Search

Identifying protein-protein interactions (PPIs) represents a challenging research problem in computational biology. Even though machine learning methods have significantly advanced the research in this field, current approaches struggle to accurately predict interactions for previously unseen proteins. Supervised autoencoders, which are neural networks trained to simultaneously predict labels and reconstruct their inputs, have been proved to offer generalization guarantees in the linear case. Moreover, the addition of a reconstruction branch was empirically shown to improve the performance of standard neural networks classifiers on several tasks. In this paper, we introduce a two-stage sequence-based PPI prediction method based on supervised autoencoders. The proposed approach consists of initially training a denoising autoencoder on protein sequences, followed by a supervised training stage in which the model learns to both predict whether two proteins interact and to reconstruct the two proteins in the pair. An experimental analysis was performed on two public PPI data sets containing testing pairs formed using both seen and unseen protein sequences. The results show that our approach surpasses, on the two data sets, multiple machine learning classifiers proposed in the literature. [ABSTRACT FROM AUTHOR]

Biological, social and economical systems expose enormous levels of complexity, and studying situations of cooperation of conflict encompassing such systems is of particular interest. The N-Player Iterated Prisoner's Dilemma (NIPD) is a general game-theoretic model that captures realistic scenarios of cooperation and conflict. This paper studies the emergence of cooperation in the NIPD, using methods borrowed from the field of reinforcement learning. Such methods attracted immense amounts of attention over the last decade, and have shown promising results in robotics, game-playing, multi-agent systems, etc. However, it is well-known that plain reinforcement learning applied in the NIPD will converge to the Nash equilibrium of the game, which is defection. Therefore, additional mechanisms are needed to foster the development and upkeep of cooperation. This work considers different interconnection topologies for players of the NIPD, and uses dynamic rewiring (players can severe some connections and form others) as a mechanism to encourage cooperation. In such scenarios, the initial interconnection topology is also a crucial aspect. Thus, conducted experiments are concerned with the NIPD over dynamic complex networks, where players can rewire edges in order to position themselves in more cooperative settings. Our results show that cooperation can emerge in such settings. [ABSTRACT FROM AUTHOR]

The paper describes a solution to utterance style transfer within the speaker's identity and emotional tone exchange while maintaining the utterance's content. Using deep generative neural networks, we developed two models that differ in introducing information related to the speaker's identity and an additional variable representing the expected emotional category. The embedding of emotions is taken from the convolutional network that was trained to classify emotional categories. Siamese network is responsible for learning the content embeddings. The models can perform the style transfer between any two speakers with a satisfactory result. This assessment is based on a survey that considers the degree of content retention, the quality of transferring voice features related to identity, and the degree of converting emotional features into a desirable category. [ABSTRACT FROM AUTHOR]

The paper focuses on simulating an artificial life in which neural networks (recurrent (RNN) and Long Short Term Memory (LSTM) networks) control prey and predator agents. The research goal was to check whether a simple genetic algorithm evolves the LSTM based controller that competes with the classic RNN controller in the real world. We also examined the impact of audio communication within a given species on the survival of agents. Our experiments evidenced the LSTM network results were slightly worse than the RNN controller. We also showed that prey agents developed herd behavior in response to predator pressure. They learned to form herds that allowed them to resist predator attacks. It was also possible to observe the prey agent's cooperation in searching for food when the plants formed clusters. [ABSTRACT FROM AUTHOR]

The analysis of the transitions that a protein conformation can undergo is an important step towards achieving a thorough understanding of protein functions. In this context, learning informative and transferable low-dimensional representations of proteins conformations is of great importance for developing robust computational approaches that can model proteins dynamical systems. In this paper we perform an experimental evaluation of the ability of the latent space representations learned by a Variational Autoencoder (VAE) trained on trajectories of proteins belonging to a superfamily to characterize an unseen protein from the same superfamily. In our study, we compare the latent space encodings obtained using two representations for the protein conformations and evaluate the impact of the latent space dimensionality on the embeddings' quality. The obtained results outline that the VAE's latent space can capture structural information characterizing a certain protein superfamily and, consequently, generalize to a different testing protein from that superfamily. [ABSTRACT FROM AUTHOR]

