Hamid Alinejad-Rokny

Proposing a classifier ensemble framework based on classifier selection and decision tree

Abstract One of the most important tasks in pattern, machine learning, and data mining is classification problem. Introducing a general classifier is a challenge for pattern recognition communities, which enables one to learn each problem׳s dataset. Many classifiers have been proposed to learn any problem thus far. However, many of them have their own positive and negative aspects. So they are good only for specific problems. But there is no strong solution to recognize which classifier is better or good for a specific problem. Fortunately, ensemble learning provides a good way to have a near-optimal classifying system for any problem. One of the most challenging problems in classifier ensemble is introducing a suitable ensemble of base classifiers. Every ensemble needs diversity. It means that if a group of classifiers is to be a successful ensemble, they must be diverse enough to cover their errors. Therefore, during ensemble creation, a mechanism is needed to ensure that the ensemble classifiers are diverse. Sometimes this mechanism can select/remove a subset of base classifiers with respect to maintaining the diversity of the ensemble. This paper proposes a novel method, named the Classifier Selection Based on Clustering (CSBS), for ensemble creation. To insure diversity in ensemble classifiers, this method uses the clustering of classifiers technique. Bagging is used to produce base classifiers. During ensemble creation, every type of base classifier is the same as a decision tree classier or a multilayer perceptron classifier. After producing a number of base classifiers, CSBC partitions them by using a clustering algorithm. Then CSBC produces a final ensemble by selecting one classifier from each cluster. Weighted majority vote method is used as an aggregator function. In this paper we investigate the influence of cluster number on the performance of the CSBC method; we also probe how we can select a good approximate value for cluster number in any dataset. We base our study on a large number of real datasets of UCI repository to reach a definite result.

Effects of resampling method and adaptation on clustering ensemble efficacy

Clustering ensembles combine multiple partitions of data into a single clustering solution of better quality. Inspired by the success of supervised bagging and boosting algorithms, we propose non-adaptive and adaptive resampling schemes for the integration of multiple independent and dependent clusterings. We investigate the effectiveness of bagging techniques, comparing the efficacy of sampling with and without replacement, in conjunction with several consensus algorithms. In our adaptive approach, individual partitions in the ensemble are sequentially generated by clustering specially selected subsamples of the given dataset. The sampling probability for each data point dynamically depends on the consistency of its previous assignments in the ensemble. New subsamples are then drawn to increasingly focus on the problematic regions of the input feature space. A measure of data point clustering consistency is therefore defined to guide this adaptation. Experimental results show improved stability and accuracy for clustering structures obtained via bootstrapping, subsampling, and adaptive techniques. A meaningful consensus partition for an entire set of data points emerges from multiple clusterings of bootstraps and subsamples. Subsamples of small size can reduce computational cost and measurement complexity for many unsupervised data mining tasks with distributed sources of data. This empirical study also compares the performance of adaptive and non-adaptive clustering ensembles using different consensus functions on a number of datasets. By focusing attention on the data points with the least consistent clustering assignments, whether one can better approximate the inter-cluster boundaries or can at least create diversity in boundaries and this results in improving clustering accuracy and convergence speed as a function of the number of partitions in the ensemble. The comparison of adaptive and non-adaptive approaches is a new avenue for research, and this study helps to pave the way for the useful application of distributed data mining methods.

Data weighing mechanisms for clustering ensembles

Inspired by bagging and boosting algorithms in classification, the non-weighing and weighing-based sampling approaches for clustering are proposed and studied in the paper. The effectiveness of non-weighing-based sampling technique, comparing the efficacy of sampling with and without replacement, in conjunction with several consensus algorithms have been invested in this paper. Experimental results have shown improved stability and accuracy for clustering structures obtained via bootstrapping, subsampling, and boosting techniques. Subsamples of small size can reduce the computational cost and measurement complexity for many unsupervised data mining tasks with distributed sources of data. This empirical research study also compares the performance of boosting and bagging clustering ensembles using different consensus functions on a number of datasets.

A new classifier ensemble methodology based on subspace learning

Different classifiers with different characteristics and methodologies can complement each other and cover their internal weaknesses; so classifier ensemble is an important approach to handle the weakness of single classifier based systems. In this article we explore an automatic and fast function to approximate the accuracy of a given classifier on a typical dataset. Then employing the function, we can convert the ensemble learning to an optimisation problem. So, in this article, the target is to achieve a model to approximate the performance of a predetermined classifier over each arbitrary dataset. According to this model, an optimisation problem is designed and a genetic algorithm is employed as an optimiser to explore the best classifier set in each subspace. The proposed ensemble methodology is called classifier ensemble based on subspace learning (CEBSL). CEBSL is examined on some datasets and it shows considerable improvements.

