Ismael T. Freire
@ishmajl.bsky.social
Postdoc in NeuroAI at Sorbonne University.
Studying collaboration and morality in humans and machines. Computacional ethics, Cybernetics, ALife, self-organization, complexity, ecology, cultural evolution.
Studying collaboration and morality in humans and machines. Computacional ethics, Cybernetics, ALife, self-organization, complexity, ecology, cultural evolution.
I’m very happy to announce that "The Nature and Dynamics of Collaboration" is now out from @mitpress.bsky.social!
I’m honored to have contributed to two chapters to this incredibly multidisciplinary and exciting book, which couldn’t be more timely in these complex and challenging times.
I’m honored to have contributed to two chapters to this incredibly multidisciplinary and exciting book, which couldn’t be more timely in these complex and challenging times.
January 22, 2025 at 9:49 PM
I’m very happy to announce that "The Nature and Dynamics of Collaboration" is now out from @mitpress.bsky.social!
I’m honored to have contributed to two chapters to this incredibly multidisciplinary and exciting book, which couldn’t be more timely in these complex and challenging times.
I’m honored to have contributed to two chapters to this incredibly multidisciplinary and exciting book, which couldn’t be more timely in these complex and challenging times.
We tested SEC on several benchmarks of the Animal-AI testbed:
✅ Outperformed DQN, MFEC, and ERLAM
✅ Achieved greater policy stability
✅ Better scalability under memory limitations
✅ Outperformed DQN, MFEC, and ERLAM
✅ Achieved greater policy stability
✅ Better scalability under memory limitations
December 31, 2024 at 1:34 PM
We tested SEC on several benchmarks of the Animal-AI testbed:
✅ Outperformed DQN, MFEC, and ERLAM
✅ Achieved greater policy stability
✅ Better scalability under memory limitations
✅ Outperformed DQN, MFEC, and ERLAM
✅ Achieved greater policy stability
✅ Better scalability under memory limitations
SEC stores entire event sequences in memory, keeping their temporal order intact.
Unlike traditional episodic models that treat memories as isolated events, SEC uses sequential retrieval for direct control.
Result? Faster learning, higher rewards, and efficient memory use.
Unlike traditional episodic models that treat memories as isolated events, SEC uses sequential retrieval for direct control.
Result? Faster learning, higher rewards, and efficient memory use.
December 31, 2024 at 1:34 PM
SEC stores entire event sequences in memory, keeping their temporal order intact.
Unlike traditional episodic models that treat memories as isolated events, SEC uses sequential retrieval for direct control.
Result? Faster learning, higher rewards, and efficient memory use.
Unlike traditional episodic models that treat memories as isolated events, SEC uses sequential retrieval for direct control.
Result? Faster learning, higher rewards, and efficient memory use.
What a great way to end the year! 🎉
Thrilled to announce our paper is now out in @natmachintell.bsky.social
How can agents achieve both sample and memory efficiency?
We present Sequential Episodic Control (SEC), a hippocampal-inspired model that uses sequential memory to guide actions!
🧵
Thrilled to announce our paper is now out in @natmachintell.bsky.social
How can agents achieve both sample and memory efficiency?
We present Sequential Episodic Control (SEC), a hippocampal-inspired model that uses sequential memory to guide actions!
🧵
December 31, 2024 at 1:34 PM
What a great way to end the year! 🎉
Thrilled to announce our paper is now out in @natmachintell.bsky.social
How can agents achieve both sample and memory efficiency?
We present Sequential Episodic Control (SEC), a hippocampal-inspired model that uses sequential memory to guide actions!
🧵
Thrilled to announce our paper is now out in @natmachintell.bsky.social
How can agents achieve both sample and memory efficiency?
We present Sequential Episodic Control (SEC), a hippocampal-inspired model that uses sequential memory to guide actions!
🧵