Unverified50% confidenceFactExact time
稀疏编码理论由Olshausen与Field于1996年提出，视觉皮层神经元通过稀疏激活高效表征自然图像
1
Sources
50%
Confidence
Long-term
Relevance
7/15/2026
First Seen
Sources
Related Claims
UnverifiedOlshausen和Field在1996年的工作表明，V1神经元的感受野与ICA在自然图像上学到的基函数高度吻合76% similarUnverifiedAnthropic's sparse autoencoder 'feature' research is a neuron-level interpretability study that attempts to disentangle distributed representations.75% similarUnverifiedIn image recognition, shallow neurons respond to low-level features such as edges and textures, while deep neurons respond to overall concepts.71% similarUnverifiedConvolutional Neural Networks (CNNs), inspired by the structure of the visual cortex, extract local features by sliding convolutional kernels across images, with the parameter sharing mechanism significantly reducing computational load.68% similarUnverifiedIn 2015, Gatys et al. proposed neural style transfer, using the Gram matrix of convolutional neural networks to match the feature distributions of content and style images64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/523915API
curl https://kongchang.com/api/v1/knowledge/claims/523915MCP
get_claim(id=523915)