Supporting Scientific Research Through Machine and Deep Learning: Fluorescence Microscopy and Operational Intelligence Use Cases

Although the debate of what data science is has a long history and has not reached a complete consensus yet, Data Science can be summarized as the process of learning from data. Guided by the above vision, this thesis presents two independent data science projects developed in the scope of multidisciplinary applied research. The first part analyzes fluorescence microscopy images typically produced in life science experiments, where the objective is to count how many marked neuronal cells are present in each image. Aiming to automate the task for supporting research in the area, we propose a neural network architecture tuned specifically for this use case, cell ResUnet (c-ResUnet), and discuss the impact of alternative training strategies in overcoming particular challenges of our data. The approach provides good results in terms of both detection and counting, showing performance comparable to the interpretation of human operators. As a meaningful addition, we release the pre-trained model and the Fluorescent Neuronal Cells dataset collecting pixel-level annotations of where neuronal cells are located. In this way, we hope to help future research in the area and foster innovative methodologies for tackling similar problems. The second part deals with the problem of distributed data management in the context of LHC experiments, with a focus on supporting ATLAS operations concerning data transfer failures. In particular, we analyze error messages produced by failed transfers and propose a Machine Learning pipeline that leverages the word2vec language model and K-means clustering. This provides groups of similar errors that are presented to human operators as suggestions of potential issues to investigate. The approach is demonstrated on one full day of data, showing promising ability in understanding the message content and providing meaningful groupings, in line with previously reported incidents by human operators.

Data driven regularization models of non-linear ill-posed inverse problems in imaging

Imaging technologies are widely used in application fields such as natural sciences, engineering, medicine, and life sciences. A broad class of imaging problems reduces to solve ill-posed inverse problems (IPs). Traditional strategies to solve these ill-posed IPs rely on variational regularization methods, which are based on minimization of suitable energies, and make use of knowledge about the image formation model (forward operator) and prior knowledge on the solution, but lack in incorporating knowledge directly from data. On the other hand, the more recent learned approaches can easily learn the intricate statistics of images depending on a large set of data, but do not have a systematic method for incorporating prior knowledge about the image formation model. The main purpose of this thesis is to discuss data-driven image reconstruction methods which combine the benefits of these two different reconstruction strategies for the solution of highly nonlinear ill-posed inverse problems. Mathematical formulation and numerical approaches for image IPs, including linear as well as strongly nonlinear problems are described. More specifically we address the Electrical impedance Tomography (EIT) reconstruction problem by unrolling the regularized Gauss-Newton method and integrating the regularization learned by a data-adaptive neural network. Furthermore we investigate the solution of non-linear ill-posed IPs introducing a deep-PnP framework that integrates the graph convolutional denoiser into the proximal Gauss-Newton method with a practical application to the EIT, a recently introduced promising imaging technique. Efficient algorithms are then applied to the solution of the limited electrods problem in EIT, combining compressive sensing techniques and deep learning strategies. Finally, a transformer-based neural network architecture is adapted to restore the noisy solution of the Computed Tomography problem recovered using the filtered back-projection method.

The impact of architecture on the performance of artificial neural networks

A number of researchers have investigated the impact of network architecture on the performance of artificial neural networks. Particular attention has been paid to the impact on the performance of the multi-layer perceptron of architectural issues, and the use of various strategies to attain an optimal network structure. However, there are still perceived limitations with the multi-layer perceptron and networks that employ a different architecture to the multi-layer perceptron have gained in popularity in recent years, particularly, networks that implement a more localised solution, where the solution in one area of the problem space does not impact, or has a minimal impact, on other areas of the space. In this study, we discuss the major architectural issues affecting the performance of a multi-layer perceptron, before moving on to examine in detail the performance of a new localised network, namely the bumptree. The work presented here examines the impact on the performance of artificial neural networks of employing alternative networks to the long established multi-layer perceptron. In particular, networks that impose a solution where the impact of each parameter in the final network architecture has a localised impact on the problem space being modelled are examined. The alternatives examined are the radial basis function and bumptree neural networks, and the impact of architectural issues on the performance of these networks is examined. Particular attention is paid to the bumptree, with new techniques for both developing the bumptree structure and employing this structure to classify patterns being examined.

Symbolic Knowledge Extraction from Trained Neural Networks Governed by Lukasiewicz Logics

This work describes a methodology to extract symbolic rules from trained neural networks. In our approach, patterns on the network are codified using formulas on a Lukasiewicz logic. For this we take advantage of the fact that every connective in this multi-valued logic can be evaluated by a neuron in an artificial network having, by activation function the identity truncated to zero and one. This fact simplifies symbolic rule extraction and allows the easy injection of formulas into a network architecture. We trained this type of neural network using a back-propagation algorithm based on Levenderg-Marquardt algorithm, where in each learning iteration, we restricted the knowledge dissemination in the network structure. This makes the descriptive power of produced neural networks similar to the descriptive power of Lukasiewicz logic language, minimizing the information loss on the translation between connectionist and symbolic structures. To avoid redundance on the generated network, the method simplifies them in a pruning phase, using the "Optimal Brain Surgeon" algorithm. We tested this method on the task of finding the formula used on the generation of a given truth table. For real data tests, we selected the Mushrooms data set, available on the UCI Machine Learning Repository.

