Chapter
The Power of Neural Evolution
1:10:10 - 1:18:44 (08:34)
The combination of neural networks and evolutionary computation can produce a powerful mechanism for constructing brains and behaviors, allowing the evolution of parameters and weight values in neural networks for specific purposes.
Clips
Neural evolution combines neural networks and evolutionary computation to construct a neural network using evolution to determine parameters, like weight values, activation functions, and loss functions, instead of backpropagation or stochastic gradient descent.
1:10:10 - 1:12:39 (02:29)
Summary
Neural evolution combines neural networks and evolutionary computation to construct a neural network using evolution to determine parameters, like weight values, activation functions, and loss functions, instead of backpropagation or stochastic gradient descent.
ChapterThe Power of Neural Evolution
Episode#177 – Risto Miikkulainen: Neuroevolution and Evolutionary Computation
PodcastLex Fridman Podcast
The interaction of two timescales, where evolution gives baby neural networks that eventually learn during their lifetime, coupled with an intelligent environment can potentially be a powerful method to construct brains and behaviors.
1:12:39 - 1:13:53 (01:14)
Summary
The interaction of two timescales, where evolution gives baby neural networks that eventually learn during their lifetime, coupled with an intelligent environment can potentially be a powerful method to construct brains and behaviors.
ChapterThe Power of Neural Evolution
Episode#177 – Risto Miikkulainen: Neuroevolution and Evolutionary Computation
PodcastLex Fridman Podcast
This podcast explores the fascinating space between neural networks, evolution and computation with a focus on growing neural networks and optimizing their hyperparameters.
1:13:53 - 1:14:54 (01:00)
Summary
This podcast explores the fascinating space between neural networks, evolution and computation with a focus on growing neural networks and optimizing their hyperparameters. The guest speaker delves into the challenges of growing a neural network and what kind of architectures are more amenable to this idea.
ChapterThe Power of Neural Evolution
Episode#177 – Risto Miikkulainen: Neuroevolution and Evolutionary Computation
PodcastLex Fridman Podcast
The podcast discusses building a tiny network that can grow into something state-of-the-art using intelligent methods and figuring out the right representations and operators.
1:14:54 - 1:17:27 (02:32)
Summary
The podcast discusses building a tiny network that can grow into something state-of-the-art using intelligent methods and figuring out the right representations and operators. The host suggests breaking down winning architectures and creating better representations to approach a problem effectively.
ChapterThe Power of Neural Evolution
Episode#177 – Risto Miikkulainen: Neuroevolution and Evolutionary Computation
PodcastLex Fridman Podcast
A potential approach to simulating neural network growth involves evolving a starting point and then continuously training the network.
1:17:27 - 1:18:44 (01:17)
Summary
A potential approach to simulating neural network growth involves evolving a starting point and then continuously training the network. However, creating a simulation environment that enables interactions in the real world is challenging.