• Based on the relationship between different dimensions of semantic subspace, a concise learning model is constructed. • A simple yet effective approach is proposed to simplify the complex and intractable semantic features. • Two new regular terms are introduced into the loss function to ensure the direct sum decomposition. The core problem of zero-shot learning (ZSL) is extracting concise feature from cumbersome and intricate semantic space. While semantic vectors were mainly used to constrain the distance between different classes in embedding space in existing ZSL methods, we first emphasize the importance of the relationship between different dimensions in semantic space. In this paper, we propose a novel inductive approach, Multiple Semantic Subspaces Network (MSSN), to simplify the complex and intractable semantic features. Our method generates multiple disentangled subfeatures via direct sum decomposition, which not only retain completely semantic information, but also obtain simple independent hierarchical features. Specially, instead of original space, using projection subspace to map embedding space can reduce the difficulty of model optimization and enhance the generalization ability of the model. Full experiments are achieved on almost all datasets and comparision with many algorithms which exist in the present latest literature. Compared with nineteen competitors, the results show that our model outperforms the state-of-the-art on conventional ZSL setting and has a very competitive performance on generalized ZSL setting. [ABSTRACT FROM AUTHOR]

Predicting weather, and particularly severe weather, is an important challenge both for meteorological and machine learning researchers. The complexity and difficulty of the problem is mainly due to the chaotic character of the atmosphere and the implicit large set of meteorological information (radar, satellite or ground meteorological observations) which have to be analyzed by meteorologists. Thus, understanding the relationships between various meteorological parameters extracted from radar observations may be useful for providing additional comprehension about severe weather development and would help to identify situations when severe weather can occur. Self-organizing maps are being explored as an unsupervised classification model for detecting patterns in radar data which are relevant in predicting short-term weather changes. Experiments are performed on real radar data provided by the Romanian National Meteorological Administration. With the main goal of analyzing how the values for the weather radar products are evolving between consecutive radar scans, we empirically show that in general there is a slow change in the values over time, except for the situations when certain severe phenomena occur. The study conducted in this paper is aimed to provide a better insight regarding how the values of weather radar products are evolving in time both in calm and severe weather conditions, with the broader goal of using these findings for weather nowcasting. [ABSTRACT FROM AUTHOR]

• A novel cloud-based decision support system AgroDSS was developed. • Users can upload own data and process it using the online data mining tools. • AgroDSS provides predictive modeling, explanation and time series analysis. • Case study of pest population dynamics illustrates the potential for the use. Decision support systems, data analysis and data mining have become significant tools for improving business in professional world. The emerging technologies are making the precision agriculture omnipresent and allow potential for enriching it with computer-assisted decision support systems for farm management. In this paper we describe a novel system AgroDSS that bridges the gap between agricultural systems and state-of-the-art decision support methodology. The described system is intended for integration into the existing farm management information systems and provides a cloud-based decision support toolbox, allowing farmers to upload their own data, utilize several data analysis methods and retrieve their outputs. The implemented tools include predictive modeling with explanation, accuracy evaluation, time series clustering and decomposition, and structural change detection. They can help users make predictions for simulated scenarios and better understand the dependencies (interactions) within their domain. We apply the AgroDSS system on a case study of pest population dynamics, illustrating the potential for its use. [ABSTRACT FROM AUTHOR]

We are concerned with the solution of the matrix equation A X B = D in real time by means of the gradient based neural network (GNN) model, called GNN ( A, B, D ). The convergence analysis shows that the result of global asymptotic convergence is determined by the choice of the initial state and coincides with the general solution of the matrix equation A X B = D . Several applications of the GNN ( A, B, D ) model in online approximation of various inner and outer inverses with prescribed range and/or null space are considered. An appropriate adaptation of proposed models for finding an online solution of a set of linear equations A x = b is defined and investigated. The influence of various nonlinear activation functions on the convergence of GNN ( A, B, D ) is investigated both theoretically as well as using computer-simulation results. [ABSTRACT FROM AUTHOR]

Regularized empirical risk minimization using kernels and their corresponding reproducing kernel Hilbert spaces (RKHSs) plays an important role in machine learning. However, the actually used kernel often depends on one or on a few hyperparameters or the kernel is even data dependent in a much more complicated manner. Examples are Gaussian RBF kernels, kernel learning, and hierarchical Gaussian kernels which were recently proposed for deep learning. Therefore, the actually used kernel is often computed by a grid search or in an iterative manner and can often only be considered as an approximation to the “ideal” or “optimal” kernel. The paper gives conditions under which classical kernel based methods based on a convex Lipschitz loss function and on a bounded and smooth kernel are stable, if the probability measure P, the regularization parameter λ , and the kernel K may slightly change in a simultaneous manner. Similar results are also given for pairwise learning. Therefore, the topic of this paper is somewhat more general than in classical robust statistics, where usually only the influence of small perturbations of the probability measure P on the estimated function is considered. [ABSTRACT FROM AUTHOR]

