Christopher DiMattina, PhD

In this study, we quantitatively investigate human performance on a natural occlusion boundary detection task. We utilize natural stimuli taken from a large database of calibrated color images whose occlusion boundaries were hand-labeled by multiple observers. We measure the statistical properties of occlusion boundaries in both grayscale and color images and compare their statistics to those measured from texture regions taken from the same images. In order to understand the locally available cues which may be utilized by human observers in the detection of occlusion boundaries, we perform a simple two-alternative forced choice experiment on human observers whose task was to discriminate image patches containing occlusion edges and relate human performance to the statistics measured from our databases. We find that local luminance cues are insufficient to explain human performance, and that ideal observer models incorporating textural information are required. Finally, we consider the question of how various models of texture representation solve natural edge-detection and region segmentation tasks which have been well studied in computer vision.

Recent efforts to model higher-order statistical structure in natural images have failed to adequately capture the statistical properties of edges. A recent study by Cadieu and Olshausen suggests that this may be due to the fact that statistical models fail to take into account correlated phase structure in natural images. In this study, we empirically investigate the conditional amplitude and phase correlations which are present in natural images, in particular those observed at occlusion edges. We will define a hierarchical model which learns joint dependencies between amplitude and phase and train this model with naturl images in order to learn a basis which captures the joint dependencies between amplitude and phase.

In recent years, many investigators have trained hierarchical statistical image models in efforts to learn higher-order statistical properties of natural images. However, these studies have generally employed grayscale stimuli. We will use two models developed in our laboratory in order to learn higher-order structure in color images. Of particular interest is whether we can observe sensitivity for edges which are defined by second-order color cues, or combinations of color and texture.

Many previous studies have considered the statistical differences between areas which people fixate when free-viewing natural images, and recent work has indirectly suggested that people may be more likely to fixate edges. However, this has yet to be directly investigated using a database of annotated images, and we are currently doing this experiment using a set of images whose occlusion edges have been hand-labeled by human observers. In addition, we are interested in testing a statistical notion of salience by measuring the Shannon surprisal of natural image patches using hierarchical generative models of natural images.

Sensory neurophysiology experiments can potentially benefit from adaptively choosing stimuli on-line in order to maximize the information gained about the underlying system. It has been shown that this active data collection strategy can greatly reduce the number of trials needed to estimate the parameters of a sensory processing model. Here we demonstrate that for nonlinear neural network models active data collection is not simply useful for speeding up the convergence of estimates, but may in fact be necessary for accurate recovery of parameters. We also present one practical algorithm for combining the dual goals of model estimation and model comparison in on-line experiments and apply it to neural network models which extend and generalize popular linear receptive field models. (pdf)

What is the relationship between the values of the parameters (weights, threholds, etc..) in a neural network and the input-output relationship? We derive a general mathematical condition for there to be a continuum of functionally equivalent neural network models. We find that in general this is only possible when the hidden unit gain functions are given by power, exponential or logarithmic forms. However, since the standard tanh and sigmoid gain functions commonly used in neural modeling may be well approximated by these forms over limited ranges of inputs, one may practically observe a continuum of parameters in these networks giving rise to identical functionality, making unique identification of the network parameters from finite noisy data impossible and suggesting a need for active learning methods when identifying nonlinear models. (pdf)

The exact relationship between a neural circuit's architecture and stimulus-response properties remains poorly understood. Here we obtain general theoretical results for quadratic analysis on multilayer feedforward neural networks and classify all possible quadratic behaviors in relation to the optimal stimulus and invariant stimuli. The optimal stimulus, or the stimulus that best drives a neuron, describes a neuron's selectivity, wheras stimulus invariance describes the generalization of a neuron's response to a continuum of equivalently effective stimuli. Despite the intuitive notion of the optimal stimulus as a peak in the stimulus-response relationship, diverse quadratic behaviors are possible and can occur in the same network. We demonstrate that invariant stimuli are ubiquitous in hierarchical networks, and identify two classes of invariant stimulus transformations which leave neural responses unchanged. (preprint)

Neurophysiologists have long been interested in finding the 'optimal stimulus' for sensory neurons, which is defined as the stimulus which produces the maximum firing rate response. However, it is not clear what constraints the architecture of the underlying neural system which actually generates the responses to sensory stimuli may place on the location of the 'optimal stimulus'. In this study, we analyze feed-forward and recurrent neural network models and discover that for convergent networks whose connections between processing layers form a non-degenerate weight matrix that it is impossible for an strict firing rate maximum to exist in the interior of a compact stimulus space isomorphic to the peripheral representation. This result may explain why many sensory neurons exhibit their strongest responses to stimuli of maximum contrast. (pdf)

Most studies of species-specific vocalization coding have made use of pre-recorded token vocalization stimuli. The inability to systematically manipulate token stimuli makes it difficult to determine which of the many acoustical features present in the vocalization are driving neural responses. Other studies have made use of synthetic vocalizations, but these are typically based on a single exemplar and thus fail to capture the full range of acoustic variability. We utilized a database of thousands of marmoset vocalizations to develop synthetic "Virtual Vocalization" stimuli. These vocalizations can be systematically varied both within and outside of the range of natural acoustical variation along several parameter dimensions and provide a useful tool for the study of vocalization coding in the auditory cortex. (pdf)

It is known that the inhibitory projection from the MNTB to the LSO nucleus in the auditory brainstem exhibits an age-dependent form of long-lasting depression when activated at a low rate. However, since this synapse releases both Glycine and GABA during maturation, the mechanism of this age-dependent synaptic depresssion is unclear. Using GABA(B) receptor anatoginists, it was found that synaptic depression was blocked, suggesting a role for GABA-ergic transmission for inhibitory synaptic depression. Using bath applied BDNF and TRK receptor agonists it was also found that neurotrophins may play a role in this age-dependent inhibitory synaptic depression. My role in this work was the creation of the SLICE software package used to collect the data, and helping to develop methods to analyze of the synaptic data. (pdf)