WSEAS Transactions on Information Science and Applications archive, Jun 1, 2008
Data mining is the process of extraction of hidden predictive information from large databases an... more Data mining is the process of extraction of hidden predictive information from large databases and expressing them in a simple and meaningful manner. This paper explains the use of Fuzzy logic as a data mining process to generate decision trees from a pavement (road) database obtained from Ohio Department of Transportation containing historical pavement information from 1985 to 2006. Generally there are many attributes in the pavement database and often it is a complicated process to develop a mathematical model to classify the data. This study demonstrates the use of fuzzy logic to generate decision tree to classify the pavement data. Further, the fuzzy decision tree is then converted to fuzzy rules. These fuzzy rules will assist decision-making process for selecting a particular type of repair on a pavement based on its current condition. The fuzzy decision tree induction method used is based on minimizing the measure of classification ambiguity for different attributes. These models overcome the sharp boundary problems, providing soft controller surface and good accuracy dealing with continuous attributes and prediction problems. This method was compared with common logistic regression model for predicting the pavement treatment. The results show that the fuzzy decision method outperforms the logistic regression model by 10%. The fuzzy decision tree method generates the rules, which gives the better understanding of the relationship between the parameters and the pavement treatment prediction.
This paper aims to automatically design optimal LSTM topologies using the clonal selection algori... more This paper aims to automatically design optimal LSTM topologies using the clonal selection algorithm (CSA) to solve text classification tasks such as sentiment analysis and SMS spam classification. Designing optimal topologies involves determining the best configuration of hyperparameters that will give the best performance. The current state-of-the-art LSTM topologies are often designed using trial and error approaches which are incredibly time-consuming and require domain experts. Our proposed method, referred to as CSA-LSTM, is evaluated using the Large Movie Review Dataset (IMDB). Furthermore, to verify the robustness of the hyperparameters discovered by CSA for the IMDB dataset, we have used them for the other datasets, viz. the Twitter US Airline Sentiment and the SMS Spam Collection. Additionally, the discovered hyperparameters for the LSTM are combined with predetermined convolutional neural network (CNN) layers to achieve the same or better results to fast the training time and fewer trainable parameters. For further verification and evaluation of the generalization ability and effectiveness of the proposed approach, it is compared with four machine learning algorithms widely used for text classification tasks: (1) random forest (RF), (2) logistic regression (LR), (3) support vector machine (SVM), and (4) multinomial naive Bayes (NB). The results of our experiments show that the LSTM topologies automatically designed by our CSA method are less expensive, reusable and outperform the machine learning algorithms and other models in the literature evaluated on the same three datasets. Through our proposed method, LSTM's best topology can be self-determined without any human intervention, making CSA-based algorithms a promising approach to automatically design optimal LSTM topologies that provide the best performance for a given task.
Proceedings of WSEAS International Conference on …
Abstract: This paper presents a soft computing technique using neuro fuzzy approach to predict ... more Abstract: This paper presents a soft computing technique using neuro fuzzy approach to predict the future pavement condition based on the current pavement age and current pavement condition. The Ohio Department of Transportation (ODOT) database for the asphalt ...
2018 International Conference on Computational Science and Computational Intelligence (CSCI), 2018
In this paper, Convolution Neural Network (CNN) and a special variant of Recurrent Neural Network... more In this paper, Convolution Neural Network (CNN) and a special variant of Recurrent Neural Network (RNN) named Long Short-Term Memory Model (LSTM) with peep hole connection is developed for optical character recognition (OCR). Data-set of mathematical equations known as Image to Latex 100K is retrieved from OPEN-AI and used for testing the model. First, the mathematical equations from the images are converted to Latex texts. Then this Latex text is used to render the mathematical equations. The proposed method uses the tokenized data, which is sequentially given to the deep learning network. The sequential process helps the algorithms to keep track of the processed data and yield high accuracy. A new variant of LSTM called "LSTM with peephole connections" and Stochastic "Hard" Attention model was used. The performance of the proposed deep learning neural network is compared with INFTY (which uses no RNN) and WYGIWYS (which uses RNN). The proposed algorithm gives a better accuracy of 76% as compared to 74% achieved by WYGIWYS.-Convolutional Neural Network, Recurrent Neural Network, Long Short-Term Memory (LSTM) with peephole connections. IMAGE2LATEX 100K.
