Research map / Vision-language and multimodal / Human and embodied modeling
Brain-like Flexible Visual Inference by Harnessing Feedback Feedforward Alignment
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Co-optimize discrimination and reconstruction with mutually trained feedforward and feedback pathways (FFA) instead of training a discriminator with BP | MNIST · MNIST | test accuracy on Gaussian-noisy inputs | Helped |
| Co-optimize discrimination and reconstruction with mutually trained feedforward and feedback pathways (FFA) instead of training a discriminator with BP | MNIST · MNIST | robustness to FGSM adversarial attacks | Helped |
| Use FFA instead of BP for a convolutional classifier | MNIST · MNIST | discrimination performance | Hurt |
| Train a feedback reconstruction pathway jointly with the discriminator using FFA | MNIST convolutional architecture · MNIST | reconstruction performance | Helped |
| Train a feedback reconstruction pathway jointly with the discriminator using FFA | MNIST fully connected architecture · MNIST | reconstruction performance | Mixed |
| Use FFA instead of BP | CIFAR10 ResNet · CIFAR10 | discrimination performance | Hurt |
| Use FFA instead of BP or FA for closed-loop inference | FFA-trained network · MNIST closed-loop denoising | discrimination and reconstruction performance across iterations | Helped |
| Use higher inference noise (β=0.2) instead of lower inference noise (β=0.99) | FFA-trained network · MNIST hallucination from Gaussian-noise initialization | hallucinated digit quality | Helped |
| Increase inference noise | FFA-trained network · MNIST mental imagery | imagined digit quality | Helped |
| Use high-noise rather than low-noise inference for occlusion completion | FFA-trained network · MNIST occlusion completion | resolved image quality | No clear effect |
Closest papers by meaning
- Flexible Context-Driven Sensory Processing in Dynamical Vision Models (NeurIPS 2024 · similarity 0.86)
- Bridging Critical Gaps in Convergent Learning: How Representational Alignment Evolves Across Layers, Training, and Distribution Shifts (NeurIPS 2025 · similarity 0.86)
- Sparse components distinguish visual pathways & their alignment to neural networks (ICLR 2025 · similarity 0.86)
- Biologically Inspired Learning Model for Instructed Vision (NeurIPS 2024 · similarity 0.86)
- CoreMatching: A Co-adaptive Sparse Inference Framework with Token and Neuron Pruning for Comprehensive Acceleration of Vision-Language Models (ICML 2025 · similarity 0.85)
- Forward Learning with Top-Down Feedback: Empirical and Analytical Characterization (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.