Artificial neural networks (ANN): prediction of sensory measurements from instrumental data

The objective of this study was to predict by means of Artificial Neural Network (ANN), multilayer perceptrons, the texture attributes of light cheesecurds perceived by trained judges based on instrumental texture measurements. Inputs to the network were the instrumental texture measurements of light cheesecurd (imitative and fundamental parameters). Output variables were the sensory attributes consistency and spreadability. Nine light cheesecurd formulations composed of different combinations of fat and water were evaluated. The measurements obtained by the instrumental and sensory analyses of these formulations constituted the data set used for training and validation of the network. Network training was performed using a back-propagation algorithm. The network architecture selected was composed of 8-3-9-2 neurons in its layers, which quickly and accurately predicted the sensory texture attributes studied, showing a high correlation between the predicted and experimental values for the validation data set and excellent generalization ability, with a validation RMSE of 0.0506.

A Reinforcement Learning Approach to Economic Dispatch using Neural Networks

This paper presents a Reinforcement Learning (RL) approach to economic dispatch (ED) using Radial Basis Function neural network. We formulate the ED as an N stage decision making problem. We propose a novel architecture to store Qvalues and present a learning algorithm to learn the weights of the neural network. Even though many stochastic search techniques like simulated annealing, genetic algorithm and evolutionary programming have been applied to ED, they require searching for the optimal solution for each load demand. Also they find limitation in handling stochastic cost functions. In our approach once we learn the Q-values, we can find the dispatch for any load demand. We have recently proposed a RL approach to ED. In that approach, we could find only the optimum dispatch for a set of specified discrete values of power demand. The performance of the proposed algorithm is validated by taking IEEE 6 bus system, considering transmission losses

Approximation of non-autonomous dynamic systems by continuous time recurrent neural networks

This work provides a framework for the approximation of a dynamic system of the form x˙=f(x)+g(x)u by dynamic recurrent neural network. This extends previous work in which approximate realisation of autonomous dynamic systems was proven. Given certain conditions, the first p output neural units of a dynamic n-dimensional neural model approximate at a desired proximity a p-dimensional dynamic system with n>p. The neural architecture studied is then successfully implemented in a nonlinear multivariable system identification case study.

Contribuição para o estudo do embarque de uma rede neural artificial em field programmable gate array (FPGA)

This study shows the implementation and the embedding of an Artificial Neural Network (ANN) in hardware, or in a programmable device, as a field programmable gate array (FPGA). This work allowed the exploration of different implementations, described in VHDL, of multilayer perceptrons ANN. Due to the parallelism inherent to ANNs, there are disadvantages in software implementations due to the sequential nature of the Von Neumann architectures. As an alternative to this problem, there is a hardware implementation that allows to exploit all the parallelism implicit in this model. Currently, there is an increase in use of FPGAs as a platform to implement neural networks in hardware, exploiting the high processing power, low cost, ease of programming and ability to reconfigure the circuit, allowing the network to adapt to different applications. Given this context, the aim is to develop arrays of neural networks in hardware, a flexible architecture, in which it is possible to add or remove neurons, and mainly, modify the network topology, in order to enable a modular network of fixed-point arithmetic in a FPGA. Five synthesis of VHDL descriptions were produced: two for the neuron with one or two entrances, and three different architectures of ANN. The descriptions of the used architectures became very modular, easily allowing the increase or decrease of the number of neurons. As a result, some complete neural networks were implemented in FPGA, in fixed-point arithmetic, with a high-capacity parallel processing

Artificial neural networks for load flow and external equivalents studies

In this paper an artificial neural network (ANN) based methodology is proposed for (a) solving the basic load flow, (b) solving the load flow considering the reactive power limits of generation (PV) buses, (c) determining a good quality load flow starting point for ill-conditioned systems, and (d) computing static external equivalent circuits. An analysis of the input data required as well as the ANN architecture is presented. A multilayer perceptron trained with the Levenberg-Marquardt second order method is used. The proposed methodology was tested with the IEEE 30- and 57-bus, and an ill-conditioned 11-bus system. Normal operating conditions (base case) and several contingency situations including different load and generation scenarios have been considered. Simulation results show the excellent performance of the ANN for solving problems (a)-(d). (C) 2010 Elsevier B.V. All rights reserved.

