Search

This paper reports the development and testing of a new intervention to recruit terminal Master’s students into computer science (CS) PhD programs and to enhance diversity in academic CS. We introduced peer assistants (i.e., successful PhD students) to CSCI 549: Intelligent Systems, a popular course for Master’s students with approximately 40–50% women and 90% international students. Although there is a fair amount of diversity in this course and in the CS Master’s program generally (i.e., international female students), many of these students do not continue to earn a PhD or get involved in research during the Master’s program. Because increasing the diversity of CS professors is imperative for enhancing the diversity of CS majors and the CS workforce, it is critical to test new methods to recruit Master’s students from underrepresented groups into academic CS. To address this need, we introduced PhD student peer assistants into Intelligent Systems not only to help Master’s students with in-class research projects but to also act as role models to promote Master’s students’ interest in CS research and PhD programs. Thus, this paper suggests a new and innovative technique for an enhancing diversity in academic CS and STEM generally.

Watershed planning over a geographic area is a difficult task primarily due to the presence of large number of stakeholders and decision makers whose intrinsic conflicting and/or subjective preferences often lead to uncertainty in perceived fitness of planning decisions. Deciding which watershed strategy should be implemented at what location requires a participatory approach to design and decision making, if adoption of landscape decisions is critical to success. Analytical participatory design (APD) approaches aim to enable farmers, environmentalists, government agencies, and other stakeholders to visualize the landscape, explore and design competitive scenarios of implementing certain management practices on the landscape. Since these approaches improve decision makers’ awareness of opportunities and constraints in the co-existing physical and human systems, it is hypothesized that they can be used to generate acceptable decisions that are robust to uncertainties in stakeholder preferences.An APD method based on Interactive optimization is described in this paper and tested for design of wetlands in a study watershed site (Eagle Creek Watershed) in the state of Indiana. The method is then used to test research hypothesis by involving multiple virtual stakeholders as surrogates to diverse human users and their preferences. The results indicate that, while, as expected, the interactive optimization approach results in lower values of the financial and environmental objective criteria (which are being traded off against users’ diverse subjective personal criteria), it also results in a relatively high degree of user consensus, indicating high likelihood of adoption of the generated solutions by the stakeholders.

Interactive Genetic Algorithms (IGA) are advanced human-in-the-loop optimization methods that enable humans to give feedback, based on their subjective and unquantified preferences and knowledge, during the algorithm's search process. While these methods are gaining popularity in multiple fields, there is a critical lack of data and analyses on (a) the nature of interactions of different humans with interfaces of decision support systems (DSS) that employ IGA in water resources planning problems and on (b) the effect of human feedback on the algorithm's ability to search for design alternatives desirable to end-users. In this paper, we present results and analyses of observational experiments in which different human participants (surrogates and stakeholders) interacted with an IGA-based, watershed DSS called WRESTORE to identify plans of conservation practices in a watershed. The main goal of this paper is to evaluate how the IGA adapts its search process in the objective space to a user's feedback, and identify whether any similarities exist in the objective space of plans found by different participants. Some participants focused on the entire watershed, while others focused only on specific local subbasins. Additionally, two different hydrology models were used to identify any potential differences in interactive search outcomes that could arise from differences in the numerical values of benefits displayed to participants. Results indicate that stakeholders, in comparison to their surrogates, were more likely to use multiple features of the DSS interface to collect information before giving feedback, and dissimilarities existed among participants in the objective space of design alternatives.

Learning theory has been studied for a long time by philosophers, and in the last century by psychologists and engineers. Yet, all learning is carried out in a general deterministic or stochastic environment, mostly by one isolated learner. This paper discusses the concept of mutual learning, where two or more entities attempt to learn from each other. The question posed is: “If two or more entities are learning in the same or similar environments trying to solve the same or similar tasks, how can they share their learning to improve themselves?” The authors believe that this is a central question that will keep researchers busy for many years. The paper merely introduces this question for discussion, and suggests some preliminary answers using the well known stochastic learning automaton framework for reinforcement learning.

This paper discusses a method for dealing with limited data in deep networks based on calculating the uncertainty associated with remaining training data. The method was developed for the Watershed REstoration using Spatio-Temporal Optimization of REsources (WRESTORE) system, an interactive decision support system designed for performing multi-criteria decision analysis with a distributed system of conservation practices on the Eagle Creek Watershed in Indiana, USA. Our results show faster and more stable convergence when using an uncertainty-based incremental sampling method than when using a standard random incremental sampling method. This work describes the existing WRESTORE system, provides details about the implementation of our uncertainty-based incremental sampling method, and provides a discussion of our results and future work. The primary contribution of the paper is an uncertainty-based incremental sampling method which can be applied to limited data watershed design problems.