Linking Pig-Tailed Macaque Major Histocompatibility Complex Class I Haplotypes and Cytotoxic T Lymphocyte Escape Mutations in Simian Immunodeficiency Virus Infection

ABSTRACT The influence of major histocompatibility complex class I (MHC-I) alleles on human immunodeficiency virus (HIV) diversity in humans has been well characterized at the population level. MHC-I alleles likely affect viral diversity in the simian immunodeficiency virus (SIV)-infected pig-tailed macaque (Macaca nemestrina) model, but this is poorly characterized. We studied the evolution of SIV in pig-tailed macaques with a range of MHC-I haplotypes. SIVmac251 genomes were amplified from the plasma of 44 pig-tailed macaques infected with SIVmac251 at 4 to 10 months after infection and characterized by Illumina deep sequencing. MHC-I typing was performed on cellular RNA using Roche/454 pyrosequencing. MHC-I haplotypes and viral sequence polymorphisms at both individual mutations and groups of mutations spanning 10-amino-acid segments were linked using in-house bioinformatics pipelines, since cytotoxic T lymphocyte (CTL) escape can occur at different amino acids within the same epitope in different animals. The approach successfully identified 6 known CTL escape mutations within 3 Mane-A1*084-restricted epitopes. The approach also identified over 70 new SIV polymorphisms linked to a variety of MHC-I haplotypes. Using functional CD8 T cell assays, we confirmed that one of these associations, a Mane-B028 haplotype-linked mutation in Nef, corresponded to a CTL epitope. We also identified mutations associated with the Mane-B017 haplotype that were previously described to be CTL epitopes restricted by Mamu-B*017:01 in rhesus macaques. This detailed study of pig-tailed macaque MHC-I genetics and SIV polymorphisms will enable a refined level of analysis for future vaccine design and strategies for treatment of HIV infection. IMPORTANCE Cytotoxic T lymphocytes select for virus escape mutants of HIV and SIV, and this limits the effectiveness of vaccines and immunotherapies against these viruses. Patterns of immune escape variants are similar in HIV type 1-infected human subjects that share the same MHC-I genes, but this has not been studied for SIV infection of macaques. By studying SIV sequence diversity in 44 MHC-typed SIV-infected pigtail macaques, we defined over 70 sites within SIV where mutations were common in macaques sharing particular MHC-I genes. Further, pigtail macaques sharing nearly identical MHC-I genes with rhesus macaques responded to the same CTL epitope and forced immune escape. This allows many reagents developed to study rhesus macaques to also be used to study pigtail macaques. Overall, our study defines sites of immune escape in SIV in pigtailed macaques, and this enables a more refined level of analysis of future vaccine design and strategies for treatment of HIV infection.

Computational intelligence approaches for classification of medical data: State-of-the-art, future challenges and research directions

Abstract The explosive growth of data in volume, velocity and diversity that are produced by medical applications has contributed to abundance of big data. Current solutions for efficient data storage and management cannot fulfill the needs of heterogeneous data. Therefore, by applying computational intelligence (CI) approaches in medical data helps get better management, faster performance and higher level of accuracy in detection. This paper aims to investigate the state-of-the-art of computational intelligence approaches in medical data and to categorize the existing CI techniques, used in medical fields, as single and hybrid. In addition, the techniques and methodologies, their limitations and performances are presented in this study. The limitations are addressed as challenges to obtain a set of requirements for Computational Intelligence Medical Data (CIMD) in establishing an efficient CIMD architectural design. The results show that on the one hand Support Vector Machine (SVM) and Artificial Immune Recognition System (AIRS) as a single based computational intelligence approach were the best methods in medical applications. On the other hand, the hybridization of SVM with other methods such as SVM-Genetic Algorithm (SVM-GA), SVM-Artificial Immune System (SVM-AIS), SVM-AIRS and fuzzy support vector machine (FSVM) had great performances achieving better results in terms of accuracy, sensitivity and specificity.

Insights into the motif preference of APOBEC3 enzymes.

