Articles

Environmental and Economic Prospects of Low-carbon Vehicles in Support of European Commission 2030 City Logistics Fleet Goals

Giordano A.

PhD Thesis

2020

Abstract:

Active learning (AL) methods create an optimized labeled training set from unlabeled data. We introduce a novel online active deep learning method for medical image analysis. We extend our MedAL AL framework to present new results in this paper. A novel sampling method queries the unlabeled examples that maximize the average distance to all training set examples. Our online method enhances performance of its underlying baseline deep network. These novelties contribute to significant performance improvements, including improving the model’s underlying deep network accuracy by 6.30%, using only 25% of the labeled dataset to achieve baseline accuracy, reducing backpropagated images during training by as much as 67%, and demonstrating robustness to class imbalance in binary and multiclass tasks. This article is categorized under: Technologies > Machine Learning Technologies > Classification Application Areas > Health Care

Evaluating the benefits of commercial building energy codes and improving federal incentives for code adoption

Gilbraith N., Azevedo I.L., Jaramillo P.

Environmental Science and Technology

2014

Abstract:

The federal government has the goal of decreasing commercial building energy consumption and pollutant emissions by incentivizing the adoption of commercial building energy codes. Quantitative estimates of code benefits at the state level that can inform the size and allocation of these incentives are not available. We estimate the state-level climate, environmental, and health benefits (i.e., social benefits) and reductions in energy bills (private benefits) of a more stringent code (ASHRAE 90.1–2010) relative to a baseline code (ASHRAE 90.1–2007). We find that reductions in site energy use intensity range from 93 MJ/m2 of new construction per year (California) to 270 MJ/m2 of new construction per year (North Dakota). Total annual benefits from more stringent codes total $506 million for all states, where $372 million are from reductions in energy bills, and $134 million are from social benefits. These total benefits range from $0.6 million in Wyoming to $49 million in Texas. Private benefits range from $0.38 per square meter in Washington State to $1.06 per square meter in New Hampshire. Social benefits range from $0.2 per square meter annually in California to $2.5 per square meter in Ohio. Reductions in human/environmental damages and future climate damages account for nearly equal shares of social benefits.

Evaluation of word vector representations by subspace alignment

Tsvetkov Y., Faruqui M., Ling W., Lample G., Dyer C.

Conference Proceedings - EMNLP 2015: Conference on Empirical Methods in Natural Language Processing

2015

Abstract:

Unsupervisedly learned word vectors have proven to provide exceptionally effective features in many NLP tasks. Most common intrinsic evaluations of vector quality measure correlation with similarity judgments. However, these often correlate poorly with how well the learned representations perform as features in downstream evaluation tasks. We present QVEC—a computationally inexpensive intrinsic evaluation measure of the quality of word embeddings based on alignment to a matrix of features extracted from manually crafted lexical resources—that obtains strong correlation with performance of the vectors in a battery of downstream semantic evaluation tasks.1

Event-based summarization using a centrality-as-relevance model

Marujo L., Ribeiro R., Gershman A., de Matos D.M., Neto J.P., Carbonell J.

Knowledge and Information Systems

2016

Abstract:

Event detection is a fundamental information extraction task, which has been explored largely in the context of question answering, topic detection and tracking, knowledge base population, news recommendation, and automatic summarization. In this article, we explore an event detection framework to improve a key phrase-guided centrality-based summarization model. Event detection is based on the fuzzy fingerprint method, which is able to detect all types of events in the ACE 2005 Multilingual Corpus. Our base summarization approach is a two-stage method that starts by extracting a collection of key phrases that will be used to help the centrality-as-relevance retrieval model. We explored three different ways to integrate event information, achieving state-of-the-art results in text and speech corpora: (1) filtering of nonevents, (2) event fingerprints as features, and (3) combination of filtering of nonevents and event fingerprints as features.

Evolutionary learning methodology: A case study of R&D strategy development

Hooshangi S., Arasti M. R., Hounshell D. A., Sahebzamani S.