In this paper, we answer the question of to what extend selective decentralization could enhance the learning and control performance when the system is noisy and unknown. Compared to the previous works in selective decentralization, in this paper, we add the system noise as another complexity in the learning and control problem. Thus, we only perform analysis for some simple toy examples of noisy linear system. In linear system, the Halminton-Jaccobi-Bellman (HJB) equation becomes Riccati equation with closed-form solution. Our previous framework in learning and control unknown system is based on the following principle: approximating the system using identification in order to apply model-based solution. Therefore, this paper would explore the learning and control performance on two aspects: system identification error and system stabilization. Our results show that selective decentralization show better learning performance than the centralization when the noise level is low.

In this paper, we explore the capability of selectively decentralized Q-learning approach in learning how to optimally stabilize control systems, as compared to the centralized approach. We focus on problems in which the systems are completely unknown except the possible domain knowledge that allow us to decentralize into subsystems. In selective decentralization, we explore all of the possible communication policies among subsystems and use the cumulative gained Q-value as the metric to decide which decentralization scheme should be used for controlling. The results show that the selectively decentralized approach not only stabilizes the system faster but also shows superior converging speed on gained Q-value in different systems with different interconnection strength. In addition, the selectively decentralized converging time does not seem to grow exponentially with the system dimensionality. Practically, this fact implies that the selectively decentralized Q-learning could be used as an alternative approach in large-scale unknown control system, where in theory, the Hamilton-Jacobi-Bellman-equation approach is difficult to derive the close-form solution.

Determining the optimal design of a watershed is a highly subjective process which involves the consideration of many distinct factors by several different stakeholder groups. We describe additional functionality for our watershed planning system, called WRESTORE (Watershed REstoration Using Spatio-Temporal Optimization of REsources) (http://wrestore.iupui.edu), where stakeholders can collaboratively optimize best management practices on to the watershed. WRESTORE utilizes the USDA's public domain Soil and Water Assessment Tool hydrologic model for watershed simulations. Reinforcement learning and interactive genetic algorithms are applied for the search process. The new functionality described is a user modeling component that develops a computational model of a user's decision-making, based on real-time user-provided ratings for a subset of possible designs. The user modeling task utilizes neural network approaches, such as deep learning. We believe the originality of our approach centers on integrating user models in to the hydrological decision support process. This paper thus has three objectives: (i) outline current work in user modeling and watershed design, (ii) describe our system for interactive optimization of watershed design, and (iii) describe our work on implementing accurate and stable user predictive models to boost optimization performance.

In this paper, we explore the capability of selective decentralization in improving the control performance for unknown large-scale systems using model-based approaches. In selective decentralization, we explore all of the possible communication policies among subsystems and show that with the appropriate switching among the resulting multiple identification models (with corresponding communication policies), such selective decentralization significantly outperforms a centralized identification model when the system is weakly interconnected, and performs at least equivalent to the centralized model when the system is strongly interconnected. To derive the sub-optimal control, our control design include two phases. First, we apply system identification to train the approximation model for the unknown system. Second, we find the suboptimal solution of the Halminton-Jacobi-Bellman (HJB) equation to derive the suboptimal control. In linear systems, the HJB equation transforms to the well-solved Riccati equation with closed-form solution. In nonlinear systems, we discretize the approximation model in order to acquire the control unit by using dynamic programming methods for the resulting Markov Decision Process (MDP). We compare the performance among the selective decentralization, the complete decentralization and the centralization in our two-phase control design. Our results show that selective decentralization outperforms the complete decentralization and the centralization approaches when the systems are completely decoupled or strongly interconnected.

Associations among biological objects such as genes, proteins, and drugs can be discovered automatically from the scientific literature. TransMiner is a system for finding associations among objects by mining the Medline database of the scientific literature. The direct associations among the objects are discovered based on the principle of co-occurrence in the form of an association graph. The principle of transitive closure is applied to the association graph to find potential transitive associations. The potential transitive associations that are indeed direct are discovered by iterative retrieval and mining of the Medline documents. Those associations that are not found explicitly in the entire Medline database are transitive associations and are the candidates for hypothesis generation. The transitive associations were ranked based on the sum of weight of terms that co-occur with both the objects. The direct and transitive associations are visualized using a graph visualization applet. TransMiner was tested by finding associations among 56 breast cancer genes and among 24 objects in the calpain signal transduction pathway. TransMiner was also used to rediscover associations between magnesium and migraine.

