Performance-based Algorithm Research Articles

Moving-Target Intelligent Tutoring System for Marksmanship Training

Intelligent tutoring systems (ITSs) may augment military training systems and mitigate existing limitations in training personnel and resources. A study was conducted to investigate the effectiveness of an embedded rifle marksmanship ITS for Moving Targets (MT-ITS). MT-ITS has two main components: (1) a Smart Sight System that provides a perceptual cue to help trainees adjust their point of aim to account for a target’s speed, direction of movement, and distance, and (2) a performance-based algorithm that delivers shooting performance feedback to trainees.The MT-ITS was tested in an experiment where participants engaged moving targets in a virtual shooting range. Moving targets were presented at different speeds, direction of movement, and distances. Two types of marksmanship training were compared: with ITS and without ITS (a standard training). The ITS training group produced better hit rate and aiming accuracy scores than the standard training group, requiring less practice to achieve asymptotic results. Implications for the design of embedded trainers with ITS for marksmanship specifically and for training motor skills in general are discussed in the context of future research directions.

A novel approach for evaluating contact patterns and risk mitigation strategies for COVID-19 in English primary schools with application of structured expert judgement.

Personal contacts drive COVID-19 infections. After being closed (23 March 2020) UK primary schools partially re-opened on 1 June 2020 with social distancing and new risk mitigation strategies. We conducted a structured expert elicitation of teachers to quantify primary school contact patterns and how contact rates changed upon re-opening with risk mitigation measures in place. These rates, with uncertainties, were determined using a performance-based algorithm. We report mean number of contacts per day for four cohorts within schools, with associated 90% confidence ranges. Prior to lockdown, younger children (Reception and Year 1) made 15 contacts per day [range 8.35] within school, older children (Year 6) 18 contacts [range 5.55], teaching staff 25 contacts [range 4.55] and non-classroom staff 11 contacts [range 2.27]. After re-opening, the mean number of contacts was reduced by 53% for young children, 62% for older children, 60% for classroom staff and 64% for other staff. Contacts between teaching and non-teaching staff reduced by 80%. The distributions of contacts per person are asymmetric with heavy tail reflecting a few individuals with high contact numbers. Questions on risk mitigation and supplementary structured interviews elucidated how new measures reduced daily contacts in-school and contribute to infection risk reduction.

Performance-based semi-active control algorithm for protecting base isolated buildings from near-fault earthquakes

Base isolated structures have been found to be at risk in near-fault regions as a result of long period pulses that may exist in near-source ground motions. Various control strategies, including passive, active and semi-active control systems, have been investigated to overcome this problem. This study focuses on the development of a semi-active control algorithm based on several performance levels anticipated from an isolated building during different levels of ground shaking corresponding to various earthquake hazard levels. The proposed performance-based algorithm is based on a modified version of the well-known semi-active skyhook control algorithm. The proposed control algorithm changes the control gain depending on the level of shaking imposed on the structure. The proposed control system has been evaluated using a series of analyses performed on a base isolated benchmark building subjected to seven pairs of scaled ground motion records. Simulation results show that the newly proposed algorithm is effective in improving the structural and nonstructural performance of the building for selected earthquakes.

Neural Networks Prediction of Soil Hydraulic Functions for Alluvial Soils Using Multistep Outflow Data

Indirect methods for prediction of soil hydraulic properties play an important role in understanding site-specific unsaturated water flow and transport processes, usually via numerical simulation models. Specifically, pedotransfer functions (PTFs) to predict soil-water retention have been widely developed. However, few datasets that include unsaturated hydraulic conductivity data are available for prediction purposes. Moreover, those available employ a variety of measurement techniques. We show that prediction of soil-water retention and unsaturated hydraulic conductivity curves from basic soil properties can be improved if hydraulic data are determined by a single measurement method that is consistently applied to all soil samples. Here, we present a unique dataset that consists of 310 soil-water retention and unsaturated hydraulic conductivity functions, all of which were obtained from the multistep outflow method. With this dataset, neural networks coupled with bootstrap aggregation were used to predict the soil-water retention and hydraulic conductivity characteristics from basic soil properties, that is, sand, silt, and clay content, bulk density (ρb), saturated water content, and saturated hydraulic conductivity. The prediction errors of water content were about 3 to 4% by volume. Unsaturated hydraulic conductivity predictions improved significantly when a so-called performance-based algorithm was utilized to minimize residuals of soil hydraulic data rather than hydraulic parameters. The root mean squared of residuals for predicted values of water content and unsaturated hydraulic conductivity were reduced by about 50% when compared with predicted hydraulic functions using a published neural networks program Rosetta Results from a sensitivity analysis suggest that the hydraulic parameters are mostly sensitive to sand content and saturated water content.

Performance-based variable selection for scoring models

The performance of a direct marketing scoring model at a particular mailing depth, d, is usually measured by the total amount of revenue generated by sending an offer to the customers with the 100 d% largest scores (predicted values). Commonly used variable selection algorithms optimize some function of model fit (squared difference between training and predicted values). This article (1) discusses issues involved in selecting a mailing depth, d, and (2) proposes a variable selection algorithm that optimizes the performance as the primary objective. The relationship between fit and performance is discussed. The performance-based algorithm is compared with fit-based algorithms using two real direct marketing data sets. These experimental results indicate that performance-based variable selection is 3–4% better than corresponding fit-based models, on average, when the mailing depth is between 20% and 40%.