Proceedings of the 10th annual conference companion on Genetic and evolutionary computation, 2008
We introduce quotient graphs for modeling neutrality in evolutionary search. We demonstrate that ... more We introduce quotient graphs for modeling neutrality in evolutionary search. We demonstrate that for a variety of evolutionary computing problems, search can be characterized by grouping genes with similar fitness and search behavior into quotient sets. These sets can potentially reduce the degrees of freedom needed for modeling evolutionary behavior without any loss of accuracy in such models. Quotients sets, which are also shown to be Markov models, aid in understanding the nature of search. We explain how to calculate Fitness Distance Correlation (FDC) through quotient graphs, and why different problems can have the same FDC but have different dynamics. Quotient models also allow visualization of correlated evolutionary drives.
The paper proposes a new hierarchical hypernet which is an extension of hypernet suggested by Hwa... more The paper proposes a new hierarchical hypernet which is an extension of hypernet suggested by Hwang and Ghosh. The hypernet is generalised for any value ofG = 2* for 0 < x <</-2 where d is the dimension of the basic hyper cube. The mathematical formulas for the network parameters are developed as a function ofG. The hier (d,h,G) hypernet consists of(d,l,G), (d,2,G), (d,3,G), (d,h-l,G), (d,h,G) hypernets connected in top down fashion, in h physical levels. The general methodology to construct hierarchical hypernet is explained. The relationship between network parameters of hierarchical hypernets and hypernets suggested earlier 1 is developed. Some algorithms are mapped onto hierarchical hypernets to illustrate their ability to support parallel processing in a hierarchical structured or data dependent environment. The hierarchical hypernets are suitable for pipelining and hence the speed up achieved in various algorithms is illustrated using space time diagrams. The advantage with these networks is that they do not require reservation of one subnetwork in each recursive level to emulate the next higher level for pipelining.
In this paper Evolutionary Strategy for uncorrelated mutations such as self-adaptive one step and... more In this paper Evolutionary Strategy for uncorrelated mutations such as self-adaptive one step and self-adaptive k step are compared with correlated mutation. The two offspring selection techniques of direct replacement (µ,λ) and best fit (µ+λ) are used for comparison. These techniques were applied to a standard multi peak function to evaluate their performance. It was found that none of these approaches always found the global maximum. The results were very much dependent on the selection of the initial random parents. Therefore a new approach of correlated mutation using additional geometric translation has been proposed. It is illustrated that this technique was successful in finding the global maximum.
Data mining is the process of extraction of hidden predictive information from large databases an... more Data mining is the process of extraction of hidden predictive information from large databases and expressing them in a simple and meaningful manner. This paper explains the use of Fuzzy logic as a data mining process to generate decision trees from a pavement (road) ...
The Santa Fe Ant model problem has been extensively used to investigate, test and evaluate Evolut... more The Santa Fe Ant model problem has been extensively used to investigate, test and evaluate Evolutionary Computing systems and methods over the past two decades. There is however no literature on its program structures that are systematically used for fitness improvement, the geometries of those structures and their dynamics during optimization. This paper analyzes the Santa Fe Ant Problem using a new phenotypic schema and landscape analysis based on executed instruction sequences. For the first time we detail systematic structural features that give high fitness and the evolutionary dynamics of such structures. The new schema avoids variances due to introns. We develop a phenotypic variation method that tests the new understanding of the landscape. We also develop a modified function set that tests newly identified synchronization constraints. We obtain favorable computational efforts compared to those in the literature, on testing the new variation and function set on both the Sant...
In this paper, an explainable intelligence model that gives the logic behind the decisions unmann... more In this paper, an explainable intelligence model that gives the logic behind the decisions unmanned aerial vehicle (UAV) makes when it is on a predefined mission and chooses to deviate from its designated path is developed. The explainable model is on a visual platform in the format of if-then rules derived from the Sugeno-type fuzzy inference model. The model is tested using the data recorded from three different missions. In each mission, adverse weather, conditions and enemy locations are introduced at random locations along the path of the mission. There are two phases to the model development. In the first phase, the Mamdani fuzzy model is used to create rules to steer the UAV along the designated mission and the rules of engagement when it encounters weather and enemy locations along and near its chosen mission. The data are gathered as UAV traverses on each mission. In the second phase, the data gathered from these missions are used to create a reverse model using a Sugeno-type fuzzy inference system based on the subtractive clustering in the data. The model has seven inputs (time, x-coordinate, y-coordinate, heading direction, engage in attack, continue mission, and steer UAV) and two outputs (weather conditions and distance from the enemy). This model predicts the outputs regarding the weather conditions and enemy positions whenever UAV deviates from the predefined path. The model is optimized with respect to the number of rules and prediction accuracy by adjusting subtractive clustering parameters. The model is then fine-tuned with ANFIS. The final