A5010392887- Rough Sets and Fuzzy Logic
- Data Mining Algorithms and Applications
- Biomedical Text Mining and Ontologies
- Infrastructure Maintenance and Monitoring
- Nuclear Engineering Thermal-Hydraulics
- Semantic Web and Ontologies
- Advanced Neural Network Applications
- Gene expression and cancer classification
- Fuzzy Logic and Control Systems
- Preterm Birth and Chorioamnionitis
- Advancements in Semiconductor Devices and Circuit Design
- Neonatal and fetal brain pathology
- CCD and CMOS Imaging Sensors
- Data Visualization and Analytics
- Metaheuristic Optimization Algorithms Research
- Machine Learning and Data Classification
- Advanced Text Analysis Techniques
- Data Quality and Management
- Insect and Arachnid Ecology and Behavior
- Lepidoptera: Biology and Taxonomy
- Radio Frequency Integrated Circuit Design
- Engineering and Test Systems
- Electrical and Bioimpedance Tomography
- Electronic Health Records Systems
- Machine Learning in Bioinformatics
University of Sheffield
2022-2024
Google (United States)
2021
Smithsonian Tropical Research Institute
2019
Leeds Beckett University
2016-2018
University of Manchester
2006-2018
Manchester Academic Health Science Centre
2018
Carnegie Mellon University
2010-2013
Sheffield Hallam University
2010-2013
Alaska Biological Research (United States)
2013
University of Hawaii System
2013
The genes that are required for organismal survival annotated as 'essential genes'. Identifying all the essential of an animal species can reveal critical functions needed during development organism. To inform studies on mouse development, we developed a supervised machine learning classifier based phenotype data from knockout experiments. We used this to predict essentiality lacking experimental data. Validation our predictions against blind test set recent indicated high level accuracy...
This paper proposes a Bayesian association rule mining algorithm (BAR) which combines the Apriori with networks. Two interesting-ness measures of rules: confidence (BC) and lift (BL) measure conditional dependence independence relationships between items are defined based on joint probabilities represented by networks rules. BAR outputs best rules according to BC BL. is evaluated for its performance using two anonymized clinical phenotype datasets from UCI Repository: Thyroid disease...
A set of well-integrated clinical terminologies is at the core delivering an efficient trial system. The design and outcomes a can be improved significantly through unambiguous consistent used in participating institute. However, due to lack generalised legal technical standards, heterogeneity exists between prominent as well within systems several levels, e.g., data, schema, medical codes. This article specifically addresses problem integrating local or proprietary with globally defined...
Feature selection algorithms select the most relevant features of a data set to improve classification performance machine learning classifiers trained using set. This paper proposes feature algorithm called ultiobjective genetic local search (MOGLS) which integrates 3-objective with heuristic find subsets maximum prediction accuracy, smallest sizes and minimum redundancy. The MOGLS is compared 4 algorithms: wrapper algorithm, correlation-based selection, mutual information ranking C4.5 on 8...
Classification of yeast genes based on their expression levels obtained from micro array hybridization experiments is an important and challenging application domain in data mining knowledge discovery. Over the past decade, neural networks support vector machines (SVMs) have achieved good results for classification. This paper presents a methodology which uses two to classify unseen levels. In order remove some noise deal with imbalanced class distribution dataset, pre-processing firstly...
Rough set feature selection (RSFS) can be used to improve classifier performance. RSFS removes redundant attributes whilst keeping important ones that preserve the classification power of original dataset. The subsets selected by are termed reducts. intersection all reducts is core. As works on discrete only, for real-valued datasets discretization real performed before RSFS. core size discretized determined process. Previous work has shown dataset critically affects performance This paper...
A 2GS/s 6-bit flash sub-ADC with an op-amp free track-and-hold (T&H) for use in 8GS/s 4-way time-interleaved ADC was implemented 45nm SOI CMOS. The T&H utilizes a passive charge sharing technique and achieves 4GHz input bandwidth at clock rate without op-amp. consumes 74mW occupies area of 0.2mm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> . measured INL DNL are -0.9/1.0LSB -1.35/0.9LSB, respectively. SNDR is 33.9dB 125MHz 30.6dB input.
Outphasing power amplifiers (PAs) with reclamation networks are an attractive solution for maintaining a constant PA load during outphasing while minimizing energy lost. Power have only been demonstrated at frequencies below 2 GHz. In this work, we examine topologies and investigate how their efficiency scales frequency. Our results show that mm-wave frequencies, network is dominated by the current in resonant portion of rectifier loss can increase disproportionately frequency, thus placing...
One major accident of a nuclear power plant (NPP) is the loss coolant (LOCA) which caused by large break in an inlet header (IH) reactor. This work proposes constraint-based random search algorithm for optimizing neural network (NN) architectures and ensemble construction three stages detecting size IH NPP. In stage one, number 2-hidden layer, 3-hidden layer 4-hiddden are created using proposed constraint satisfaction algorithm. Then, optimised network, 4-hidden chosen from these training...
Early detection of the failure a nuclear system is an important topic in energy. This paper proposes three machine learning methodologies to detect modes (FM) Lead-Bismuth Eutectic eXperimental Accelerator Driven System (LBE-XADS) after first 10%, 50% and 90% time periods 3000 seconds mission LBEXADS. The methodology detects FM LBE-XADS 10% period consists two Gaussian mixture-based (GM-based) classifiers. second GM-based classifier neural network MLP1. third mode MLP2. proposed outperformed...
Feature selection is a pre-processing step for train- ing of classifiers in order to improve their performance. Rough Set Selection (RSFS) novel feature ap- proach. RSFS removes the redundant attributes only while keeping all important ones that preserve classification power original dataset. The subsets selected by are called reducts. intersection reducts core. This paper investigates effect on performance decision trees terms accuracy and number tree nodes. 9 datasets from different...
In this innovative practice WIP paper, we present PyRoboCar, a low-cost deep neural network-based autonomous car project. PyRoboCar is small-scale replication of real self-driving using convolutional network (CNN), which takes images from front fisheye camera as input and produces steering angles output. uses similar architecture industry-level cars can drive itself in real-time camera, an additional Tensor Processing Unit, Raspberry Pi 4 platform. We have also made project open online,...
The development of a mobile medical robot using ER1 technology