Physical Sciences and Mathematics Commons^™

Using Chatgpt To Generate Gendered Language, Shweta Soundararajan, Manuela Nayantara Jeyaraj, Sarah Jane Delany

Conference papers

Gendered language is the use of words that denote an individual's gender. This can be explicit where the gender is evident in the actual word used, e.g. mother, she, man, but it can also be implicit where social roles or behaviours can signal an individual's gender - for example, expectations that women display communal traits (e.g., affectionate, caring, gentle) and men display agentic traits (e.g., assertive, competitive, decisive). The use of gendered language in NLP systems can perpetuate gender stereotypes and bias. This paper proposes an approach to generating gendered language datasets using ChatGPT which will provide data for data-driven …

Poly-Gan: Regularizing Polygons With Generative Adversarial Networks, Lasith Niroshan, James Carswell

Conference Papers

Regularizing polygons involves simplifying irregular and noisy shapes of built environment objects (e.g. buildings) to ensure that they are accurately represented using a minimum number of vertices. It is a vital processing step when creating/transmitting online digital maps so that they occupy minimal storage space and bandwidth. This paper presents a data-driven and Deep Learning (DL) based approach for regularizing OpenStreetMap building polygon edges. The study introduces a building footprint regularization technique (Poly-GAN) that utilises a Generative Adversarial Network model trained on irregular building footprints and OSM vector data. The proposed method is particularly relevant for map features …

Wifi-Based Human Activity Recognition Using Attention-Based Bilstm, Amany Elkelany, Robert J. Ross, Susan Mckeever

Conference papers

Recently, significant efforts have been made to explore human activity recognition (HAR) techniques that use information gathered by existing indoor wireless infrastructures through WiFi signals without demanding the monitored subject to carry a dedicated device. The key intuition is that different activities introduce different multi-paths in WiFi signals and generate different patterns in the time series of channel state information (CSI). In this paper, we propose and evaluate a full pipeline for a CSI-based human activity recognition framework for 12 activities in three different spatial environments using two deep learning models: ABiLSTM and CNN-ABiLSTM. Evaluation experiments have demonstrated that the …

Determining Child Sexual Abuse Posts Based On Artificial Intelligence, Susan Mckeever, Christina Thorpe, Vuong Ngo

Conference papers

The volume of child sexual abuse materials (CSAM) created and shared daily both surface web platforms such as Twitter and dark web forums is very high. Based on volume, it is not viable for human experts to intercept or identify CSAM manually. However, automatically detecting and analysing child sexual abusive language in online text is challenging and time-intensive, mostly due to the variety of data formats and privacy constraints of hosting platforms. We propose a CSAM detection intelligence algorithm based on natural language processing and machine learning techniques. Our CSAM detection model is not only used to remove CSAM on …

The Interaction Of Normalisation And Clustering In Sub-Domain Definition For Multi-Source Transfer Learning Based Time Series Anomaly Detection, Matthew Nicholson, Rahul Agrahari, Clare Conran, Haythem Assem, John D. Kelleher

Articles

This paper examines how data normalisation and clustering interact in the definition of sub-domains within multi-source transfer learning systems for time series anomaly detection. The paper introduces a distinction between (i) clustering as a primary/direct method for anomaly detection, and (ii) clustering as a method for identifying sub-domains within the source or target datasets. Reporting the results of three sets of experiments, we find that normalisation after feature extraction and before clustering results in the best performance for anomaly detection. Interestingly, we find that in the multi-source transfer learning scenario clustering on the target dataset and identifying subdomains in the …

Identity Term Sampling For Measuring Gender Bias In Training Data, Nasim Sobhani, Sarah Jane Delany

Conference Papers

Predictions from machine learning models can reflect biases in the data on which they are trained. Gender bias has been identified in natural language processing systems such as those used for recruitment. The development of approaches to mitigate gender bias in training data typically need to be able to isolate the effect of gender on the output to see the impact of gender. While it is possible to isolate and identify gender for some types of training data, e.g. CVs in recruitment, for most textual corpora there is no obvious gender label. This paper proposes a general approach to measure …

