A5068590850- Anomaly Detection Techniques and Applications
- COVID-19 diagnosis using AI
- Machine Learning and Data Classification
- Adversarial Robustness in Machine Learning
- AI in cancer detection
- Imbalanced Data Classification Techniques
- Explainable Artificial Intelligence (XAI)
- Domain Adaptation and Few-Shot Learning
- Hydrological Forecasting Using AI
- Radiomics and Machine Learning in Medical Imaging
- Coffee research and impacts
- Spam and Phishing Detection
- Healthcare and Environmental Waste Management
- Recycling and Waste Management Techniques
- Crime Patterns and Interventions
- Stock Market Forecasting Methods
- Municipal Solid Waste Management
- Fuzzy Logic and Control Systems
- Data Stream Mining Techniques
- Muscle metabolism and nutrition
- Digital Media Forensic Detection
- Growth and nutrition in plants
- Time Series Analysis and Forecasting
- Tea Polyphenols and Effects
- Advanced Neural Network Applications
Universidade de São Paulo
2001-2025
IBM Research - Brazil
2024
Lancaster University
2019-2023
Duke University
2023
Fundo Brasil
2023
Universidade Federal de Lavras
2017-2020
Centro Universitário Cesumar
2019
Universidade Federal de São Carlos
2018
Abstract This paper provides a brief analytical review of the current state‐of‐the‐art in relation to explainability artificial intelligence context recent advances machine learning and deep learning. The starts with historical introduction taxonomy, formulates main challenges terms building on recently formulated National Institute Standards four principles explainability. Recently published methods related topic are then critically reviewed analyzed. Finally, future directions for research...
A bstract The COVID-19 disease has widely spread all over the world since beginning of 2020. On January 30, 2020 World Health Organization (WHO) declared a global health emergency. At time writing this paper number infected about 2 million people worldwide and took 125,000 lives, advanced public systems European countries as well USA were overwhelmed. In paper, we propose an eXplainable Deep Learning approach to detect from computer tomography (CT) - Scan images. rapid detection any case is...
As COVID-19 hounds the world, common cause of finding a swift solution to manage pandemic has brought together researchers, institutions, governments, and society at large. The Internet Things (IoT), artificial intelligence (AI)-including machine learning (ML) Big Data analytics-as well as Robotics Blockchain, are four decisive areas technological innovation that have been ingenuity harnessed fight this future ones. While these highly interrelated smart connected health technologies cannot...
A computer vision approach to classify garbage into recycling categories could be an efficient way process waste. This project aims take waste images and them four classes: glass, paper, metal and, plastic. We use a image database that contains around 400 for each class. The models used in the experiments are Pre-trained VGG-16 (VGG16), AlexNet, Support Vector Machine (SVM), K-Nearest Neighbor (KNN) Random Forest (RF). Experiments showed our reached accuracy 93%.
Traditionally, in supervised machine learning, (a significant) part of the available data (usually 50%-80%) is used for training and rest—for validation. In many problems, however, are highly imbalanced regard to different classes or does not have good coverage feasible space which, turn, creates problems validation usage phase. this paper, we propose a technique synthesizing likely help balance as well boost performance terms confusion matrix overall. The idea, nutshell, synthesize samples...
This article describes a novel approach to the problem of developing explainable machine learning models. We consider deep reinforcement (DRL) model representing highway path planning policy for autonomous driving [1]. The constitutes mapping from continuous multidimensional state space characterizing vehicle positions and velocities discrete set actions in longitudinal lateral direction. It is obtained by applying customized version double Q-network algorithm [2]. main idea approximate DRL...
Abstract The infection by SARS-CoV-2 which causes the COVID-19 disease has spread widely over whole world since beginning of 2020. Following epidemic started in Wuhan, China on January 30, 2020 World Health Organization (WHO) declared a global health emergency and pandemic. In this paper, we describe publicly available multiclass CT scan dataset for identification. Which currently contains 4173 CT-scans 210 different patients, out 2168 correspond to 80 patients infected with confirmed...
The cold brew method consists of soaking roasted and ground coffee beans either in or ambient water (4–23 °C) for up to 24 h. Using this technique, a drink with unique sensory profile is obtained. This study was conducted determine the shelf life organic (pH~5.0) made from subjected three roast levels: light, medium dark. pasteurized at 90 °C/30 s, ultra-clean filled into high-density polyethylene bottles, stored 4 °C Physicochemical, enzymic tests, instrumental color analysis,...
