GOAL-DIRECTED MODEL ACQUISITION AND RENDERING

点击翻译按钮获取中文摘要

The long term goal of this project is to develop a new interactive paradigm for data collection and analysis in the neurosciences. The specific aim is to create a "teleological pipeline" for making datasets, models, and images. The pipeline algorithmically incorporates data and image "goals" of the investigator, expressing and executing the goals through interactive computational means. The user would make man)' data acquisition and image decisions interactively, closing the loop while the experiment is in progress. The teleological pipeline will be applied to Magnetic Resonance Micro- imaging. The pipeline is roughly analogous to a "spread sheet environment" for MR data collection and analysis. The teleological paradigm helps to convert "what the user wants" into constrained optimization problems, and other tasks that are computationally tractable. The approach builds on Cal tech expertise in teleological modeling [see reprints]. The teleological pipeline is a robust conversion process, converting goals into computational procedures. Examples are tools which use constrained optimization methods (as opposed to more heuristic methods). The approach involves creating soft and hard constraints to determine computational tasks that achieve what the user has requested. 'Closing the loop' interactively through use of a teleological pipeline may have a fundamental impact in the neurosciences, perhaps as significant as the transition from computer punch cards to interactive computer graphics CRT screens.