Mohammed Maniruzzaman

Learning semantics-preserving distance metrics for clustering graphical data

In mining graphical data the default Euclidean distance is often used as a notion of similarity. However this does not adequately capture semantics in our targeted domains, having graphical representations depicting results of scientific experiments. It is seldom known a-priori what other distance metric best preserves semantics. This motivates the need to learn such a metric. A technique called LearnMet is proposed here to learn a domain-specific distance metric for graphical representations. Input to LearnMet is a training set of correct clusters of such graphs. LearnMet iteratively compares these correct clusters with those obtained from an arbitrary but fixed clustering algorithm. In the first iteration a guessed metric is used for clustering. This metric is then refined using the error between the obtained and correct clusters until the error is below a given threshold. LearnMet is evaluated rigorously in the Heat Treating domain which motivated this research. Clusters obtained using the learned metric and clusters obtained using Euclidean distance are both compared against the correct clusters over a separate test set. Our results show that the learned metric provides better clusters.

Quenchml: A semantics-preserving markup language for knowledge representation in quenching

Abstract Knowledge representation (KR) is an important area in artificial intelligence (AI) and is often related to specific domains. The representation of knowledge in domain-specific contexts makes it desirable to capture semantics as domain experts would. This motivates the development of semantics-preserving standards for KR within the given domain. In addition to the storage and analysis of information using such standards, the effect of globalization today necessitates the publishing of information on the Web. Thus, it is advisable to use formats that make the information easily publishable and accessible while developing KR standards. In this article, we propose such a standard called Quenching Markup Language (QuenchML). This follows the syntax of the eXtensible Markup Language and captures the semantics of the quenching domain within the heat treating of materials. We describe the development of QuenchML, a multidisciplinary effort spanning the realms of AI, database management, and materials science, considering various aspects such as ontology, data modeling, and domain-specific constraints. We also explain the usefulness of QuenchML in semantics-preserving information retrieval and in text mining guided by domain knowledge. Furthermore, we outline the significance of this work in software tools within the field of AI.

The Effect of Agitation and Quenchant Temperature on the Heat Transfer Coefficients for 6061 Aluminum Alloy Quenched in Distilled Water

A 6061 aluminum alloy probe was quenched using a Center for Heat Treating Excellence probe-quenching system in distilled water while varying bath temperature and the agitation level. Time versus temperature data were collected during the quench using a thermocouple embedded inside the probe. The surface heat transfer coefficients as a function of temperature were calculated using the Newtonian cooling approximation. The maximum heat transfer coefficient values ranged from 1054 W/m2⋅K for 100°C water with no agitation to 3822 W/m2⋅K for 5°C water with the agitation of 1850 r/min. The data also showed that at higher levels of agitation and lower bath temperatures, the maximum heat transfer coefficient increased.

Comparing mathematical and heuristic approaches for scientific data analysis

Abstract Scientific data is often analyzed in the context of domain-specific problems, for example, failure diagnostics, predictive analysis, and computational estimation. These problems can be solved using approaches such as mathematical models or heuristic methods. In this paper we compare a heuristic approach based on mining stored data with a mathematical approach based on applying state-of-the-art formulae to solve an estimation problem. The goal is to estimate results of scientific experiments given their input conditions. We present a comparative study based on sample space, time complexity, and data storage with respect to a real application in materials science. Performance evaluation with real materials science data is also presented, taking into account accuracy and efficiency. We find that both approaches have their pros and cons in computational estimation. Similar arguments can be applied to other scientific problems such as failure diagnostics and predictive analysis. In the estimation problem in this paper, heuristic methods outperform mathematical models.

LearnMet: learning domain-specific distance metrics for plots of scientific functions

Scientific experimental results are often depicted as plots of functions to aid their visual analysis and comparison. In computationally comparing these plots using techniques such as similarity search and clustering, the notion of similarity is typically distance. However, it is seldom known which distance metric(s) best preserve(s) semantics in the respective domain. It is thus desirable to learn such domain-specific distance metrics for the comparison of plots. This paper describes a technique called LearnMet proposed to learn such metrics. The input to LearnMet is a training set with actual clusters of plots. These are iteratively compared with clusters over the same plots predicted using an arbitrary but fixed clustering algorithm. Using a guessed initial metric for clustering, adjustments are made to the metric in each epoch based on the error between the predicted and actual clusters until the error is minimal or below a given threshold. The metric giving the lowest error is output as the learned metric. The proposed LearnMet technique and its enhancements are discussed in detail in this paper. The primary application of LearnMet is clustering plots in the Heat Treating domain. Hence it is rigorously evaluated using Heat Treating data. Given distinct test sets for evaluation, clusters of plots predicted using the learned metrics are compared with given actual clusters over the same plots. The extent to which the predicted and actual clusters match each other denotes the accuracy of the learned metrics.