Assistant robots are designed to perform specific tasks for the user, but their performance is rarely optimal, hence they are required to adapt to user preferences or new task requirements. In the previous work, the potential of an interactive learning framework based on user intervention and reinforcement learning (RL) was assessed. The framework allowed the user to correct an unfitted segment of the robot trajectory by using hand movements to guide the robot along a corrective path. So far, only the usability of the framework was evaluated through experiments with users. In the current work, the framework is described in detail and its ability to learn from a set of sample trajectories using an RL algorithm is analyzed. To evaluate the learning performance, three versions of the framework are proposed that differ in the method used to obtain the sample trajectories, which are: human-guided learning, autonomous learning, and combined human-guided with autonomous learning. The results show that the combination of the human-guided and autonomous learning achieved the best performance, and although it needed a higher number of sample trajectories than the human-guided learning, it required less user involvement. Autonomous learning alone obtained the lowest reward value and needed the highest number of sample trajectories. [ABSTRACT FROM AUTHOR]

The time-varying nature of consumption patterns is critical in the development of reliable electric vehicles and real-time schemes for assessing energy autonomy. Most of these schemes use battery voltage observations as a primary source of information and neglect variables external to the vehicle that affect its autonomy and help to characterise the behaviour of the battery as main energy storage device. Using an electric bicycle as case study, we show that the incorporation of external variables (e.g., altitude measurements) improves predictions associated with evolution of the battery voltage in time. We achieve this by proposing a novel kernel adaptive filter for multiple inputs and with a data-dependent dictionary construction. This allows us to model the dependency between battery voltage and altitude variations in a sequential manner. The proposed methodology combines automatic discovery of the relationship between voltage and altitude from data, and a kernel-based voltage predictor to address an important issue in reliability of electric vehicles. The proposed method is validated against a standard kernel adaptive filter, fixed linear filters and adaptive linear filters as baselines on the short- and long-term prediction of real-world battery voltage data. [ABSTRACT FROM AUTHOR]

Within the software testing domain, determining the order in which classes are tested in integration testing is an important problem called class integration test order identification (CITO). This problem is useful in integration testing , as it contributes to reducing the time needed for testing a software system and refers to the process of identifying an optimal order in which the application classes should be combined and tested as a group. This paper introduces a novel approach based on reinforcement learning for class integration test order optimization in the context of integration testing . The experimental evaluation is performed on four synthetic examples and on six existing software systems often used in the literature for this problem. The results obtained are analyzed and compared to similar related work from the literature, highlighting the potential of the current proposal. The proposed reinforcement learning based approach outperforms most methods existing in the software engineering literature for optimizing the test order for class-based integration. Moreover, it is general, and can be easily extended for optimizing the order in which software components should be tested during component integration testing of a software system. [ABSTRACT FROM AUTHOR]

Learning from data streams (incremental learning) is increasingly attracting research focus due to many real-world streaming problems and due to many open challenges, among which is the detection of concept drift – a phenomenon when the data distribution changes and makes the current prediction model inaccurate or obsolete. Current state-of-the art detection methods can be roughly split into performance monitoring algorithms and distribution comparing algorithms. In this work we propose a novel concept drift detector that can be combined with an arbitrary classification algorithm. The proposed concept drift detector is based on computing multiple model explanations over time and observing the magnitudes of their changes. The model explanation is computed using a methodology that yields attribute-value contributions for prediction outcomes and thus provides insight into the model’s decision-making process and enables its transparency. The evaluation has revealed that the methods surpass the baseline methods in terms of concept drift detection, accuracy, robustness and sensitivity. To even further augment interpretability, we visualized the detection of concept drift, enabling macro and micro views of the data. [ABSTRACT FROM AUTHOR]

Online data mining techniques are used to uncover relevant patterns in complex data which are dynamic by nature and thus continuously extended with real-time arriving data streams. Relational association rules (RARs), a data analysis and mining concept, extend the classical association rules so as to capture different relations between the attributes characterizing the data. This paper introduces a new Incremental Relational Association Rule Mining (IRARM) approach with the aim of progressively adapting the interesting relational association rules identified in a data set, when it is enlarged with new instances. We have experimentally evaluated IRARM on publicly available data sets. The reduction in mining time when using IRARM against mining from scratch emphasizes its efficiency in adapting the rules to real-time data extension. [ABSTRACT FROM AUTHOR]

