquNEIfJo[/media]
[QUOTE=Nigey Nige;50261263]hello friends, which perspective do you think works better
[/QUOTE]
I like the first, it adds character to the game and doesn't seem to impede the player
first one looks better but i'm not sure if the controls would be annoying from that perspective
[QUOTE=Nigey Nige;50261263]hello friends, which perspective do you think works better
[/QUOTE]
Why not the option to switch between them?
But the first definitely looks better.
[QUOTE=FlakTheMighty;50261573]Why not the option to switch between them?
But the first definitely looks better.[/QUOTE]
Problem then is that all design will take twice as long to account for two different perspectives
go with the first one but make the camera lead/zoom out when you're going fast
Definitely the first, in my opinion.
Sorry, you need to Log In to post a reply to this thread.