We used a multivariate data analysis approach to identify motifs associated with HIV hypermutation by different APOBEC3 enzymes. The analysis showed that APOBEC3G targets G mainly within GG, TG, TGG, GGG, TGGG and also GGGT. The G nucleotides flanked by a C at the 3′ end (in +1 and +2 positions) were indicated as disfavoured targets by APOBEC3G. The G nucleotides within GGGG were found to be targeted at a frequency much less than what is expected. We found that the infrequent G-to-A mutation within GGGG is not limited to the inaccessibility, to APOBEC3, of poly Gs in the central and 3′polypurine tracts (PPTs) which remain double stranded during the HIV reverse transcription. GGGG motifs outside the PPTs were also disfavoured. The motifs GGAG and GAGG were also found to be disfavoured targets for APOBEC3. The motif-dependent mutation of G within the HIV genome by members of the APOBEC3 family other than APOBEC3G was limited to GA→AA changes. The results did not show evidence of other types of context dependent G-to-A changes in the HIV genome.

Selection of the best well control system by using fuzzy multiple-attribute decision-making methods

There are numerous difficulties involved in drilling operations of an oil well, one of the most important of them being well control. Well control systems are applied when we have irruption of liquids or unwanted intrusion of the reservoirs liquid (oil, gas or brine) into the well, during drilling when the pressure of well fluid column is less than formation pressure, and the permeability of the reservoir has a value that is able to pass the liquid through. For this purpose, a variety of methods including Driller, wait and weight, and the concurrent methods were used to control the well at different drilling sites. In this study, we investigate the optimum method for well control using a fussy method based on many parameters, including technical factors (mud weight, drilling rate, blockage of pipes, sensitivity to drilling network changes, etc.) and security factors (existence of effervescent mud, drilling circuit control, etc.), and cost of selection, which is one of the most important decisions that are made under critical conditions such as irruption. Till now, these methods were selected based on the experience of field personnel in drilling sites. The technical criteria and standards were influenced by experience, so the soft computerizing system (fuzzy method) was used. Thus, both these criteria and standards would be of greater importance and indicate whether the optimum numerical method is the same one that is expressed by human experience. The concurrent method was selected as the best for well control, using the fuzzy method at the end of the evaluation, while field personnel experience suggests the Driller method.

High fidelity simian immunodeficiency virus reverse transcriptase mutants have impaired replication in vitro and in vivo.

The low fidelity of HIV replication facilitates immune and drug escape. Some reverse transcriptase (RT) inhibitor drug-resistance mutations increase RT fidelity in biochemical assays but their effect during viral replication is unclear. We investigated the effect of RT mutations K65R, Q151N and V148I on SIV replication and fidelity in vitro, along with SIV replication in pigtailed macaques. SIVmac239-K65R and SIVmac239-V148I viruses had reduced replication capacity compared to wild-type SIVmac239. Direct virus competition assays demonstrated a rank order of wild-type>K65R>V148I mutants in terms of viral fitness. In single round in vitro-replication assays, SIVmac239-K65R demonstrated significantly higher fidelity than wild-type, and rapidly reverted to wild-type following infection of macaques. In contrast, SIVmac239-Q151N was replication incompetent in vitro and in pigtailed macaques. Thus, we showed that RT mutants, and specifically the common K65R drug-resistance mutation, had impaired replication capacity and higher fidelity. These results have implications for the pathogenesis of drug-resistant HIV.

Epitope-Specific CD8+ T Cell Kinetics Rather than Viral Variability Determine the Timing of Immune Escape in Simian Immunodeficiency Virus Infection

CD8+ T cells are important for the control of chronic HIV infection. However, the virus rapidly acquires “escape mutations” that reduce CD8+ T cell recognition and viral control. The timing of when immune escape occurs at a given epitope varies widely among patients and also among different epitopes within a patient. The strength of the CD8+ T cell response, as well as mutation rates, patterns of particular amino acids undergoing escape, and growth rates of escape mutants, may affect when escape occurs. In this study, we analyze the epitope-specific CD8+ T cells in 25 SIV-infected pigtail macaques responding to three SIV epitopes. Two epitopes showed a variable escape pattern and one had a highly monomorphic escape pattern. Despite very different patterns, immune escape occurs with a similar delay of on average 18 d after the epitope-specific CD8+ T cells reach 0.5% of total CD8+ T cells. We find that the most delayed escape occurs in one of the highly variable epitopes, and that this is associated with a delay in the epitope-specific CD8+ T cells responding to this epitope. When we analyzed the kinetics of immune escape, we found that multiple escape mutants emerge simultaneously during the escape, implying that a diverse population of potential escape mutants is present during immune selection. Our results suggest that the conservation or variability of an epitope does not appear to affect the timing of immune escape in SIV. Instead, timing of escape is largely determined by the kinetics of epitope-specific CD8+ T cells.