A Theory of How the Brain Might Work

I wish to propose a quite speculative new version of the grandmother cell theory to explain how the brain, or parts of it, may work. In particular, I discuss how the visual system may learn to recognize 3D objects. The model would apply directly to the cortical cells involved in visual face recognition. I will also outline the relation of our theory to existing models of the cerebellum and of motor control. Specific biophysical mechanisms can be readily suggested as part of a basic type of neural circuitry that can learn to approximate multidimensional input-output mappings from sets of examples and that is expected to be replicated in different regions of the brain and across modalities. The main points of the theory are: -the brain uses modules for multivariate function approximation as basic components of several of its information processing subsystems. -these modules are realized as HyperBF networks (Poggio and Girosi, 1990a,b). -HyperBF networks can be implemented in terms of biologically plausible mechanisms and circuitry. The theory predicts a specific type of population coding that represents an extension of schemes such as look-up tables. I will conclude with some speculations about the trade-off between memory and computation and the evolution of intelligence.

Trajectory Mapping ("TM''): A New Non-Metric Scaling Technique

Trajectory Mapping "TM'' is a new scaling technique designed to recover the parameterizations, axes, and paths used to traverse a feature space. Unlike Multidimensional Scaling (MDS), there is no assumption that the space is homogenous or metric. Although some metric ordering information is obtained with TM, the main output is the feature parameterizations that partition the given domain of object samples into different categories. Following an introductory example, the technique is further illustrated using first a set of colors and then a collection of textures taken from Brodatz (1966).

Multisensor Modeling Underwater with Uncertain Information

This thesis develops an approach to the construction of multidimensional stochastic models for intelligent systems exploring an underwater environment. It describes methods for building models by a three- dimensional spatial decomposition of stochastic, multisensor feature vectors. New sensor information is incrementally incorporated into the model by stochastic backprojection. Error and ambiguity are explicitly accounted for by blurring a spatial projection of remote sensor data before incorporation. The stochastic models can be used to derive surface maps or other representations of the environment. The methods are demonstrated on data sets from multibeam bathymetric surveying, towed sidescan bathymetry, towed sidescan acoustic imagery, and high-resolution scanning sonar aboard a remotely operated vehicle.

Image-Based View Synthesis

We present a new method for rendering novel images of flexible 3D objects from a small number of example images in correspondence. The strength of the method is the ability to synthesize images whose viewing position is significantly far away from the viewing cone of the example images ("view extrapolation"), yet without ever modeling the 3D structure of the scene. The method relies on synthesizing a chain of "trilinear tensors" that governs the warping function from the example images to the novel image, together with a multi-dimensional interpolation function that synthesizes the non-rigid motions of the viewed object from the virtual camera position. We show that two closely spaced example images alone are sufficient in practice to synthesize a significant viewing cone, thus demonstrating the ability of representing an object by a relatively small number of model images --- for the purpose of cheap and fast viewers that can run on standard hardware.