In recent years, great progress has been made in the field of forecasting meteorological variables. Recently, deep learning architectures have made a major breakthrough in forecasting the daily average temperature over a ten-day horizon. However, advances in forecasting events related to the maximum temperature over short horizons remain a challenge for the community. A problem that is even more complex consists in making predictions of the maximum daily temperatures in the short, medium, and long term. In this work, we focus on forecasting events related to the maximum daily temperature over medium-term periods (90 days). Therefore, instead of addressing the problem from a meteorological point of view, this article tackles it from a climatological point of view. Due to the complexity of this problem, a common approach is to frame the study as a temporal classification problem with the classes: maximum temperature above normal, normal or below normal. From a practical point of view, we created a large historical dataset (from 1981 to 2018) collecting information from weather stations located in South America. In addition, we also integrated exogenous information from the Pacific, Atlantic, and Indian Ocean basins. We applied the AutoGluonTS platform to solve the above-mentioned problem. This AutoML tool shows competitive forecasting performance with respect to large operational platforms dedicated to tackling this climatological problem; but with a "relatively" low computational cost in terms of time and resources.

Since the 1990s, Self-Organizing Maps (SOMs) have been instrumental in reducing dimensionality and visualizing high-dimensional data.This study adapts SOMs to explore the neural representation of human concepts, their neural 'word net' mapping, and the deterioration of these mappings in certain neurological disorders.Our model draws inspiration from semantic dementia, a severe condition that degrades semantic knowledge in the brain.Although our exploration utilizes a low-dimensional model -a rough simplification with respect of our brains -it successfully replicates observed clinical patterns.These promising results inspire further research to enhance our understanding of language pathophysiology in neurological disorders.

Modeling non-stationary data is a challenging problem in the field of continual learning, and data distribution shifts may result in negative consequences on the performance of a machine learning model. Classic learning tools are often vulnerable to perturbations of the input covariates, and are sensitive to outliers and noise, and some tools are based on rigid algebraic assumptions. Distribution shifts are frequently occurring due to changes in raw materials for production, seasonality, a different user base, or even adversarial attacks. Therefore, there is a need for more effective distribution shift detection techniques.In this work, we propose a continual learning framework for monitoring and detecting distribution changes. We explore the problem in a latent space generated by a bio-inspired self-organizing clustering and statistical aspects of the latent space. In particular, we investigate the projections made by two topology-preserving maps: the Self-Organizing Map and the Scale Invariant Map. Our method can be applied in both a supervised and an unsupervised context. We construct the assessment of changes in the data distribution as a comparison of Gaussian signals, making the proposed method fast and robust. We compare it to other unsupervised techniques, specifically Principal Component Analysis (PCA) and Kernel-PCA. Our comparison involves conducting experiments using sequences of images (based on MNIST and injected shifts with adversarial samples), chemical sensor measurements, and the environmental variable related to ozone levels. The empirical study reveals the potential of the proposed approach.

Detecting fake news is currently one of the critical challenges facing modern societies. The problem is particularly relevant, as disinformation is readily used for political warfare but can also cause significant harm to the health of citizens, such as by promoting false data on the harmfulness of selected therapies. One way to combat disinformation is to treat fake news detection as a machine learning task. This paper presents such an approach, which additionally addresses an important problem related to the non-stationarity characteristics of the fake news. We elaborated a stream data with the simulation of domain shift based on two popular benchmark datasets dedicated to the fake news classification problem (Kaggle Fake News and Constraint@AAAI2021–COVID19 Fake News Detection). The proposed learning system works in a Continual Learning (CL) framework and integrates a self domain shift adaptation in a machine learning scheme. The method was built following state-of-the-art techniques, that includes Word2Vec as a feature extractor and the LSTM model as a classifier. The performance of the approach has been evaluated over the generated data stream. The convenience of our approach is showed in the results, where the accuracy gain with respect to a CL approach without domain adaptation is observed to be significant.