An Investigation Of The Reconstruction Capacity Of Stacked Convolutional Autoencoders For Log-Mel-Spectrograms, Anastasia Natsiou, Luca Longo, Seán O'Leary

Conference Papers

In audio processing applications, the generation of expressive sounds based on high-level representations demonstrates a high demand. These representations can be used to manipulate the timbre and influence the synthesis of creative instrumental notes. Modern algorithms, such as neural networks, have inspired the development of expressive synthesizers based on musical instrument timbre compression. Unsupervised deep learning methods can achieve audio compression by training the network to learn a mapping from waveforms or spectrograms to low-dimensional representations. This study investigates the use of stacked convolutional autoencoders for the compression of time-frequency audio representations for a variety of instruments for a single …

Self-Supervised Learning For Invariant Representations From Multi-Spectral And Sar Images, Pallavi Jain, Bianca Schoen Phelan, Robert J. Ross

Articles

Self-Supervised learning (SSL) has become the new state of the art in several domain classification and segmentation tasks. One popular category of SSL are distillation networks such as Bootstrap Your Own Latent (BYOL). This work proposes RS-BYOL, which builds on BYOL in the remote sensing (RS) domain where data are non-trivially different from natural RGB images. Since multi-spectral (MS) and synthetic aperture radar (SAR) sensors provide varied spectral and spatial resolution information, we utilise them as an implicit augmentation to learn invariant feature embeddings. In order to learn RS based invariant features with SSL, we trained RS-BYOL in two ways, …

Machine Learning With Kay, Lasith Niroshan, James Carswell

Conference Papers

Computational power is very important when training Deep Learning (DL) models with large amounts of data (Wooldridge, 2021). Hence, High-Performance Computing (HPC) can be leveraged to reduce computational cost, and the Irish Centre for High-End Computing (ICHEC) provides significant infrastructure and services for research and development to both academia and industry. A portion of ICHEC's HPC system has been allocated for institutional access, and this paper presents a case study of how to use Kay (Ireland's national supercomputer) in the remote sensing domain. Specifically, this study uses clusters of Kay Graphics Processing Units (GPUs) for training DL models to extract …

Assessing Feature Representations For Instance-Based Cross-Domain Anomaly Detection In Cloud Services Univariate Time Series Data, Rahul Agrahari, Matthew Nicholson, Clare Conran, Haythem Assem, John D. Kelleher

Articles

In this paper, we compare and assess the efficacy of a number of time-series instance feature representations for anomaly detection. To assess whether there are statistically significant differences between different feature representations for anomaly detection in a time series, we calculate and compare confidence intervals on the average performance of different feature sets across a number of different model types and cross-domain time-series datasets. Our results indicate that the catch22 time-series feature set augmented with features based on rolling mean and variance performs best on average, and that the difference in performance between this feature set and the next best …

Explaining Deep Learning Models For Tabular Data Using Layer-Wise Relevance Propagation, Ihsan Ullah, Andre Rios, Vaibhov Gala, Susan Mckeever

Articles

Trust and credibility in machine learning models are bolstered by the ability of a model to explain its decisions. While explainability of deep learning models is a well-known challenge, a further challenge is clarity of the explanation itself for relevant stakeholders of the model. Layer-wise Relevance Propagation (LRP), an established explainability technique developed for deep models in computer vision, provides intuitive human-readable heat maps of input images. We present the novel application of LRP with tabular datasets containing mixed data (categorical and numerical) using a deep neural network (1D-CNN), for Credit Card Fraud detection and Telecom Customer Churn prediction use …

Notions Of Explainability And Evaluation Approaches For Explainable Artificial Intelligence, Giulia Vilone, Luca Longo

Articles

Explainable Artificial Intelligence (XAI) has experienced a significant growth over the last few years. This is due to the widespread application of machine learning, particularly deep learning, that has led to the development of highly accurate models that lack explainability and interpretability. A plethora of methods to tackle this problem have been proposed, developed and tested, coupled with several studies attempting to define the concept of explainability and its evaluation. This systematic review contributes to the body of knowledge by clustering all the scientific studies via a hierarchical system that classifies theories and notions related to the concept of explainability …

