This paper improves a recently developed multi-objective particle swarm optimizer (D2MOPSO) that incorporates dominance with decomposition used in the context of optimization. Decomposition simplifies problem (MOP) by transforming it to set aggregation problems, whereas plays major role building leaders' archive. D2MOPSO introduces new archiving technique facilitates attaining better diversity and coverage both objective solution spaces. The improved method is evaluated on standard...

In this paper, we propose a novel Convolutional Neural Network (CNN) approach for the classification of raw dry-EEG signals without any data pre-processing. To illustrate effectiveness our approach, utilise Steady State Visual Evoked Potential (SSVEP) paradigm as use case. SSVEP can be utilised to allow people with severe physical disabilities such Complete Locked-In Syndrome or Amyotrophic Lateral Sclerosis aided via BCI applications, it requires only subject fixate upon sensory stimuli...

Despite significant recent progress in the area of Brain-Computer Interface (BCI), there are numerous shortcomings associated with collecting Electroencephalography (EEG) signals real-world environments. These include, but not limited to, subject and session data variance, long arduous calibration processes predictive generalisation issues across different subjects or sessions. This implies that many downstream applications, including Steady State Visual Evoked Potential (SSVEP) based...

Chest X-Ray (CXR) images play a crucial role in clinical practice, providing vital support for diagnosis and treatment. Augmenting the CXR dataset with synthetically generated annotated radiology reports can enhance performance of deep learning models various tasks. However, existing studies have primarily focused on generating unimodal data either or reports. In this study, we propose an integrated model, CXR-IRGen, designed specifically image-report pairs. Our model follows modularized...

A common goal of mechanistic interpretability is to decompose the activations neural networks into features: interpretable properties input computed by model. Sparse autoencoders (SAEs) are a popular method for finding these features in LLMs, and it has been postulated that they can be used find \textit{canonical} set units: unique complete list atomic features. We cast doubt on this belief using two novel techniques: SAE stitching show incomplete, meta-SAEs not atomic. involves inserting or...

Automatic Personality Perception is the task of automatically predicting personality traits people attribute to others. This work presents experiments where such a performed by mapping facial appearance into Big-Five traits, namely Openness, Conscientiousness, Extraversion, Agreeableness and Neuroticism. The are over pictures FERET corpus, originally collected for biometrics purposes, total 829 individuals. results show that it possible predict whether person perceived be above or below...

Recent work reports disparate performance for intersectional racial groups across face recognition tasks: verification and identification. However, the definition of those has a significant impact on underlying findings such bias analysis. Previous studies define these based either demographic information (e.g. African, Asian etc.) or skin tone lighter darker skins). The use sensitive broad group definitions disadvantages investigation subsequent counter-bias solutions design. By contrast,...

The development of natural language processing techniques for deriving useful information from unstructured clinical narratives is a fast-paced and rapidly evolving area machine learning research. Large volumes veterinary now exist curated by projects such as the Small Animal Veterinary Surveillance Network (SAVSNET) VetCompass, application to these datasets already (and will continue to) improve our understanding disease patterns within medicine. In part one this two article series, we...

This paper addresses the challenge of humanoid robot teleoperation in a natural indoor environment via Brain-Computer Interface (BCI). We leverage deep Convolutional Neural Network (CNN) based image and signal understanding to facilitate both real-time bject detection dry-Electroencephalography (EEG) human cortical brain bio-signals decoding. employ recent advances dry-EEG technology stream collect waveforms from subjects while they fixate on variable Steady State Visual Evoked Potential...

Deep Learning of neural networks has progressively become more prominent in healthcare with models reaching, or even surpassing, expert accuracy levels. However, these success stories are tainted by concerning reports on the lack model transparency and bias against some medical conditions patients' sub-groups. Explainable methods considered gateway to alleviate many concerns. In this study we demonstrate that generated explanations volatile changes training perpendicular classification task...

Abstract Effective public health surveillance requires consistent monitoring of disease signals such that researchers and decision-makers can react dynamically to changes in occurrence. However, whilst initiatives exist production animal veterinary medicine, comparable frameworks for companion animals are lacking. First-opinion electronic records (EHRs) have the potential reveal often represent initial reporting clinical syndromes presenting medical attention, highlighting their possible...

Explainable machine learning has become increasingly prevalent, especially in healthcare where explainable models are vital for ethical and trusted automated decision making. Work on the susceptibility of deep to adversarial attacks shown ease designing samples mislead a model into making incorrect predictions. In this work, we propose agnostic explainability-based method accurate detection two datasets with different complexity properties: Electronic Health Record (EHR) chest X-ray (CXR)...

Clinical data, such as evaluations, treatments, vital sign and lab test results, are usually observed recorded in hospital systems. Making use of data to help physicians evaluate the mortality risk in-hospital patients provides an invaluable source information that can ultimately with improving healthcare services. In particular, quick accurate predictions be valuable for who making decisions about interventions. this work we introduce a predictive Deep Learning model patients. Stacked...

Background Predicting the risk of glycated hemoglobin (HbA1c) elevation can help identify patients with potential for developing serious chronic health problems, such as diabetes. Early preventive interventions based upon advanced predictive models using electronic records data identifying ultimately provide better outcomes. Objective Our study investigated performance to forecast HbA1c levels by employing several machine learning models. We also examined use patient record longitudinal in...

In this study, we address the challenging problem of automatic detection transient deformation Earth's crust in time series differential satellite radar [interferometric synthetic aperture (InSAR)] images. The these events is important for a wide range natural hazard and solid earth applications, InSAR an ideal data source purpose due to its frequent global observational coverage. However, size dataset precludes systematic manual analysis, low signal-to-noise ratio makes task difficult. We...

TVQA is a large scale video question answering (video-QA) dataset based on popular TV shows. The questions were specifically designed to require "both vision and language understanding answer". In this work, we demonstrate an inherent bias in the towards textual subtitle modality. We infer said both directly indirectly, notably finding that models trained with subtitles learn, on-average, suppress feature contribution. Our results only visual information can answer ~45% of questions, while...

XAI with natural language processing aims to produce human-readable explanations as evidence for AI decision-making, which addresses explainability and transparency. However, from an HCI perspective, the current approaches only focus on delivering a single explanation, fails account diversity of human thoughts experiences in language. This paper thus this gap, by proposing generative framework, INTERACTION (explain aNd predicT thEn queRy contextuAl CondiTional varIational autO-eNcoder). Our...