Yamisleydi Salgueiro

Yamisleydi Salgueiro

Especialidad: Machine learning, multi-objective optimization, knowledge-driven optimization.
Yamisleydi es graduada en ciencias de la computación y máster en informática aplicada. Recibió su Ph.D. en ciencias técnicas de la Universidad Central de Las Villas, Cuba en 2017. Actualmente es académica del Departamento de Ingeniería Industrial de la Facultad de Ingeniería en la Universidad de Talca.

PUBLICACIONES

Publisher: Elsevier,  Knowledge-Based Systems  Link>

ABSTRACT

Time series similarity evaluation is a crucial processing task performed either as a stand-alone action or as a part of extensive data analysis schemes. Among essential procedures that rely on measuring time series similarity, we find time series clustering and classification. While the similarity of regular (not temporal) data frames is studied extensively, there are not many methods that account for the time flow. In particular, there is a need for methods that are easy to interpret by a human being. In this paper, we present a concept-based approach for time series similarity evaluation. Firstly, a global model describing a given dataset of time series (made of two or more time series) is built. Then, for each time series in the dataset, we create the corresponding local model. Comparing time series is performed with the aid of their local models instead of raw time series values. In the paper, the described processing scheme is implemented using fuzzy sets representing concepts. The proposed approach has been applied to the task of time series classification, yielding highly satisfactory results.


Publisher:  IEEE Computational Intelligence Magazine Link>

ABSTRACT

This paper proposes an algorithm called Forward Composition Propagation (FCP) to explain the predictions of feed-forward neural networks operating on structured classification problems. In the proposed FCP algorithm, each neuron is described by a composition vector indicating the role of each problem feature in that neuron. Composition vectors are initialized using a given input instance and subsequently propagated through the whole network until reaching the output layer. The sign of each composition value indicates whether the corresponding feature excites or inhibits the neuron, while the absolute value quantifies its impact. The FCP algorithm is executed on a post-hoc basis, i.e., once the learning process is completed. Aiming to illustrate the FCP algorithm, this paper develops a case study concerning bias detection in a fairness problem in which the ground truth is known. The simulation results show that the composition values closely align with the expected behavior of protected features. The source code and supplementary material for this paper are available at https://github.com/igraugar/fcp.

Publisher: Springer, Neural Computing and Applications  Link>

ABSTRACT

In this paper, we present a recurrent neural system named long short-term cognitive networks (LSTCNs) as a generalization of the short-term cognitive network (STCN) model. Such a generalization is motivated by the difficulty of forecasting very long time series efficiently. The LSTCN model can be defined as a collection of STCN blocks, each processing a specific time patch of the (multivariate) time series being modeled. In this neural ensemble, each block passes information to the subsequent one in the form of weight matrices representing the prior knowledge. As a second contribution, we propose a deterministic learning algorithm to compute the learnable weights while preserving the prior knowledge resulting from previous learning processes. As a third contribution, we introduce a feature influence score as a proxy to explain the forecasting process in multivariate time series. The simulations using three case studies show that our neural system reports small forecasting errors while being significantly faster than state-of-the-art recurrent models.


Publisher: Elsevier, Renewable Energy Link>

ABSTRACT

Time series processing is an essential aspect of wind turbine health monitoring. In this paper, we propose two new approaches for analyzing wind turbine health. Both methods are based on abstract concepts, implemented using fuzzy sets, which allow aggregating and summarizing the underlying raw data in terms of relative low, moderate, and high power production. By observing a change in concepts, we infer the difference in a turbine's health. The first method evaluates the decrease or increase in relatively high and low power production. This task is performed using a regression model. The second method evaluates the overall drift of extracted concepts. A significant drift indicates that the power production process undergoes fluctuations in time. Concepts are labeled using linguistic labels, which makes our model easier to interpret. We applied the proposed approach to publicly available data describing four wind turbines, while exploring different external conditions (wind speed and temperature). The simulation results have shown that turbines with IDs T07 and T06 degraded the most. Moreover, the deterioration was clearer when we analyzed data concerning relatively low atmospheric temperature and relatively high wind speed.


Publisher: MDPI, Link>

ABSTRACT

Software Defined Networking (SDN) is a communication alternative to increase the scalability and resilience of microgrid hierarchical control. The common architecture has a centralized and monolithic topology, where the controller is highly susceptible to latency problems, resiliency, and scalability issues. This paper proposes a novel and intelligent control network to improve the performance of microgrid communications, solving the typical drawback of monolithic SDN controllers. The SDN controller’s functionalities are segregated into microservices groups and distributed through a bare-metal Kubernetes cluster. Results are presented from PLECS hardware in the loop simulation to validate the seamless transition between standard hierarchical control to the SDN networked microgrid. The microservices significantly impact the performance of the SDN controller, decreasing the latency by 10.76% compared with a monolithic architecture. Furthermore, the proposed approach demonstrates a 42.23% decrease in packet loss versus monolithic topologies and a 53.41% reduction in recovery time during failures. Combining Kubernetes with SDN microservices can eliminate the single point of failure in hierarchical control, improve application recovery time, and enhance containerization benefits, including security and portability. This proposal represents a reference framework for future edge computing and intelligent control approaches in networked microgrids.


