🌟 New Research Alert! 🌟 Excited to share our latest work (accepted to NeurIPS2024) on understanding working memory in multi-task RNN models using naturalistic stimuli!: with @takuito.bsky.social and @bashivan.bsky.social #tweeprint below: - ThreadSky

About ThreadSky

🌟 New Research Alert! 🌟
Excited to share our latest work (accepted to NeurIPS2024) on understanding working memory in multi-task RNN models using naturalistic stimuli!: with @takuito.bsky.social and @bashivan.bsky.social
#tweeprint below:

Comments

averyryoo.bsky.social•91 days ago

Congrats! Excited to read this ✨

andrewzeng.bsky.social•90 days ago

Congrats Xiaoxuan!

xiaoxuanlei.bsky.social•91 days ago

Previous models of working memory have mainly focused on using abstract stimuli (Mante et al., 2013, Yang et al., 2019, Driscoll et al., 2024, Fascianelli et al., 2024, Piwek & Stokes, 2023 etc)

xiaoxuanlei.bsky.social•91 days ago

It’s unclear how high-dimensional naturalistic sensory information is encoded, retained and recalled in these models to accommodate various task demands.

xiaoxuanlei.bsky.social•91 days ago

🖥️ To answer this question, we trained multi-task RNNs (vanilla, GRU, LSTM) on 9 N-back tasks using naturalistic 3D object stimuli to study encoding, retention, & retrieval dynamics.

xiaoxuanlei.bsky.social•91 days ago

🧠 We found that multi-task RNNs (unlike single-task ones) retain both task-relevant & irrelevant info but reusable representations only emerged in simple gateless architectures.

xiaoxuanlei.bsky.social•91 days ago

📐 Surprisingly, object features are less orthogonalized in RNN representations compared to perceptual space.

xiaoxuanlei.bsky.social•91 days ago

🎯 With training, RNNs implemented chronological memory subspaces allowing them to track object information using rotational dynamics—supporting resource-based models of working memory.

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply