Rapid de novo shape encoding: a challenge to connectionist modeling

TL;DR

This paper challenges traditional connectionist models by highlighting their inability to encode and recognize new shapes immediately, as humans do, despite advanced models achieving invariance through training.

Contribution

It presents evidence that human perceptual skills can encode unknown shapes instantly, posing a challenge to existing models that need extensive training.

Findings

Humans can encode unknown shapes after a single exposure.

Connectionist models require many training trials for recognition.

Advanced models achieve invariance through training, not immediate recognition.

Abstract

Neural network (connectionist) models are designed to encode image features and provide the building blocks for object and shape recognition. These models generally call for: a) initial diffuse connections from one neuron population to another, and b) training to bring about a functional change in those connections so that one or more high-tier neurons will selectively respond to a specific shape stimulus. Advanced models provide for translation, size, and rotation invariance. The present discourse notes that recent work on human perceptual skills has demonstrated immediate encoding of unknown shapes that were seen only once. Further, the perceptual mechanism provided for translation, size, and rotation invariance. This finding represents a challenge to connectionist models that require many training trials to achieve recognition and invariance.