Proteins are complex macromolecules which contribute to maintaining cellular environments and thus have fundamental roles in biological processes of living organisms. Understanding the conformational transitions of proteins represents an important stage towards comprehending protein function and would help to identify situations when mutations can occur. In this paper we use clustering as an unsupervised classification method in order to study the relevance of the residues’ relative solvent accessibility (RSA) values to analyze protein internal transitions. With the main goal of studying the evolution of RSA values between conformational transitions, we experimentally show that RSA values are slowly modifying as the protein undergoes conformational changes. The study conducted in this paper is aimed to provide a better apprehension of how proteins’ conformational transitions are evolving in time, with the broader goal of better understanding protein internal dynamics. [ABSTRACT FROM AUTHOR]

Zeroing neural network (ZNN, or termed Zhang neural network after its inventor), being a special type of neurodynamic methodology, has shown powerful abilities to solve a great variety of time-varying problems with monotonically increasing odd activation functions. However, the existing results on ZNN cannot handle the inequality constraint in the optimization problem and nonconvex function cannot applied to accelerating the convergence speed of ZNN. This work breaks these limitations by proposing ZNN models, allowing nonconvex sets for projection operations in activation functions and incorporating new techniques for handing inequality constraint arising in optimizations. Theoretical analyses reveal that the proposed ZNN models are of global stability with timely convergence. Finally, illustrative simulation examples are provided and analyzed to substantiate the efficacy and superiority of the proposed ZNN models for real-time dynamic quadratic programming subject to equality and inequality constraints. [ABSTRACT FROM AUTHOR]

In this paper the concept of sparsity for complex-valued variables is introduced in the following three types: directly in complex domain and for two real-valued pairs phase/amplitude and real/imaginary parts of complex variables. The nonlocal block-matching technique is used for sparsity implementation and filter design for each type of sparsity. These filters are complex domain generalizations of the Block Matching 3D collaborative (BM3D) filter based on the high-order singular value decomposition (HOSVD) in order to generate group-wise adaptive analysis/synthesis transforms. Complex domain denoising is developed and studied as a test-problem for comparison of the designed filters as well as the different types of sparsity modeling. [ABSTRACT FROM AUTHOR]

Moving least square regression is an important local learning algorithm. In this paper, we consider a regularized moving least square regression algorithm in reproducing kernel Hilbert space. The localized representer theorem is different from the classical representer theorems for regularized kernel machines. It shows that, regularization not only ensures the computational stability, it is also necessary for the algorithm to preserve localization property. We also studied the learning performance of the regularized moving least square algorithm and conducted a rigorous error analysis. Compared with the unregularized method, convergence analysis of regularized moving least square regression requires more natural and much simpler conditions and achieves fast rates. [ABSTRACT FROM AUTHOR]

Nonnegative matrix factorization (NMF) has been attracting intensive attention due to its wide applications. However due to the non-convexity of the NMF models, most of the existing methods are easily stuck into a bad local minima, especially in presence of noise or outliers. To alleviate this deficiency, in this paper we propose a novel NMF method by incorporating self-paced learning (SPL) methodology with traditional NMF model, to sequentially include matrix elements into NMF training from easy to complex, which draws the merits of SPL that have been demonstrated to be beneficial in avoiding bad local minima. To make the SPL methodology play a more stable and efficient role in NMF, we suggest to base on multicriteria to select training elements. The effectiveness of the proposed multicriteria self-paced NMF (MSPNMF) method is demonstrated by a series of numerical experiments on synthetic and real face image data. We also discuss the effects of different initializations on MSPNMF. Experimental results show that MSPNMF is sensitive to the starting values and different initializations should be adopted for MSPNMF based on different situations. [ABSTRACT FROM AUTHOR]

In this study, we present Swift Linked Data Miner, an interruptible algorithm that can directly mine an online Linked Data source (e.g., a SPARQL endpoint) for OWL 2 EL class expressions to extend an ontology with new SubClassOf: axioms. The algorithm works by downloading only a small part of the Linked Data source at a time, building a smart index in the memory and swiftly iterating over the index to mine axioms. We propose a transformation function from mined axioms to RDF Data Shapes. We show, by means of a crowdsourcing experiment, that most of the axioms mined by Swift Linked Data Miner are correct and can be added to an ontology. We provide a ready to use Protégé plugin implementing the algorithm, to support ontology engineers in their daily modeling work. [ABSTRACT FROM AUTHOR]