model has six rules and root mean square error value that is less than 0.05. Furthermore, to check the robustness of the model, the Gaussian random noise is added to a UAV path, and the prediction accuracy is validated. INDEX TERMS Explainable artificial intelligence (XAI), fuzzy logic, ANFIS, unmanned aerial vehicle (UAV), subtractive clustering. I. INTRODUCTION Unmanned Air Vehicles(UAVs) are used for many purposes including agriculture, industry, law enforcement, and defense. These autonomous systems have several advantages over manned aerial vehicles as not only they reduce expenses by avoiding human error, but they also save the lives of fighter jet pilots.The incoming generation of artificial intelligence(AI) systems are showing significant success through the use of various machine learning techniques. These systems offer a wide range of benefits when it comes to simplifying the lives of individuals as well as military operations. Continued advances promise to produce autonomous systems that will perceive, learn, decide, and act on their own. However, the effectiveness of today's AI systems is limited by the inability of the machine to explain its decisions and actions to human users [1]-[3]. This is where the concept of Explainable Artificial Intelligence (XAI) comes in to play. XAI aims to create a suite of machine learning techniques that will produce more explainable models while maintaining a high level of learning performance (prediction accuracy). As a rule, XAI enables human users to understand, appropriately trust, and effectively manage [1], [2] the emerging generation of artificially intelligent partners. The central problem of machine learning models is that they are regarded as black-box models. Meaning, even if we understand the underlying mathematical principles, they lack an explicit declarative knowledge representation [1], [3]. The lack of knowledge representation and explainable
International Journal of Computational Intelligence Systems, 2010
The paper demonstrates performance enhancement using selective cloning on evolutionary neural net... more The paper demonstrates performance enhancement using selective cloning on evolutionary neural network over the conventional genetic algorithm and neural back propagation algorithm for data classification. Introduction of selective cloning improves the convergence rate of the genetic algorithm without compromising on the classification errors. The selective cloning is tested on five data sets. The Iris data problem is used as a benchmark to compare the selective cloning technique with the conventional GA and the back-propagation algorithm. For comparative analysis, same neural network architecture is used for both the back propagation and the genetic algorithms. The selective cloning approach is based on the schema theorem. By using selective cloning, it has been shown that GA is 27.78% more efficient than the conventional GA and 83.33% more efficient than the back propagation approach. The results of selective cloning on other data sets are also discussed.
Journal of the Air & Waste Management Association, 2013
The present study developed a novel approach to modeling indoor air quality (IAQ) of a public tra... more The present study developed a novel approach to modeling indoor air quality (IAQ) of a public transportation bus by the development of hybrid genetic-algorithm-based neural networks (also known as evolutionary neural networks) with input variables optimized from using the regression trees, referred as the GART approach. This study validated the applicability of the GART modeling approach in solving complex nonlinear systems by accurately predicting the monitored contaminants of carbon dioxide (CO 2), carbon monoxide (CO), nitric oxide (NO), sulfur dioxide (SO 2), 0.3-0.4 µm sized particle numbers, 0.4-0.5 µm sized particle numbers, particulate matter (PM) concentrations less than 1.0 µm (PM 1.0), and PM concentrations less than 2.5 µm (PM 2.5) inside a public transportation bus operating on 20% grade biodiesel in Toledo, OH. First, the important variables affecting each monitored in-bus contaminant were determined using regression trees. Second, the analysis of variance was used as a complimentary sensitivity analysis to the regression tree results to determine a subset of statistically significant variables affecting each monitored in-bus contaminant. Finally, the identified subsets of statistically significant variables were used as inputs to develop three artificial neural network (ANN) models. The models developed were regression tree-based backpropagation network (BPN-RT), regression tree-based radial basis function network (RBFN-RT), and GART models. Performance measures were used to validate the predictive capacity of the developed IAQ models. The results from this approach were compared with the results obtained from using a theoretical approach and a generalized practicable approach to modeling IAQ that included the consideration of additional independent variables when developing the aforementioned ANN models. The hybrid GART models were able to capture majority of the variance in the monitored in-bus contaminants. The genetic-algorithm-based neural network IAQ models outperformed the traditional ANN methods of the back-propagation and the radial basis function networks. Implications: The novelty of this research is the development of a novel approach to modeling vehicular indoor air quality by integration of the advanced methods of genetic algorithms, regression trees, and the analysis of variance for the monitored in-vehicle gaseous and particulate matter contaminants, and comparing the results obtained from using the developed approach with conventional artificial intelligence techniques of back propagation networks and radial basis function networks. This study validated the newly developed approach using holdout and threefold cross-validation methods. These results are of great interest to scientists, researchers, and the public in understanding the various aspects of modeling an indoor microenvironment. This methodology can easily be extended to other fields of study also.
