- 1
- 8 254
David Schäfer
Приєднався 18 січ 2024
NEAT Algorithm Visually Explained
NeuroEvolution of Augmenting Topologies (NEAT) is a genetic algorithm (GA) for training artificial neural networks based on concepts taken from evolutionary biology. This animation was made in python using the package 'manim'. The voiceover was created using ElevenLabs.
Music by Vincent Rubinetti.
Music by Vincent Rubinetti.
Переглядів: 8 292
Is (bestbirdw1 + 2ndbestbirdw1) / 2 = new w1?
Wonderful explanation, keep making videos it’s really helpful for beginners
the production quality is amazing, wow!
Now try dense multilayer NN, and evolve it using CMA-ES algorithm, and you will be surprised how fast it will converge.
Thats neat
I love this algorithm
This is a superb demo David. Are you able to share your github code? Thanks for the video!
sounds quite unusual of NEAT to take this long for a simple problem Great video though!
Great explaination! Looking forward to more videos
"more of a lucky coincidence than a well thought out strategy" Or as it's called when Space-X boosters use this exact strategy intentionally in real life, a "suicide burn".
Please make more videos on different algorithms.
It was Awsome! Tahnk you.
Bless you in Jesus name
????
why u not become the famous how
Do I understand that correct, that one always needs to define a reward function (in this case score) for this type of problem?
yes
(Reward function = score function = fitness function) Basicly all the same, and they are just to tell the algorithm if it was a bad neural network, or a good little neurie :)
It might be out of context, but you have the the voice of OmniMan
lmao fr
Really great visuals, sad it has so small amount of views
This is really great for beginners! Keep it up. I would love to see other games and configurations for the algorithm and how they could affect the simulation e.g. faster training time etc.
Amazing explanations!