Recurrent Neural Network Visualization
Channel:
Subscribers:
27
Published on ● Video Link: https://www.youtube.com/watch?v=aWcvIKM_VQo
A 128x128 fully recurrent neural network. The brightness of each pixel corresponds to that neurons activation level. Each neuron is modeled as a node that is connected to it's 8 neighbors by weighted connections, and also to a small amount of neurons in a random location within the grid (modeling an axon), also with weighted connections. A neuron receives a decaying input and if it surpasses the threshold value the neuron fires, sending a signal to its neighbors as well as down it's axon. The program uses ~4Mb of memory, but is less than .12mb in size :D
Other Videos By Xalriel
2021-02-21 | Laser Buttholes?!? |
2021-01-25 | Gladius VR Funny Moments, Murders & Glitches 1/25/2021 [Oculus Quest] |
2021-01-24 | Let's Play Ancient Dungeon VR 1/24/21 [Oculus Quest] |
2020-08-09 | Jennie Wade House Groan EVP |
2020-07-29 | Speed Modeling! - The SpaceBar |
2020-07-26 | VRChat World - Rave Cave by Crypto [Neo] |
2020-07-08 | VRChat World Creation Fun! |
2020-06-11 | Beat Saber Training Session |
2020-06-07 | Beat Saber Fail Session |
2017-04-23 | Recurrent Spiking Neural Network Visualization |
2017-04-13 | Recurrent Neural Network Visualization |
2016-08-31 | 3D Printed Robot Claw |
Tags:
simulation
neural
network
recurrent
model
ann
brain
ai
artificial intelligence