The Echo State Network (ESN) is a class of Recurrent Neural Network with a large number of hidden-hidden weights (in the so-called reservoir). Canonical ESN and its variations have recently received significant attention due to their remarkable success in the modeling of non-linear dynamical systems. The reservoir is randomly connected with fixed weights that don't change in the learning process. Only the weights from reservoir to output are trained. Since the reservoir is fixed during the training procedure, we may wonder if the computational power of the recurrent structure is fully harnessed. In this article, we propose a new computational model of the ESN type, that represents the reservoir weights in the Fourier space and performs a fine-tuning of these weights applying genetic algorithms in the frequency domain. The main interest is that this procedure will work in a much smaller space compared to the classical ESN, thus providing a dimensionality reduction transformation of the initial method. The proposed technique allows us to exploit the benefits of the large recurrent structure avoiding the training problems of gradient-based method. We provide a detailed experimental study that demonstrates the good performances of our approach with well-known chaotic systems and real-world data.

Recently, continual learning has received a lot of attention. One of the significant problems is the occurrence of concept drift, which consists of changing probabilistic characteristics of the incoming data. In the case of the classification task, this phenomenon destabilizes the model’s performance and negatively affects the achieved prediction quality. Most current methods apply statistical learning and similarity analysis over the raw data. However, similarity analysis in streaming data remains a complex problem due to time limitation, non-precise values, fast decision speed, scalability, etc. This article introduces a novel method for monitoring changes in the probabilistic distribution of multi-dimensional data streams. As a measure of the rapidity of changes, we analyze the popular Kullback-Leibler divergence. During the experimental study, we show how to use this metric to predict the concept drift occurrence and understand its nature. The obtained results encourage further work on the proposed methods and its application in the real tasks where the prediction of the future appearance of concept drift plays a crucial role, such as predictive maintenance.

In this article, we apply metaheuristics and Neural Networks for classifying human mental activities using EEG signals. We developed a Brain-Computer Interface system that is able to classify mental concentration versus relaxation. We collect the brain information during specific activities of the subject. Besides, we selected the best combination of the input features using the following two metaheuristic techniques: Simulated Annealing and Geometric Particle Swarm Optimization. The classification problem is solved using Neural Networks. We show that is possible to identify the human concentration using few EEG signals. In addition, the proposed system is developed with a fast and robust learning technique that can be easily adapted according to each subject.

The Echo State Network (ESN) is a specific recurrent network, which has gained popularity during the last years. The model has a recurrent network named reservoir, that is fixed during the learning process. The reservoir is used for transforming the input space in a larger space. A fundamental property that provokes an impact on the model accuracy is the Echo State Property (ESP). There are two main theoretical results related to the ESP. First, a sufficient condition for the ESP existence that involves the singular values of the reservoir matrix. Second, a necessary condition for the ESP. The ESP can be violated according to the spectral radius value of the reservoir matrix. There is a theoretical gap between these necessary and sufficient conditions. This article presents an empirical analysis of the accuracy and the projections of reservoirs that satisfy this theoretical gap. It gives some insights about the generation of the reservoir matrix. From previous works, it is already known that the optimal accuracy is obtained near to the border of stability control of the dynamics. Then, according to our empirical results, we can see that this border seems to be closer to the sufficient conditions than to the necessary conditions of the ESP.

We present an approach for emotion recognition using information of the pupil. In last years, the pupil variables have been used as an assessment of emotional arousal. In this article, we generate signals of pupil size and gaze position monitored during image viewing. The emotions are provoked by visual stimuli of colored images. Those images were taken from the International Affective Picture System which has been the reference for objective emotional assessment based on visual stimuli. For recognising the emotions we use the evolution of the eye tracking data during a window of time. The learning dataset is composed by the evolution of the pupil size and the gaze position, and labels associated to the emotional states. We study two kinds of learning tools based on Neural Networks. We obtain promising empirical results that show the potential of using temporal learning tools for emotion recognition.

In last years -- especially due to the development of telecommunications -- fairness modelling has received a strong attention. This article presents an approach for categorizing unknown relations according to their "closeness" to known relations. We consider as reference relations, the well-known: Pareto dominance, Leximin and Proportional fairness relation. We simulate each relation generating a learning dataset that is used for learning Neural Networks. The learning performance evaluation is based in several metrics, which are used as a "signature" of each relation. Besides, we develop a new function that gives an estimation about the "closeness" between relations. This concept permits us to categorise a new dataset (generated by an unknown relation) according its "closeness" with the Pareto dominance, Leximin and Proportional fairness relations know relations. Our experimental results are coherent with the alpha fairness concept.