A Quantitative Evaluation Of Global, Rule-Based Explanations Of Post-Hoc, Model Agnostic Methods, Giulia Vilone, Luca Longo

Articles

Understanding the inferences of data-driven, machine-learned models can be seen as a process that discloses the relationships between their input and output. These relationships consist and can be represented as a set of inference rules. However, the models usually do not explicit these rules to their end-users who, subsequently, perceive them as black-boxes and might not trust their predictions. Therefore, scholars have proposed several methods for extracting rules from data-driven machine-learned models to explain their logic. However, limited work exists on the evaluation and comparison of these methods. This study proposes a novel comparative approach to evaluate and compare the …

Exploring The Personality Of Virtual Tutors In Conversational Foreign Language Practice, Johanna Dobbriner, Cathy Ennis, Robert J. Ross

Conference papers

Fluid interaction between virtual agents and humans requires the understanding of many issues of conversational pragmatics. One such issue is the interaction between communication strategy and personality. As a step towards developing models of personality driven pragmatics policies, in this paper, we present our initial experiment to explore differences in user interaction with two contrasting avatar personalities. Each user saw a single personality in a video-call setting and gave feedback on the interaction. Our expectations, that a more extroverted outgoing positive personality would be a more successful tutor, were only partially confirmed. While this personality did induce longer conversations in …

Classification Of Explainable Artificial Intelligence Methods Through Their Output Formats, Giulia Vilone, Luca Longo

Articles

Machine and deep learning have proven their utility to generate data-driven models with high accuracy and precision. However, their non-linear, complex structures are often difficult to interpret. Consequently, many scholars have developed a plethora of methods to explain their functioning and the logic of their inferences. This systematic review aimed to organise these methods into a hierarchical classification system that builds upon and extends existing taxonomies by adding a significant dimension—the output formats. The reviewed scientific papers were retrieved by conducting an initial search on Google Scholar with the keywords “explainable artificial intelligence”; “explainable machine learning”; and “interpretable machine learning”. …

Multi-Modal Self-Supervised Representation Learning For Earth Observation, Pallavi Jain, Bianca Schoen Phelan, Robert J. Ross

Conference papers

Self-Supervised learning (SSL) has reduced the performance gap between supervised and unsupervised learning, due to its ability to learn invariant representations. This is a boon to the domains like Earth Observation (EO), where labelled data availability is scarce but unlabelled data is freely available. While Transfer Learning from generic RGB pre-trained models is still common-place in EO, we argue that, it is essential to have good EO domain specific pre-trained model in order to use with downstream tasks with limited labelled data. Hence, we explored the applicability of SSL with multi-modal satellite imagery for downstream tasks. For this we utilised …

Flying Free: A Research Overview Of Deep Learning In Drone Navigation Autonomy, Thomas Lee, Susan Mckeever, Jane Courtney

Articles

With the rise of Deep Learning approaches in computer vision applications, significant strides have been made towards vehicular autonomy. Research activity in autonomous drone navigation has increased rapidly in the past five years, and drones are moving fast towards the ultimate goal of near-complete autonomy. However, while much work in the area focuses on specific tasks in drone navigation, the contribution to the overall goal of autonomy is often not assessed, and a comprehensive overview is needed. In this work, a taxonomy of drone navigation autonomy is established by mapping the definitions of vehicular autonomy levels, as defined by the …

An Analysis Of The Interpretability Of Neural Networks Trained On Magnetic Resonance Imaging For Stroke Outcome Prediction, Esra Zihni, John D. Kelleher, Bryony Mcgarry

Conference papers

Applying deep learning models to MRI scans of acute stroke patients to extract features that are indicative of short-term outcome could assist a clinician’s treatment decisions. Deep learning models are usually accurate but are not easily interpretable. Here, we trained a convolutional neural network on ADC maps from hyperacute ischaemic stroke patients for prediction of short-term functional outcome and used an interpretability technique to highlight regions in the ADC maps that were most important in the prediction of a bad outcome. Although highly accurate, the model’s predictions were not based on aspects of the ADC maps related to stroke pathophysiology.

