---
title: Local Weight Sharing
type: vocabulary
url: "https://www.envisioning.com/vocab/local-weight-sharing"
summary: Reusing the same weights across spatial positions to detect patterns regardless of location.
year: 1989
generality: 0.69
---

# Local Weight Sharing

Reusing the same weights across spatial positions to detect patterns regardless of location.
Local weight sharing is a design principle in convolutional neural networks (CNNs) in which a single set of learned weights, forming a convolutional filter or kernel, is applied repeatedly across different spatial positions of an input. Instead of assigning independent parameters to every connection in the network, the same filter slides across the input and computes dot products at each location. This reduces the total number of trainable parameters compared to fully connected architectures, which makes models more practical to train on high-dimensional inputs like images.

The mechanism relies on an assumption about natural data: that meaningful patterns such as edges, textures, and shapes can appear anywhere in an input, and the detector for such a pattern should not need to be relearned for every possible position. A filter trained to detect a horizontal edge in the top-left corner of an image is equally useful when that edge appears in the center or bottom-right. This property, known as translation equivariance, emerges directly from weight sharing and is one of the primary reasons CNNs generalize to visual tasks.

Local weight sharing became central to practical deep learning through Yann LeCun's development of the LeNet architecture in the late 1980s and early 1990s, which applied the principle to handwritten digit recognition. The approach built on earlier ideas from Kunihiko Fukushima's neocognitron, but LeCun's integration of backpropagation with shared convolutional weights made the technique trainable end-to-end and scalable.

Beyond image recognition, local weight sharing has influenced architectures across domains, from 1D convolutions in audio and text processing to 3D convolutions in video analysis. The principle also informs some attention-based models, where locality constraints are imposed to improve efficiency. Local weight sharing accounts for much of the parameter efficiency of CNNs and illustrates how the inductive biases built into an architecture can matter as much as raw model capacity.

---
Source: Envisioning — Technology Research Institute (https://www.envisioning.com/vocab/local-weight-sharing)