Technological Forecasting and Social Change

2013

Abstract:

This article concerns the notion of methodology in strategic management of R&D/technology. Though development of new tools and methods has received much attention during the recent decades, attention to understanding methodologies has remained disproportionally low. In this study we distinguish two methodologies that are used in strategic management of R&D/technology: planning methodology and evolutionary learning methodology. We mainly focus on defining and describing the origins, nature, and characteristics of the latter. We propose a framework for methodology selection by investigating context, content, and process factors. Using this framework, we provide supportive evidence for appropriateness of evolutionary learning methodology to develop a robust R&D strategy for Iran’s power industry. We then describe the details of operationalizing the methodology for the Iranian power industry. This study is particularly focused on delineating how evolutionary learning methodology can be applied as an effective framework to improve the formation method and content of R&D strategy. We conclude that methodological knowledge can provide a powerful lens with which to understand performance of methods, and we suggest that evolutionary learning methodology is particularly appropriate for the following situations: when the environment is uncertain or fast changing, when there exist many stakeholders with conflicting interests, and when a method needs to be applied in a context other than the one for which it was initially developed.

Exact rate for convergence in probability of averaging processes via generalized min-cut

Bajovic D., Xavier J., Moura J.M.F., Sinopoli B.

Proceedings of the IEEE Conference on Decision and Control

2012

Abstract:

We study the asymptotic exponential decay rate I for the convergence in probability of products W k W k-1 …W 1 of random symmetric, stochastic matrices W k . Albeit it is known that the probability P that the product W k W k-1 …W 1 is ∈ away from its limit converges exponentially fast to zero, i.e., P ~ e -kI , the asymptotic rate I has not been computed before. In this paper, assuming the positive entries of Wk are bounded away from zero, we explicitly characterize the rate I and show that it is a function of the underlying graphs that support the positive (non zero) entries of W k . In particular, the rate I is given by a certain generalization of the min-cut problem. Although this min-cut problem is in general combinatorial, we show how to exactly compute I in polynomial time for the commonly used matrix models, gossip and link failure. Further, for a class of models for which I is difficult to compute, we give easily computable bounds: I ≤ I ≤ I̅, where I and I̅ differ by a constant ratio. Finally, we show the relevance of I as a system design metric with the example of optimal power allocation in consensus+innovations distributed detection.

Existence of ground states of nonlocal-interaction energies

Simione R., Slepcev D., Topaloglu I.

Journal of Statistical Physics

2015

Abstract:

We investigate which nonlocal-interaction energies have a ground state (global minimizer). We consider this question over the space of probability measures and establish a sharp condition for the existence of ground states. We show that this condition is closely related to the notion of stability (i.e. -stability) of pairwise interaction potentials. Our approach uses the direct method of the calculus of variations.

Experimental study on the impact of vehicular obstructions in VANETs

Meireles R., Boban M., Steenkiste P., Tonguz O., Barros J.

2010 IEEE Vehicular Networking Conference, VNC 2010

2010

Abstract:

Channel models for vehicular networks typically disregard the effect of vehicles as physical obstructions for the wireless signal. We aim to clarify the validity of this simplification by quantifying the impact of obstructions through a series of wireless experiments. Using two cars equipped with Dedicated Short Range Communications (DSRC) hardware designed for vehicular use, we perform experimental measurements in order to collect received signal power and packet delivery ratio information in a multitude of relevant scenarios: parking lot, highway, suburban and urban canyon. Upon separating the data into line of sight (LOS) and non-line of sight (NLOS) categories, our results show that obstructing vehicles cause significant impact on the channel quality. A single obstacle can cause a drop of over 20 dB in received signal strength when two cars communicate at a distance of 10 m. At longer distances, NLOS conditions affect the usable communication range, effectively halving the distance at which communication can be achieved with 90% chance of success. The presented results motivate the inclusion of vehicles in the radio propagation models used for VANET simulation in order to increase the level of realism.

Experiments with Vizzy as a Coach for Elderly Exercise

Avelino J., Simão H., Ribeiro R., Moreno P., Figueiredo r., Duarte N., Nunes R., Bernardino A., Čaić M., Mahr D., Odekerken-Schröder G.

PREC2018, March 2018, Chicago, IL, USA

2018

Abstract:

Vizzy is a wheeled humanoid robot carefully designed for an enjoyable / pleasurable interaction with humans. In this paper evaluate Vizzy’s application as an exercise coach according to user’s perceived robot aesthetics, trust, confidence, and enjoyment. We describe the proposed robotic platform skills, interfaces, and interaction modes. We have deployed Vizzy in three care centers in Portugal for the promotion of exercise activities and tested its performance with 36 elderly users. Survey data collected after the interaction with the robot show a good acceptance of the robotic platform as an exercise coach. From observations and short interviews, we also extracted a set of guidelines that led to several improvements and ideas for future work.