Human Symmetry Uncertainty Detected by a Self-Organizing Neural Network Map

Abstract

Symmetry in biological and physical systems is a product of self-organization driven by evolutionary processes, or mechanical systems under constraints. Symmetry-based feature extraction or representation by neural networks may unravel the most informative contents in large image databases. Despite significant achievements of artificial intelligence in recognition and classification of regular patterns, the problem of uncertainty remains a major challenge in ambiguous data. In this study, we present an artificial neural network that detects symmetry uncertainty states in human observers. To this end, we exploit a neural network metric in the output of a biologically inspired Self-Organizing Map Quantization Error (SOM-QE). Shape pairs with perfect geometry mirror symmetry but a non-homogenous appearance, caused by local variations in hue, saturation, or lightness within and/or across the shapes in a given pair produce, as shown here, a longer choice response time (RT) for “yes” responses relative to symmetry. These data are consistently mirrored by the variations in the SOM-QE from unsupervised neural network analysis of the same stimulus images. The neural network metric is thus capable of detecting and scaling human symmetry uncertainty in response to patterns. Such capacity is tightly linked to the metric’s proven selectivity to local contrast and color variations in large and highly complex image data.

Color Variability Constrains Detection of Geometrically Perfect Mirror Symmetry

“…As soon as a response was given, the current image disappeared from the screen, and 800 milliseconds later the next image was delivered. Image presentation and response data encoding were, as in our previous work [45,59], controlled by a program written in Python 2.7 for Windows using the Spyder 2.0 environment. The Python codepages relative to image presentation and experimental session control are provided in the Figures S2 and S3 of the Supplementary Materials Section.…”

“…The latter uses self-organizing biological learning [42] to adapt to steadily changing external environments of the physical world. Neural network models based on self-organizing visual mapping [43,44] have proven sensitive to mirror symmetry uncertainty in visual color patterns [45,46] in a similar way as the human perceptual system [45]. The emerging consensus under the light of current state of the art biological vision converges towards the assumption that symmetry detection is accounted for by a multiple-channel model, where the response of each channel represents a combination of the output of early symmetry detectors (mechanisms) and the output of additional processing resources required to deal with external and/or intrinsic sources of noise [22,47].…”

From Biological Synapses to “Intelligent” Robots

“…They are derived from insights in neurobiology, cognitive neuroscience, physics, and, in particular, Grossberg's [2] adaptive resonance theory (ART), which provides a mechanistic, mathematically supported, account of how self-organization achieves stability and functional plasticity while minimizing structural system complexity. The principle is exploited in Kohonen's [107] self-organizing map, a computationally parsimonious example of self-organizing, brain-inspired artificial neural network (ANN) recently employed in simulations of brain-like sensory learning for automatic (sensor or robot driven) detection of microscopic changes in physical environments [108][109][110][111][112]. The SOM has a functional architecture that formally corresponds to the nonlinear, ordered, smooth mapping of high-dimensional input data to representations in terms of a regular, low-dimensional array [107].…”

Making and breaking symmetries in mind and life

“…Indeed, when we talk about the preconditions for ‘System 2’ cognition with the power of conceptual understanding through abstraction [37,38], or even our ability to generate stable percepts, are such phenomena best understood as kinds of informational symmetries? In a different direction, we may ask: to what extent does the preference for symmetry mentioned earlier come ‘for free’ via predictive coding, where symmetric structures may be easier to predict or compress, and where efficient prediction-error minimization or compression constitutes a foundation for valence for living organisms [39–41]?…”