Загрузка видео...

Не удалось загрузить видео

На главную

can a neural network learn to walk as a physical object in a physics simulation? here I train walking neural nets with an evolutionary algorithm. The input nodes/feet are activated by sine waves at learned phases & connections between two neurons extend based on their difference

339,220 просмотров • 3 лет назад •via X (Twitter)

Комментарии: 10

Фото профиля Matt Henderson
Matt Henderson3 лет назад

aka “embodied neural networks”

Фото профиля Matt Henderson
Matt Henderson3 лет назад

Remember soda constructor?

Фото профиля Patrick Ꝺoyle
Patrick Ꝺoyle3 лет назад

Whatever the result, they seem to be enjoying themselves.

Фото профиля Eric Jang
Eric Jang3 лет назад

this is amazing. would you consider open-sourcing the code? I'd love to extend this to optimizing the architecture of the net itself, so it ends up having to trade off physical bulk with function approximation power

Фото профиля Tom
Tom3 лет назад

Looks kinda similar

Фото профиля Austen Lamacraft
Austen Lamacraft3 лет назад

Surely a CNN will work better because of the strides?

Фото профиля c7ddfc
c7ddfc3 лет назад

reminds me of this, also controlled by a neural net

Фото профиля Mike Vella
Mike Vella3 лет назад

How do you think of this stuff??

Фото профиля Matt Henderson
Matt Henderson3 лет назад

I saw a cool demo of a neural net controlling an inverted pendulum, and was thinking it would be cool to make a nnet control something. I was also thinking about that old game soda constructor. Then I thought of this idea like a bad joke about learning to walk

Фото профиля Connor McCormick
Connor McCormick3 лет назад

You should make it so that the ground is a treadmill with numbers on it, and the fitness function rewards it both for walking and for adding the numbers it's touching correctly.

Похожие видео