This article introduces a metaheuristic approach to solve a variation of the well-known Vehicle Routing Problem (VRP). We present a solution for the Multi-Trip VRP with Time Windows and Heterogeneous Fleet. We add constraints to the original VRP concerning the time and the customer supply. Time constraints concerns the time windows on each customer and time horizon within which customers must be satisfied. In respect of the customer supply, we consider a heterogeneous fleet where vehicles are allowed to do multiple trips. We propose a solution for the problem using a Local Search and the Simulated Annealing technique. In order to evaluate the performance of our approach, we tested the procedure on a set of benchmark scenarios widely used for the VRP.

In the last decade, a new computational paradigm was introduced in the field of Machine Learning, under the name of Reservoir Computing (RC). RC models are neural networks which a recurrent part (the reservoir) that does not participate in the learning process, and the rest of the system where no recurrence (no neural circuit) occurs. This approach has grown rapidly due to its success in solving learning tasks and other computational applications. Some success was also observed with another recently proposed neural network designed using Queueing Theory, the Random Neural Network (RandNN). Both approaches have good properties and identified drawbacks. In this paper, we propose a new RC model called Echo State Queueing Network (ESQN), where we use ideas coming from RandNNs for the design of the reservoir. ESQNs consist in ESNs where the reservoir has a new dynamics inspired by recurrent RandNNs. The paper positions ESQNs in the global Machine Learning area, and provides examples of their use and performances. We show on largely used benchmarks that ESQNs are very accurate tools, and we illustrate how they compare with standard ESNs.

At the beginning of the 2000s was introduced the Echo State Network model (ESN). The model has been successfully used in temporal learning tasks. In spite of its success in practical applications, the model can present some stability problems when the parameters are not well initialized. The stability of the model is associated with the spectrum of the weight matrix. To compute the spectra is an expensive tasks when the network is large. Below, the initialization of the network parameters can depend of the benchmark problem. In this paper, we investigate the performance of the L2-Boost procedure, one specific boosting technique, in time-series problems. We use an ensemble of ESNs which are randomly initialized without control of the spectral radius norm as weak predictors of the L2procedure. Therefore, the procedure consists only in a random initialization of an ensemble of ESNs following of the L2-Boost steps.We evaluate this procedure on 5 widely used time-series benchmarks. Furthermore, we compare this procedure with a baseline approach which consists of averaging the prediction of an ensemble of ESNs with different initial network settings.

This paper investigates the estimation of a real time-series benchmark: the solar irradiance forcasting. The global solar irradiance is an important variable in the production of renewable energy sources. These variable is very unstable and hard to be predicted. For the prediction, we use two new models for time-series modeling: Echo State Queueing Networks and Differential polynomial Neural Networks. Both tools have been proven to be efficient for forecasting and time-series modeling. We compare their performances for this particular data set.

In the last years a new approach for designing and training artificial Recurrent Neural Network (RNN) have been investigated under the name of Reservoir Computing (RC). One important model in the field of RC has been developed under the name of Echo State Networks (ESNs). Traditionally, an ESN uses a RNN with random untrained parameters called the reservoir. The Self-Organizing Map (SOM) and the Scale Invariant Map (SIM) are two methods of topographic maps which have been used in different tasks of unsupervised learning. Recently, new works show that is effective using the SOM to set values of the reservoir parameters. The primary goal of this work is to improve the performance of ESN using the another method SIM. Here, we present the description of these two topographic map methods and the way to apply its on the ESN initialization. We specify an original algorithm to set the reservoir weights using the SOM and SIM. Furthermore, we use artificial data set to compare the use of topographic maps to initialize the ESN with random initialization. Overall, our results show the aptitude of SIM and SOM to set the reservoir parameters.

Random neural networks (RNN) have been efficiently used as learning tools in many applications of different types. The learning procedure followed so far is the gradient descent one. In this paper we explore the use of the Levenberg—Marquardt (LM) optimization procedure, more powerful when it is applicable, together with one of its major extensions, the LM procedure with adaptive momentum. We show how these methods can be used with RNN and run several experiments to evaluate their performances. The use of these techniques in the case of RNN lead to similar conclusions than when using standard artificial neural network: they clearly improve the learning efficiency.