Designing a modified Hopfield network to solve an Economic Dispatch problem with nonlinear cost function

Economic Dispatch (ED) problems have recently been solved by artificial neural networks approaches. In most of these dispatch models, the cost function must be linear or quadratic. Therefore, functions that have several minimum points represent a problem to the simulation since these approaches have not accepted nonlinear cost function. Another drawback pointed out in the literature is that some of these neural approaches fail to converge efficiently towards feasible equilibrium points. This paper discusses the application of a modified Hopfield architecture for solving ED problems defined by nonlinear cost function. The internal parameters of the neural network adopted here are computed using the valid-subspace technique, which guarantees convergence to equilibrium points that represent a solution for the ED problem. Simulation results and a comparative analysis involving a 3-bus test system are presented to illustrate efficiency of the proposed approach.

A multi-agent architecture for intelligent data analysis

The present study introduces a multi-agent architecture designed for doing automation process of data integration and intelligent data analysis. Different from other approaches the multi-agent architecture was designed using a multi-agent based methodology. Tropos, an agent based methodology was used for design. Based on the proposed architecture, we describe a Web based application where the agents are responsible to analyse petroleum well drilling data to identify possible abnormalities occurrence. The intelligent data analysis methods used was the Neural Network.

Análise da estabilidade transitória via rede neural Art-Artmap fuzzy Euclidiana modificada com treinamento continuado

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Análise da estabilidade estática de tensão de sistemas elétricos de potência usando uma rede neural baseada na teoria da ressonância adaptativa

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Correlação de poços com múltiplos perfis através da rede neural multicamadas

A correlação estratigráfica busca a determinação da continuidade lateral das rochas, ou a equivalência espacial entre unidades litológicas em subsuperfície, a partir de informações geológico-geofísicas oriundas de poços tubulares, que atravessam estas rochas. Normalmente, mas não exclusivamente, a correlação estratigráfica é realizada a partir das propriedades físicas registradas nos perfis geofísicos de poço. Neste caso, busca-se a equivalência litológica a partir da equivalência entre as propriedades físicas, medidas nos vários poços de um campo petrolífero. A técnica da correlação estratigráfica com perfis geofísicos de poço não é uma atividade trivial e sim, sujeita a inúmeras possibilidades de uma errônea interpretação da disposição geométrica ou da continuidade lateral das rochas em subsuperfície, em função da variabilidade geológica e da ambigüidade das respostas das ferramentas. Logo, é recomendável a utilização de um grande número de perfis de um mesmo poço, para uma melhor interpretação. A correlação estratigráfica é fundamental para o engenheiro de reservatório ou o geólogo, pois a partir da mesma, é possível a definição de estratégias de explotação de um campo petrolífero e a interpretação das continuidades hidráulicas dos reservatórios, bem como auxílio para a construção do modelo geológico para os reservatórios, a partir da interpretação do comportamento estrutural das diversas camadas em subsuperfície. Este trabalho apresenta um método de automação das atividades manuais envolvidas na correlação estratigráfica, com a utilização de vários perfis geofísicos de poço, através de uma arquitetura de rede neural artificial multicamadas, treinada com o algoritmo de retropropagação do erro. A correlação estratigráfica, obtida a partir da rede neural artificial, possibilita o transporte da informação geológica do datum de correlação ao longo do campo, possibilitando ao intérprete, uma visão espacial do comportamento do reservatório e a simulação dos possíveis paleoambientes. Com a metodologia aqui apresentada foi possível a construção automática de um bloco diagrama, mostrando a disposição espacial de uma camada argilosa, utilizando-se os perfis de Raio Gama (RG), Volume de Argila (Vsh), Densidade (ρ_b) e de Porosidade Neutrônica (φ_n) selecionados em cinco poços da região do Lago Maracaibo, na Venezuela.

Network-based high level data classification

Traditional supervised data classification considers only physical features (e. g., distance or similarity) of the input data. Here, this type of learning is called low level classification. On the other hand, the human (animal) brain performs both low and high orders of learning and it has facility in identifying patterns according to the semantic meaning of the input data. Data classification that considers not only physical attributes but also the pattern formation is, here, referred to as high level classification. In this paper, we propose a hybrid classification technique that combines both types of learning. The low level term can be implemented by any classification technique, while the high level term is realized by the extraction of features of the underlying network constructed from the input data. Thus, the former classifies the test instances by their physical features or class topologies, while the latter measures the compliance of the test instances to the pattern formation of the data. Our study shows that the proposed technique not only can realize classification according to the pattern formation, but also is able to improve the performance of traditional classification techniques. Furthermore, as the class configuration's complexity increases, such as the mixture among different classes, a larger portion of the high level term is required to get correct classification. This feature confirms that the high level classification has a special importance in complex situations of classification. Finally, we show how the proposed technique can be employed in a real-world application, where it is capable of identifying variations and distortions of handwritten digit images. As a result, it supplies an improvement in the overall pattern recognition rate.