Wider Vision: Enriching Convolutional Neural Networks Via Alignment To External Knowledge Bases, Xuehao Liu, Sarah Jane Delany, Susan Mckeever

Conference papers

Deep learning models suffer from opaqueness. For Convolutional Neural Networks (CNNs), current research strategies for explaining models focus on the target classes within the associated training dataset. As a result, the understanding of hidden feature map activations is limited by the discriminative knowledge gleaned during training. The aim of our work is to explain and expand CNNs models via the mirroring or alignment of the network to an external knowledge base. This will allow us to give a semantic context or label for each visual feature. Using the resultant aligned embedding space, we can match CNN feature activations to nodes …

Fairer Evaluation Of Zero Shot Action Recognition In Videos, Kaiqiang Huang, Sarah Jane Delany, Susan Mckeever

Conference Papers

Zero-shot learning (ZSL) for human action recognition (HAR) aims to recognise video action classes that have never been seen during model training. This is achieved by building mappings between visual and semantic embeddings. These visual embeddings are typically provided via a pre-trained deep neural network (DNN). The premise of ZSL is that the training and testing classes should be disjoint. In the parallel domain of ZSL for image input, the widespread poor evaluation protocol of pre-training on ZSL test classes has been highlighted. This is akin to providing a sneak preview of the evaluation classes. In this work, we investigate …

Virtual Tutor Personality In Computer Assisted Language Learning, Johanna Dobbriner, Cathy Ennis, Robert J. Ross

Conference papers

The use of intelligent virtual agents in language learning has increased in recent years. Studies into several aspects of personalisation aiming to increase user engagement are an ongoing research topic with avatar personality being one such aspect. As a step towards our development of intelligent virtual avatars, we present two of our initial experiments to explore differences in user interaction with two contrasting avatar personalities -- P1: open-minded, friendly and sociable and P2: closed-off, curt and distant. Each user interacted with a single personality in a video-call setting and gave feedback on the interaction. Our expectations, that P1 would be …

K-Nearest Neighbour Classifiers - A Tutorial, Padraig Cunningham, Sarah Jane Delany

Conference papers

Perhaps the most straightforward classifier in the arsenal or Machine Learning techniques is the Nearest Neighbour Classifier – classification is achieved by identifying the nearest neighbours to a query example and using those neighbours to determine the class of the query. This approach to classification is of particular importance because issues of poor run-time performance is not such a problem these days with the computational power that is available. This paper presents an overview of techniques for Nearest Neighbour classification focusing on; mechanisms for assessing similarity (distance), computational issues in identifying nearest neighbours and mechanisms for reducing the dimension of …

Zero-Shot Action Recognition With Knowledge Enhanced Generative Adversarial Networks, Kaiqiang Huang, Luis Miralles-Pechuán, Susan Mckeever

Conference papers

Zero-Shot Action Recognition (ZSAR) aims to recognise action classes in videos that have never been seen during model training. In some approaches, ZSAR has been achieved by generating visual features for unseen classes based on the semantic information of the unseen class labels using generative adversarial networks (GANs). Therefore, the problem is converted to standard supervised learning since the unseen visual features are accessible. This approach alleviates the lack of labelled samples of unseen classes. In addition, objects appearing in the action instances could be used to create enriched semantics of action classes and therefore, increase the accuracy of ZSAR. …

Data: The Good, The Bad And The Ethical, John D. Kelleher, Filipe Cabral Pinto, Luis M. Cortesao

Articles

It is often the case with new technologies that it is very hard to predict their long-term impacts and as a result, although new technology may be beneficial in the short term, it can still cause problems in the longer term. This is what happened with oil by-products in different areas: the use of plastic as a disposable material did not take into account the hundreds of years necessary for its decomposition and its related long-term environmental damage. Data is said to be the new oil. The message to be conveyed is associated with its intrinsic value. But as in …

