Sign in

Bernardo Esteves

@esteveste.bsky.social
71 followers 52 following 15 posts

PhD at Técnico, University of Lisbon | Deep Learning + RL bernardoesteves.com

PostsRepliesMedia
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
Check out our paper: openreview.net/forum?id=IxR... Project website: bernardoesteves.com/NeuralSolver
openreview.net
NeuralSolver: Learning Algorithms For Consistent and Efficient...
We contribute NeuralSolver, a novel recurrent solver that can efficiently and consistently extrapolate, i.e., learn algorithms from smaller problems (in terms of observation size) and execute those...
030
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
This method can also be applied to sequential decision problems, where small mistakes lead to catastrophic results. NeuralSolver can solve the doorkey environment of size 128x128 with a performance very close to the oracle, just by performing behavior cloning on smaller sizes.
130
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
With this approach, NeuralSolver achieves almost perfect algorithmic extrapolation by executing the same learned algorithm on much harder problems. While highly surpassing previous works.
120
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
We can notice this by looking at how the latent values converge to the final result and the number of recurrent iterations. NeuralSolver learns an algorithm that starts by solving the dead ends of the maze, until finding the correct path that goes from the player to the goal.
110
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
By keeping the recurrent module output size equal to the input size, the model learns small local algorithms that are executed in parallel and work with any input size.
100
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
We do this by leveraging a recurrent model with an optional pooling layer for different-size tasks. The recurrent model keeps the input size constant at each iteration. The pooling layer is then used to collapse the information from the latent state to the desired output size.
110
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
We developed NeuralSolver, a method capable of learning algorithms that solve much more complex tasks than the ones used for training. That works on tasks that have the same or different output sizes!
110
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
Check out our paper: openreview.net/forum?id=IxR... Project website: bernardoesteves.com/NeuralSolver
openreview.net
NeuralSolver: Learning Algorithms For Consistent and Efficient...
We contribute NeuralSolver, a novel recurrent solver that can efficiently and consistently extrapolate, i.e., learn algorithms from smaller problems (in terms of observation size) and execute those...
000
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
This method can also be applied to sequential decision problems, where a small mistake in the sequence can lead to catastrophic results. NeuralSolver can solve the doorkey environment of size 128x128 with a performance very close to the oracle, just by performing behavior cloning on smaller sizes.
100
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
With this approach, NeuralSolver achieves almost perfect algorithmic extrapolation by executing the same learned algorithm on much harder problems. While highly surpassing previous works.
100
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
We can notice this by looking at how the recurrent latent values converge to the final result, along a certain number of recurrent iterations. NeuralSolver learns an algorithm that starts by solving the dead ends of the maze, until finding the correct path that goes from the player to the goal.
100
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
By keeping the recurrent module output size equal to the input size, the model learns small local algorithms that are executed in parallel and work with any input size.
100
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
We are able to do this by leveraging a recurrent model with an optional pooling layer for different-size tasks. The recurrent model keeps the input size constant at each iteration. The pooling layer is then used to collapse the information from the latent state to the desired output size.
100
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
We explore problems that increase in complexity mainly through size. Our goal is to be able to train an agent on a small/easy set of tasks and then without any more training to be able to solve more complex tasks.
171
Bernardo Esteves @esteveste.bsky.social · 26/11/2024
🚀 Excited to present NeuralSolver at NeurIPS 2024! Humans can easily learn algorithms to solve much more complex tasks, from simple tasks. Machine learning methods usually fail to do this. Can we build models that have similar algorithmic extrapolation abilities? Thread 🧵:
1508