iiinova — where your digital self keeps evolving

Xiaoxuan

@xiaoxuanlei.bsky.social

102 followers 65 following 11 posts

CompNeuro & AI ❤️ FutureToBeBlackBoxBreaker👻 PhD candidate @McGill et @Mila

Posts Replies Media

Xiaoxuan @xiaoxuanlei.bsky.social · 12/12/2024

📌 Poster Session: ⏰ When: TODAY, Thu, Dec 12, 4:30 p.m. – 7:30 p.m. PST 📍 Where: East Exhibit Hall A-C, #3705 📄 What: Geometry of Naturalistic Object Representations in Recurrent Neural Network Models of Working Memory Hope to see you there! @bashivan.bsky.social @takuito.bsky.social

031 0

Xiaoxuan @xiaoxuanlei.bsky.social · 12/12/2024

Excited to be at #NeurIPS2024 in #Vancouver! Our poster session is TODAY—if you're interested in naturalistic representations in dynamic working memory models, please drop by and let’s chat!

110 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

👉 Check it out: arxiv.org/abs/2411.02685 📅 We’ll be at NeurIPS! Join us for our poster presentation on Thu 12 Dec, 7:30 p.m. EST — 10:30 p.m. EST. #AI #CognitiveScience #WorkingMemory #DeepLearning #RepresentationGeometry #MultiTask

arxiv.org

Geometry of naturalistic object representations in recurrent neural network models of working memory

Working memory is a central cognitive ability crucial for intelligent decision-making. Recent experimental and computational work studying working memory has primarily used categorical (i.e., one-hot)...

210 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

Our findings bridge cognitive science & AI, revealing how high-dimensional object information is encoded, retained, and recalled in recurrent models of working memory.

110 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

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

100 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

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

100 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

🧠 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.

100 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

🖥️ 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.

110 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

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

100 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

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)

100 0

Xiaoxuan @xiaoxuanlei.bsky.social · 28/11/2024

🌟 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:

34917 0