Canonical correlation analysis (CCA) is a powerful statistical tool quantifying correlations between two sets of multidimensional variables. CCA cannot detect nonlinear relationship, and it is costly to derive canonical variates for high-dimensional data. Kernel CCA, a nonlinear extension of the CCA method, can efficiently exploit nonlinear relations and reduce high dimensionality. However, kernel CCA yields the so called over-fitting phenomenon in the high-dimensional feature space. To handle the shortcomings of kernel CCA, this paper develops a novel robust kernel CCA algorithm (KCCA-ROB). The derived method begins with reformulating the traditional generalized eigenvalue–eigenvector problem into a new framework. Under this novel framework, we develop a stable and fast algorithm by means of singular value decomposition (SVD) method. Experimental results on both a simulated dataset and real-world datasets demonstrate the effectiveness of the developed method. [ABSTRACT FROM AUTHOR]

One of the basic principles of Approximation Theory is that the quality of approximations increase with the smoothness of the function to be approximated. Functions that are smooth in certain subdomains will have good approximations in those subdomains, and these sub-approximations can possibly be calculated efficiently in parallel, as long as the subdomains do not overlap. This paper proposes an algorithm that first calculates sub-approximations on non-overlapping subdomains, then extends the subdomains as much as possible and finally produces a global solution on the given domain by letting the subdomains fill the whole domain. Consequently, there will be no Gibbs phenomenon along the boundaries of the subdomains. The method detects faults and gradient faults with good accuracy. Throughout, the algorithm works for fixed scattered input data of the function itself, not on spectral data, and it does not resample. [ABSTRACT FROM AUTHOR]

In the Big Data Era, the management of energy consumption by servers and data centers has become a challenging issue for companies, institutions, and countries. In data-centric applications, Database Management Systems are one of the major energy consumers when executing complex queries involving very large databases. Several initiatives have been proposed to deal with this issue, covering both the hardware and software dimensions. They can be classified in two main approaches assuming that either (a) the database is already deployed on a given platform, or (b) it is not yet deployed . In this study, we focus on the first set of initiatives with a particular interest in physical design, where optimization structures (e.g., indexes, materialized views) are selected to satisfy a given set of non-functional requirements such as query performance for a given workload. In this paper, we first propose an initiative, called Eco-Physic , which integrates the energy dimension into the physical design when selecting materialized views, one of the redundant optimization structures. Secondly, we provide a multi-objective formalization of the materialized view selection problem, considering two non-functional requirements: query performance and energy consumption while executing a given workload. Thirdly, an evolutionary algorithm is developed to solve the problem. This algorithm differs from the existing ones by being interactive, so that database administrators can adjust some energy sensitive parameters at the final stage of the algorithm execution according to their specifications. Finally, intensive experiments are conducted using our mathematical cost model and a real device for energy measurements. Results underscore the value of our approach as an effective way to save energy while optimizing queries through materialized views structures. [ABSTRACT FROM AUTHOR]

Nowadays, many data sources that include multi-label learning and multi-label classification have emerged in recent application areas. To achieve high classification accuracy, the multi-label feature selection method has received much attention because its accuracy can be significantly improved by selecting important features. In previous multi-label feature selection studies, a score function was designed based on the measure of the dependency between features and labels. However, identifying the optimal feature subset is an impractical task because all possible feature subsets are 2 N , where N is the number of total features in a given dataset. Thus, the conventional methods utilized a greedy search approach that can be stuck in local optima. To circumvent the drawback of the greedy approaches, we design a score function based on mutual information and present a numerical optimization approach to avoid being stuck in the local optima. The experimental results demonstrate the superiority of the proposed multi-label feature selection method. [ABSTRACT FROM AUTHOR]

Development of approaches for reducing the prediction error has been an active research field in collaborative filtering recommender systems since the accuracy of the prediction plays a crucial role in user purchase preferences. Unlike the conventional collaborative filtering methods which directly use the computed user-to-user similarity values, this paper presents a genetic algorithm approach for refining them before using in the prediction process. The approach was found to yield promising results according to the statistical analysis performed on a variety numbers of neighbours for various similarity metrics including Pearson’s Correlation, Extended Jaccard Coefficient and Vector Cosine Similarity along with a metric that assigns random weights to be used as a benchmark. Results show that the evolutionary approach has significantly reduced the prediction error using the evolved weights and Vector Cosine Similarity has shown the best performance. [ABSTRACT FROM AUTHOR]

The first intention of this paper is to emphasize equivalence between two well known general representations of outer inverses with prescribed range and null space of a given matrix. Two dynamical state equations, corresponding to particular expressions which are included in these representations, are defined within the second objective. In this way, two gradient based neural networks, initiated by introduced dynamical equations, are exploited in generating the class of outer inverses. The results corresponding to most common generalized inverses are obtained in particular cases. Numerical results are presented. [ABSTRACT FROM AUTHOR]