Entity

Time filter

Source Type

Boulder, CO, United States

Dobres J.,Adaptive Cognitive Systems | Seitz A.R.,University of California at Riverside
Journal of Vision | Year: 2010

Classication image analysis is a psychophysical technique in which noise components of stimuli are analyzed to produce an image that reveals critical features of a task. Here we use classication images to gain greater understanding of perceptual learning. To achieve reasonable classication images within a single session, we developed an efcient classication image procedure that employed designer noise and a low-dimensional stimulus space. Subjects were trained across ten sessions to detect the orientation of a grating masked in noise, with an eleventh, test, session conducted using a stimulus orthogonal to the trained stimulus. As with standard perceptual learning studies, subjects showed improvements in performance metrics of accuracy, threshold, and reaction times. The clarity of the classication images and their correlation to an ideal target also improved across training sessions in an orientation-specic manner. Furthermore, image-based analyses revealed aspects of performance that could not be observed with standard performance metrics. Subjects with threshold improvements learned to use pixels across a wider area of the image, and, apposed to subjects without threshold improvements, showed improvements in both the bright and dark parts of the image. We conclude that classication image analysis is an important complement to traditional metrics of perceptual learning. © ARVO.


Best B.J.,Adaptive Cognitive Systems
Computational and Mathematical Organization Theory | Year: 2013

We developed an end-to-end process for inducing models of behavior from expert task performance through in-depth case study. A subject matter expert (SME) performed navigational and adversarial tasks in a virtual tank combat simulation, using the dTank and Unreal platforms. Using eye tracking and Cognitive Task Analysis, we identified the key goals pursued by and attributes used by the SME, including reliance on an egocentric spatial representation, and on the fly re-representation of terrain in qualitative terms such as "safe" and "risky". We demonstrated methods for automatic extraction of these qualitative higher-order features from combinations of surface features present in the simulation, producing a terrain map that was visually similar to the SME annotated map. The application of decision-tree and instance-based machine learning methods to the transformed task data supported prediction of SME task selection with greater than 95 % accuracy, and SME action selection at a frequency of 10 Hz with greater than 63 % accuracy, with real time constraints placing limits on algorithm selection. A complete processing model is presented for a path driving task, with the induced generative model deviating from the SME chosen path by less than 2 meters on average. The derived attributes also enabled environment portability, with path driving models induced from dTank performance and deployed in Unreal demonstrating equivalent accuracy to those induced and deployed completely within Unreal. © 2012 The Author(s).


Dobres J.,Adaptive Cognitive Systems
Journal of vision | Year: 2010

Classification image analysis is a psychophysical technique in which noise components of stimuli are analyzed to produce an image that reveals critical features of a task. Here we use classification images to gain greater understanding of perceptual learning. To achieve reasonable classification images within a single session, we developed an efficient classification image procedure that employed designer noise and a low-dimensional stimulus space. Subjects were trained across ten sessions to detect the orientation of a grating masked in noise, with an eleventh, test, session conducted using a stimulus orthogonal to the trained stimulus. As with standard perceptual learning studies, subjects showed improvements in performance metrics of accuracy, threshold, and reaction times. The clarity of the classification images and their correlation to an ideal target also improved across training sessions in an orientation-specific manner. Furthermore, image-based analyses revealed aspects of performance that could not be observed with standard performance metrics. Subjects with threshold improvements learned to use pixels across a wider area of the image, and, apposed to subjects without threshold improvements, showed improvements in both the bright and dark parts of the image. We conclude that classification image analysis is an important complement to traditional metrics of perceptual learning.


Grant
Agency: Department of Defense | Branch: Navy | Program: SBIR | Phase: Phase II | Award Amount: 741.56K | Year: 2010

A tremendous need exists for intelligent agents that can be created and edited without resorting to intensive knowledge engineering and programming, and which exhibit believable and variable behavior in the training contexts in which they are deployed. This proposal describes a novel method for creating and editing intelligent agents’ behavior based on using instance-based modeling and statistical learning methods that learn from the example of a person interacting in a virtual environment. These methods, which leverage structured knowledge in a hybrid symbolic-subsymbolic approach, support automatic incorporation of assessment feedback directly from the interface into an agent, allowing a domain expert to interact with an agent in a closed feedback loop through a participation in a virtual environment, instead of through lengthy reprogramming by a knowledge engineering expert.


Grant
Agency: National Science Foundation | Branch: | Program: SBIR | Phase: Phase I | Award Amount: 99.83K | Year: 2008

The Small Business Innovation Research (SBIR) Phase I research project aims to create a toolkit embodying cognitive capabilities for use in developing intelligent agents. These agents would provide human-like interactions with software, for desktop productivity, research, and gaming domains, by observing human interactions with the system and mimicking those interactions. Current approaches for embedding intelligent agents such as finite state machines and rule-based systems are often limited by either brittleness, or by difficulties in knowledge engineering, or often by both. In contrast, state of the art cognitive modeling approaches combine symbolic rule-based approaches with numeric statistical machine learning techniques, and do so in a computationally scalable way. The specific research objectives are: 1) exploring variations in instance-based learning techniques and their ability to simulate human learning and their computational implications; 2) examining using an expert system to elicit knowledge and produce a task skeleton for organizing knowledge; 3) exploring plan recognition techniques for mapping a stream of human behavior onto the elicited task structure; 4) exploring the extraction of strong knowledge from segmented human performance data through statistical learning techniques; and 5) developing techniques for remediating developed systems so that deficiencies noted can be translated directly into improved agent behavior. The proposed toolkit will automate computer desktop tasks, thereby enhancing productivity, and will produce gaming agents without programming, thereby satisfying the need for greater numbers of robust, believable non-player characters. For the currently installed base of PCs is estimated at 898 million units, with yearly worldwide sales at 190 million units, and with the worldwide gaming market estimated at approximately $20 billion, the proposed work will provide easier automation - through observing competent behavior rather than through programming - to both of these markets. The proposed technology is applicable to other domains not addressed specifically in this proposal such as the assistive market to produce an assistant for the handicapped that learns typical sequences of interface actions and offers to complete those actions. Additionally, the technology can also aid in building training systems where the task is collaborative and the cost of using human team mates is prohibitive.

Discover hidden collaborations