IEEE Transactions on Evolutionary Computation, 2009
We present a new perspective of search in Evolutionary Computing (EC) by using a novel model for ... more We present a new perspective of search in Evolutionary Computing (EC) by using a novel model for the analysis and visualization of genotype to phenotype maps. The model groups genes into quotient sets and shows their adjacencies. A unique quality of the quotient model is that it details geometric qualities of maps that are not otherwise easy to observe. The model shows how random mutations on genes make non-random phenotype preferences, based on the structure of a map. The interaction between such mutation-based preferences with fitness preferences is important for explaining population movements on neutral landscapes. We show the widespread applicability of our approach by applying it to different representations, encodings and problems including Grammatical Evolution (GE), Cartesian Genetic Programming, Parity and Majority Coding, OneMax, Needle-in-Haystack, Deceptive Trap and Hierarchical if-and-only-if. We also use the approach to address conflicting results in the neutral evolution literature and to analyze concepts relevant to neutral evolution including robustness, evolvability, tunneling and the relation between genetic form and function. We use the model to develop theoretical results on how mapping and neutral evolution affects search in GE. We study the two phases of mapping in GE; these being transcription (i.e. unique identification of genes with integers), and translation (i.e. many-to-one mapping of genotypes to phenotypes). It is shown that translation and transcription schemes belong to equivalence classes, therefore the properties we derive for specific schemes are applicable to classes of schemes. We present a new perspective on population diversity. We specify conditions under which increasing degeneracy (by increasing codon size) or rearranging the rules of a grammar do not affect performance. It is shown that there is a barrier to nontrivial neutral evolution with the use of the natural transcription with modulo translation combination; a necessary but not sufficient condition for such evolution is at least three bits should change on mutation within a single codon. This barrier can be avoided by using Gray transcription. We empirically validate some findings.
Abstract: - The use of Grammatical Evolution for automating the intrusion detection rules is inve... more Abstract: - The use of Grammatical Evolution for automating the intrusion detection rules is investigated in this paper. We apply this method to the KDD99 intrusion dataset and demonstrate its usefulness in this context. We achieve favorable results by evolving rules for classifying both normal ...
WSEAS Transactions on Information Science and Applications archive, 2008
Data mining is the process of extraction of hidden predictive information from large databases an... more Data mining is the process of extraction of hidden predictive information from large databases and expressing them in a simple and meaningful manner. This paper explains the use of Fuzzy logic as a...
This paper proposes and develops a fuzzy evolutionary system based on the Grey Wolf Optimizer (GW... more This paper proposes and develops a fuzzy evolutionary system based on the Grey Wolf Optimizer (GWO) algorithm to evolve Mamdani fuzzy rules that give a knowledge base for accurate classification of data set. GWO takes inspiration from nature and is modeled after the hunting behavior of the grey wolves as they move around within a pack taking cues from the leader alpha, beta, and delta wolves until they find the best position to encircle and attack the prey. The algorithm is mapped onto the data specific rule base structure of the fuzzy systems. A grammar template in the form of fuzzy rules is designed, and then the GWO algorithm is used to evolve the fuzzy rules which classify the datasets. The algorithm will generate meaningful rules that make sense of data in easy to comprehend fuzzy rules. The algorithm was extensively tested on 15 datasets. GWO was compared with the standard Particle Swarm Optimizer (PSO) algorithm in generating a similar type of rules and comparing the accuracy of these two sets of rules in data classification. It was noted that the GWO algorithm converges in a lesser number of iterations and in a shorter time as compared to PSO and achieves the best accuracy.
International Journal of Machine Learning and Computing
In this paper, the Clonal Selection Algorithm (CSA) is implemented to determine the optimal archi... more In this paper, the Clonal Selection Algorithm (CSA) is implemented to determine the optimal architecture of a convolutional neural network (CNN) for the classification of handwritten character digits. The efficacy of CNN in image recognition is one of the central motives why the world has woken up to the effectiveness of deep learning. During training, an optimal CNN architecture can extract complex features from the data that is being trained; however, the ideal architecture of a CNN for a specific problem cannot be determined by some standard procedure. In practice, CNN architectures are generally designed using human expertise and domain knowledge. By using CSA, optimal architecture of CNN can be determined autonomously through evolution of hyperparameters of the architecture for a given dataset. In this work, proposed methodology is tested on EMNIST dataset which is an enhanced version of MNIST dataset. The results have proven that the CSA based tuning is capable of generating optimal CNN architectures. Through this proposed technique, the best architecture of CNN for a given problem can be selfdetermined without any human intervention.
Uploads
Papers by Devinder Kaur