Abstract Recent advancements in large foundation models have revealed impressive capabilities mastering complex chemical language representations. These undergo a task-agnostic learning phase, characterized by pre-training on extensive unlabeled corpora followed fine-tuning specific downstream tasks. This methodology reduces reliance labeled data, facilitating data acquisition and broadening the scope of representation. However, real-world scenarios often pose challenges due to domain shift,...
Recidivism prediction provides decision makers with an assessment of the likelihood that a criminal defendant will reoffend can be used in pre-trial decision-making. It also for locations where crimes most occur, profiles are more likely to commit violent crimes. While such instruments gaining increasing popularity, their use is controversial as they may present potential discriminatory bias risk assessment. In this paper we propose new fair-by-design approach predict recidivism....
Abstract The Covid-19 disease has spread widely over the whole world since beginning of 2020. Following epidemic which started in Wuhan, China on January 30, 2020 World Health Organization (WHO) declared a global health emergency and pandemic. Researchers different disciplines work along with public officials to understand SARS-CoV-2 pathogenesis jointly policymakers urgently develop strategies control this new disease. Recent findings have observed specific image patterns from computed...
This paper presents an actively semi-supervised multi-layer neuro-fuzzy modeling method, ASSDRB, to classify different lighting conditions for driving scenes. ASSDRB is composed of a massively parallel ensemble AnYa type 0-order fuzzy rules. It uses recursive learning algorithm update its structure when new data items are provided and, therefore, able cope with nonstationarities. Different situations considered in the analysis, which used by self-driving cars as safety mechanism. Differently...
This paper describes a new self-organizing neuro-fuzzy approach to autonomously learn interpretable models by self-driving cars. A explainable architecture and density-based feature selection method are proposed. These approaches used classify different action states occurring from conditions. The proposed is able provide human understandable IF ... THEN rules representation due its learning engine which composed of massively parallel set 0-order fuzzy rules. based on the ranking densities...
In this paper we introduce the DMR - a prototype-based method and network architecture for deep learning which is using decision tree (DT)- based inference synthetic data to balance classes. It builds upon recently introduced xDNN addressing more complex multi-class problems, specifically when classes are highly imbalanced. moves away from direct on all towards layered DT of pair-wise class comparisons. addition, it forces prototypes be balanced between regardless possible imbalances...
The Internet of Things (IoT) has made it possible to include everyday objects in a connected network, allowing them intelligently process data and respond their environment. Thus, is expected that those will gain an intelligent understanding environment be able more efficiently than before. Particularly, such edge computing paradigm allowed the execution inference methods on resource-constrained devices as microcontrollers, significantly changing way IoT applications have evolved recent...
This paper presents a Gaussian fuzzy set-based evolving modeling method, FBeM-G, to predict tropical cyclone tracks 6 hours in advance. FBeM-G summarizes similar data into granules evolved from sequence of data. It uses recursive learning algorithm update its parameters and structure over time therefore is able cope with nonstationarities. Past values latitude, longitude, maximum sustained wind, pressure wind radii different quadrants the Katrina, Sandy Wilma cyclones were obtained `best...
Earth observation is fundamental for a range of human activities including flood response as it offers vital information to decision makers. Semantic segmentation plays key role in mapping the raw hyper-spectral data coming from satellites into understandable form assigning class labels each pixel. Traditionally, water index based methods have been used detecting pixels. More recently, deep learning techniques such U-Net started gain attention offering significantly higher accuracy. However,...
Deep neural networks (DNN's) have become es-sential for solving diverse complex problems and achieved considerable success in tackling computer vision tasks. How-ever, DNN's are vulnerable to human-imperceptible adversarial distortion/noise patterns that can detrimentally impact safety-critical applications such as autonomous driving. In this paper, we introduce a novel robust-by-design deep learning approach, Sim-DNN, is able detect attacks through its inner defense mechanism considers the...
In this paper we offer a method and algorithm, which make possible fully autonomous (unsupervised) detection of new classes, learning following very parsimonious training priming (few labeled data samples only). Moreover, unknown classes may appear at later stage the proposed xClass algorithm are able to successfully discover learn from autonomously. Furthermore, features (inputs classifier) automatically sub-selected by based on accumulated density per feature class. As result, highly...