Exploring The Potential Of Defeasible Argumentation For Quantitative Inferences In Real-World Contexts: An Assessment Of Computational Trust, Lucas Rizzo, Pierpaolo Dondio, Luca Longo

Articles

Argumentation has recently shown appealing properties for inference under uncertainty and conflicting knowledge. However, there is a lack of studies focused on the examination of its capacity of exploiting real-world knowledge bases for performing quantitative, case-by-case inferences. This study performs an analysis of the inferential capacity of a set of argument-based models, designed by a human reasoner, for the problem of trust assessment. Precisely, these models are exploited using data from Wikipedia, and are aimed at inferring the trustworthiness of its editors. A comparison against non-deductive approaches revealed that these models were superior according to values inferred to recognised trustworthy …

Lightgwas: A Novel Machine Learning Procedure For Genome-Wide Association Study, Ambrozio Bruno, Luca Longo, Lucas Rizzo

Articles

This paper proposes a novel machine learning procedure for genome-wide association study (GWAS), named LightGWAS. It is based on the LightGBM framework, in addition to being a single, resilient, autonomous and scalable solution to address common limitations of GWAS implementations found in the literature. These include reliance on massive manual quality control steps and specific GWAS methods for each type of dataset morphology and size. Through this research, LightGWAS has been contrasted against PLINK2, one of the current state-of-the-art for GWAS implementations based on general linear model with support to firth regularisation. The mean differences measured upon standard classification metrics, …

A Comparative Analysis Of Rule-Based, Model-Agnostic Methods For Explainable Artificial Intelligence, Giulia Vilone, Lucas Rizzo, Luca Longo

Articles

The ultimate goal of Explainable Artificial Intelligence is to build models that possess both high accuracy and degree of explainability. Understanding the inferences of such models can be seen as a process that discloses the relationships between their input and output. These relationships can be represented as a set of inference rules which are usually not explicit within a model. Scholars have proposed several methods for extracting rules from data-driven machine-learned models. However, limited work exists on their comparison. This study proposes a novel comparative approach to evaluate and compare the rulesets produced by four post-hoc rule extractors by employing …

Language-Driven Region Pointer Advancement For Controllable Image Captioning, Annika Lindh, Robert J. Ross, John D. Kelleher

Conference papers

Controllable Image Captioning is a recent sub-field in the multi-modal task of Image Captioning wherein constraints are placed on which regions in an image should be described in the generated natural language caption. This puts a stronger focus on producing more detailed descriptions, and opens the door for more end-user control over results. A vital component of the Controllable Image Captioning architecture is the mechanism that decides the timing of attending to each region through the advancement of a region pointer. In this paper, we propose a novel method for predicting the timing of region pointer advancement by treating the …

Energy-Based Neural Modelling For Large-Scale Multiple Domain Dialogue State Tracking, Anh Duong Trinh, Robert J. Ross, John D. Kelleher

Conference papers

Scaling up dialogue state tracking to multiple domains is challenging due to the growth in the number of variables being tracked. Furthermore, dialog state tracking models do not yet explicitly make use of relationships between dialogue variables, such as slots across domains. We propose using energy-based structure prediction methods for large-scale dialogue state tracking task in two multiple domain dialogue datasets. Our results indicate that: (i) modelling variable dependencies yields better results; and (ii) the structured prediction output aligns with the dialogue slot-value constraint principles. This leads to promising directions to improve state-of-the-art models by incorporating variable dependencies into their …

F-Measure Optimisation And Label Regularisation For Energy-Based Neural Dialogue State Tracking Models, Anh Duong Trinh, Robert J. Ross, John D. Kelleher

Conference papers

In recent years many multi-label classification methods have exploited label dependencies to improve performance of classification tasks in various domains, hence casting the tasks to structured prediction problems. We argue that multi-label predictions do not always satisfy domain constraint restrictions. For example when the dialogue state tracking task in task-oriented dialogue domains is solved with multi-label classification approaches, slot-value constraint rules should be enforced following real conversation scenarios.

To address these issues we propose an energy-based neural model to solve the dialogue state tracking task as a structured prediction problem. Furthermore we propose two improvements over previous methods with respect …