Research map / Probabilistic and scientific modeling / Dynamical systems
TAVAE: A VAE with Adaptable Priors Explains Contextual Modulation in the Visual Cortex
Balázs Meszéna, Keith Thomas Murray, Julien Corbo, O. Batuhan Erkat, Márton A. Hajnal, Pierre-Olivier Polack, Gergo Orban
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use a task prior matching 45 Go and 90 NoGo instead of the VAE natural prior | TAVAE · Model vs mouse V1 population profile | Pearson correlation r(TASK): 0.53 -> 0.78 | Improved |
| Keep the trained 45/135 prior rather than updating to 45/90 (ablation) | TAVAE · Model vs mouse V1 population profile | Pearson correlation r(TASK): 0.78 -> 0.54 | Hurt |
| Match the prior to each day's stimulus (eager adapter) instead of a prior fixed at 45/90 (ablation) | TAVAE · Model vs mouse V1 population profile | Pearson correlation r(TASK): 0.78 -> 0.53 | Hurt |
| Updating the prior to 45/90 versus retaining the trained prior, effect on task modulation (ablation) | TAVAE · Model vs mouse V1 task modulation | Pearson correlation r(TASK-NAIVE): 0.32 -> 0.58 | Improved |
| Eager adapter prior versus fixed 45/90 prior, effect on task modulation (ablation) | TAVAE · Model vs mouse V1 task modulation | Pearson correlation r(TASK-NAIVE): 0.58 -> -0.10 | Hurt |
| Use reduced test contrast 0.3 instead of full contrast 1.0 (ablation) | TAVAE · Model vs mouse V1 population profile | Pearson correlation r(TASK): 0.61 -> 0.78 | Improved |
| Use reduced test contrast 0.3 instead of 1.0, effect on task modulation (ablation) | TAVAE · Model vs mouse V1 task modulation | Pearson correlation r(TASK-NAIVE): 0.52 -> 0.58 | Improved |
| Add the GSM-style scaling latent to the base VAE (ablation) | TAVAE · Model vs mouse V1 population profile | Pearson correlation r(TASK): 0.67 -> 0.78 | Improved |
| Add the scaling latent, effect on task modulation (ablation) | TAVAE · Model vs mouse V1 task modulation | Pearson correlation r(TASK-NAIVE): 0.54 -> 0.58 | No clear effect |
| Apply TAVAE task prior instead of VAE on trained mouse data | VAE · Neural predictivity, trial-averaged | neural predictivity: 0.777 -> 0.804 | Improved |
| Apply TAVAE task prior instead of VAE on trained mouse data | VAE · Neural predictivity, trial-based | neural predictivity: 0.379 -> 0.387 | Improved |
| Apply TAVAE task prior instead of VAE on trained mouse data | VAE · CKA, trial-averaged | CKA: 0.82 -> 0.86 | Improved |
Closest papers by meaning
- Multi-Modal Latent Variables for Cross-Individual Primary Visual Cortex Modeling and Analysis (AAAI 2025 · similarity 0.87)
- Time-Evolving Dynamical System for Learning Latent Representations of Mouse Visual Neural Activity (NeurIPS 2025 · similarity 0.87)
- Flexible Context-Driven Sensory Processing in Dynamical Vision Models (NeurIPS 2024 · similarity 0.87)
- Modeling Dynamic Neural Activity by combining Naturalistic Video Stimuli and Stimulus-independent Latent Factors (NeurIPS 2025 · similarity 0.86)
- Poisson Variational Autoencoder (NeurIPS 2024 · similarity 0.85)
- Complex priors and flexible inference in recurrent circuits with dendritic nonlinearities (ICLR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