In this paper, we describe the COBioSIFTER project that aims at constructing a framework for managing information from the biological domain. It uses a distributed multi-agent approach based on the CORBA middleware and contains an architecture made up of functionally heterogeneous agents. A series of experiments has been carried out with a prototype of COBioSIFTER to illustrate its advantages such as high performance, scalability, heterogeneity, and fault tolerance. The results of these experiments demonstrate the effectiveness of such an agent-based biological information framework.

Although large-scale classification studies of genetic sequence data are in progress around the world, very few studies compare different classification approaches, e.g. unsupervised and supervised, in terms of objective criteria such as classification accuracy and computational complexity. In this paper, we study such criteria for both unsupervised and supervised classification of a relatively large sequence data set. The unsupervised approach involves use of different sequence alignment algorithms (e.g., Smith-Waterman, FASTA and BLAST) followed by clustering using the Maximin algorithm. The supervised approach uses a suitable numeric encoding (relative frequencies of tuples of nucleotides followed by principal component analysis) which is fed to a Multi-layer Backpropagation Neural Network. Classification experiments conducted on IBM-SP parallel computers show that FASTA with unsupervised Maximin leads to best trade-off between accuracy and speed among all methods, followed by supervised neural networks as the second best approach. Finally, the different classifiers are applied to the problem of cross-species homology detection.

Today a massive amount of information available on the WWW often makes searching for information of interest a long and tedious task. Chasing hyperlinks to find relevant information may be daunting. To overcome such a problem, a learning system, cognizant of a user's interests, can be employed to automatically search for and retrieve relevant information by following appropriate hyperlinks. In this paper, we describe the design of such a learning system for automated Web navigation using adaptive dynamic programming methods. To improve the performance of the learning system, we introduce the notion of multiple model-based learning agents operating in parallel, and describe methods for combining their models. Experimental results on the WWW navigation problem are presented to indicate that combining multiple learning agents, relying on user feedback, is a promising direction to improve learning speed in automated WWW navigation.

A neural-network inverse mapping approach was used in a structural integrity redesign problem to achieve the desired strength, corrosion, and fatigue properties. Also, through the inverse-mapping procedure the relative importance of damage parameters was obtained for two trained neural-network models. The damage parameters corresponding to a given strength and corrosion rate are predicted in one network, whereas the damage parameters corresponding to a given corrosion fatigue life are predicted in the other network. The results obtained from the inverse mapping procedure are compared with actual panel configurations and environments and experimental data. The results obtained through the inverse-mapping procedure are found to agree well with the experimental data, thus demonstrating the feasibility of the approach for the structural integrity redesign of aging aircraft structures.

Modeling users in information filtering systems is a difficult challenge due to dimensions such as nature, scope, and variability of interests. Numerous machine-learning approaches have been proposed for user modeling in filtering systems. The focus has been primarily on techniques for user model capture and representation, with relatively simple assumptions made about the type of users' interests. Although many studies claim to deal with “adaptive” techniques and thus they pay heed to the fact that different types of interests must be modeled or even changes in interests have to be captured, few studies have actually focused on the dynamic nature and the variability of user-interests and their impact on the modeling process. A simulation based information filtering environment called SIMSFITER was developed to overcome some of the barriers associated with conducting studies on user-oriented factors that can impact interests. SIMSIFTER implemented a user modeling approach known as reinforcement learning that has proven to be effective in previous filtering studies involving humans. This paper reports on several studies conducted using SIMSIFTER that examined the impact of key dimensions such as type of interests, rate of change of interests and level of user-involvement on modeling accuracy and ultimately on filtering effectiveness.

In this paper, two supervised neural networks are used to estimate the forces developed during milling. These two Artificial Neural Networks (ANNs) are compared based on a cost function that relates the size of the training data to the accuracy of the model. Training experiments are screened based on design of experiments. Verification experiments are conducted to evaluate these two models. It is shown that the Radial Basis Network model is superior in this particular case. Orthogonal design and specifically equally spaced dimensioning showed to be a good way to select the training experiments.

Recent advances in processor, networking and software technologies have made distributed computing a reality in today's world. Distributed systems offer many advantages, ranging from a higher performance to the effective utilization of physically dispersed resources. Many diverse application domains can benefit by exploiting principles of distributed computing. Information filtering is one such application domain. In this article, we present a design of a homogeneous distributed multi-agent information filtering system, called D-SIFTER. D-SIFTER is based on the language-dependent model of Java RMI. The detailed design process and various experiments carried out using D-SIFTER are also described. The results indicate that the distributed inter-agent collaboration improves the overall filtering performance.