Worcester Polytechnic Institute

About: Worcester Polytechnic Institute is a education organization based out in Worcester, Massachusetts, United States. It is known for research contribution in the topics: Population & Data envelopment analysis. The organization has 6270 authors who have published 12704 publications receiving 332081 citations. The organization is also known as: WPI.

Reinforcement learning from demonstration through shaping

Abstract: Reinforcement learning describes how a learning agent can achieve optimal behaviour based on interactions with its environment and reward feedback. A limiting factor in reinforcement learning as employed in artificial intelligence is the need for an often prohibitively large number of environment samples before the agent reaches a desirable level of performance. Learning from demonstration is an approach that provides the agent with demonstrations by a supposed expert, from which it should derive suitable behaviour. Yet, one of the challenges of learning from demonstration is that no guarantees can be provided for the quality of the demonstrations, and thus the learned behavior. In this paper, we investigate the intersection of these two approaches, leveraging the theoretical guarantees provided by reinforcement learning, and using expert demonstrations to speed up this learning by biasing exploration through a process called reward shaping. This approach allows us to leverage human input without making an erroneous assumption regarding demonstration optimality. We show experimentally that this approach requires significantly fewer demonstrations, is more robust against suboptimality of demonstrations, and achieves much faster learning than the recently developed HAT algorithm.

Sites of Rupture in Human Atherosclerotic Carotid Plaques Are Associated With High Structural Stresses An In Vivo MRI-Based 3D Fluid-Structure Interaction Study

Abstract: Background and Purpose— It has been hypothesized that high structural stress in atherosclerotic plaques at critical sites may contribute to plaque disruption. To test that hypothesis, 3D fluid-structure interaction models were constructed based on in vivo MRI data of human atherosclerotic carotid plaques to assess structural stress behaviors of plaques with and without rupture. Methods— In vivo MRI data of carotid plaques from 12 patients scheduled for endarterectomy were acquired for model reconstruction. Histology confirmed that 5 of the 12 plaques had rupture. Plaque wall stress (PWS) and flow maximum shear stress were extracted from all nodal points on the lumen surface of each plaque for analysis. A critical PWS (maximum of PWS values from all possible vulnerable sites) was determined for each plaque. Results— Mean PWS from all ulcer nodes in ruptured plaques was 86% higher than that from all nonulcer nodes (123.0 versus 66.3 kPa, P<0.0001). Mean flow maximum shear stress from all ulcer nodes in rupt...

Perceived Stakeholder Influences and Organizations' Use of Environmental Audits

Abstract: While the use of internal, external, and both types of environmental audits are becoming more pervasive in society, little is known about the stakeholder influences associated with their use, in large part because previous research has viewed them as a uniform type of management practice. This study draws on stakeholder theory to explore organizations' use of different types of environmental audits. It uses international manufacturing data to show that significant variations in the use of environmental audits are associated with differences in stakeholder influences, and that a more nuanced treatment is needed when evaluating these audits.

Efficient Algorithms for Elliptic Curve Cryptosystems

Abstract: This contribution describes three algorithms for efficient implementations of elliptic curve cryptosystems. The first algorithm is an entirely new approach which accelerates the multiplications of points which is the core operation in elliptic curve public-key systems. The algorithm works in conjunction with the k-ary or sliding window method. The algorithm explores computational advantages by computing repeated point doublings directly through closed formulae rather than from individual point doublings. This approach reduces the number of inversions in the underlying finite field at the cost of extra multiplications. For many practical implementations, where field inversion is at least four times as costly as field multiplication, the new approach proofs to be faster than traditional point multiplication methods. The second algorithm deals with efficient inversion in composite Galois fields of the form GF((2n)n). Based on an idea of Itoh and Tsujii, we optimize the algorithm for software implementation of elliptic curves. The algorithm reduced inversion in the composite field to inversion in the subfield GF(2n). The third algorithm describes the application of the Karatsuba-Ofman Algorithm to multiplication in GF((2n)n). We provide a detailed complexity analysis of the algorithm for the case that subfield arithmetic is performed through table look-up. We apply all three algorithms to an implementation of an elliptic curve system over GF((216)11). We provide absolute performance measures for the field operations and for an entire point multiplication.