Publisher: Elsevier, Expert Systems with Applications Link>

ABSTRACT

In decision-making problems, attributes can be classified as quantitative or qualitative according to their nature. For quantitative attributes, it is often necessary to normalize their values for further treatment since they are usually expressed with different measurements and scales. On the other hand, linguistic terms used by decision-makers to express their preference on qualitative attributes need to be transformed into numerical values according to linguistic-number dependencies. However, existing methods devoted to normalizing quantitative attributes and expressing the linguistic terms with numerical values have some important limitations. As a first contribution, we discuss the main deficiency in the functions commonly used to normalize quantitative attributes. Our second contribution is devoted to introducing a new property that counteracts the detected limitations, preparing the ground to present a standardization format for quantitative attributes. Our last contribution consists of a new format to transform the linguistic terms into numerical values that consider the significance of each linguistic term for the experts. Towards the end, these contributions are synthesized into a new procedure to generate a normalized consensus matrix when solving decision-making problems.


Publisher: Elsevier, Knowledge-Based Systems Link>

ABSTRACT

This paper proposes a post-hoc explanation method for computing concept attribution in Fuzzy Cognitive Map (FCM) models used for scenario analysis, based on SHapley Additive exPlanations (SHAP) values. The proposal is inspired by the lack of approaches to exploit the often-claimed intrinsic interpretability of FCM models while considering their dynamic properties. Our method uses the initial activation values of concepts as input features, while the outputs are considered as the hidden states produced by the FCM model during the recurrent reasoning process. Hence, the relevance of neural concepts is computed taking into account the model’s dynamic properties and hidden states, which result from the interaction among the initial conditions, the weight matrix, the activation function, and the selected reasoning rule. The proposed post-hoc method can handle situations where the FCM model might not converge or converge to a unique fixed-point attractor where the final activation values of neural concepts are invariant. The effectiveness of the proposed approach is demonstrated through experiments conducted on real-world case studies.


Publisher: Elsevier, Expert Systems with Applications  Link>

ABSTRACT

Forecasting windmill time series is often the basis of other processes such as anomaly detection, health monitoring, or maintenance scheduling. The amount of data generated by windmill farms makes online learning the most viable strategy to follow. Such settings require retraining the model each time a new batch of data is available. However, updating the model with new information is often very expensive when using traditional Recurrent Neural Networks (RNNs). In this paper, we use Long Short-term Cognitive Networks (LSTCNs) to forecast windmill time series in online settings. These recently introduced neural systems consist of chained Short-term Cognitive Network blocks, each processing a temporal data chunk. The learning algorithm of these blocks is based on a very fast, deterministic learning rule that makes LSTCNs suitable for online learning tasks. The numerical simulations using a case study involving four windmills showed that our approach reported the lowest forecasting errors with respect to a simple RNN, a Long Short-term Memory, a Gated Recurrent Unit, and a Hidden Markov Model. What is perhaps more important is that the LSTCN approach is significantly faster than these state-of-the-art models.


Publisher: Elsevier, Pattern Recognition Link>

ABSTRACT

The paper presents the presumably correct decision sets as a tool to analyze uncertainty in the form of inconsistency in decision systems. As a first step, problem instances are gathered into three regions containing weak members, borderline members, and strong members. This is accomplished by using the membership degrees of instances to their neighborhoods while neglecting their actual labels. As a second step, we derive the presumably correct and incorrect sets by contrasting the decision classes determined by a neighborhood function with the actual decision classes. We extract these sets from either the regions containing strong members or the whole universe, which defines the strict and relaxed versions of our theoretical formalism. These sets allow isolating the instances difficult to handle by machine learning algorithms as they are responsible for inconsistent patterns. The simulations using synthetic and real-world datasets illustrate the advantages of our model compared to rough sets, which is deemed a solid state-of-the-art approach to cope with inconsistency. In particular, it is shown that we can increase the accuracy of selected classifiers up to 36% by weighting the presumably correct and incorrect instances during the training process.  

Publisher: IEEE Transactions on Cybernetics, Link>

ABSTRACT

Machine-learning solutions for pattern classification problems are nowadays widely deployed in society and industry. However, the lack of transparency and accountability of most accurate models often hinders their safe use. Thus, there is a clear need for developing explainable artificial intelligence mechanisms. There exist model-agnostic methods that summarize feature contributions, but their interpretability is limited to predictions made by black-box models. An open challenge is to develop models that have intrinsic interpretability and produce their own explanations, even for classes of models that are traditionally considered black boxes like (recurrent) neural networks. In this article, we propose a long-term cognitive network (LTCN) for interpretable pattern classification of structured data. Our method brings its own mechanism for providing explanations by quantifying the relevance of each feature in the decision process. For supporting the interpretability without affecting the performance, the model incorporates more flexibility through a quasi-nonlinear reasoning rule that allows controlling nonlinearity. Besides, we propose a recurrence-aware decision model that evades the issues posed by the unique fixed point while introducing a deterministic learning algorithm to compute the tunable parameters. The simulations show that our interpretable model obtains competitive results when compared to state-of-the-art white and black-box models.


agencia nacional de investigación y desarrollo
Edificio de Innovación UC, Piso 2
Vicuña Mackenna 4860
Macul, Chile