Search

Large Language Models (LLMs) have demonstrated remarkable performance across various domains, including healthcare. However, their ability to effectively represent structured non-textual data, such as the alphanumeric medical codes used in records like ICD-10 or SNOMED-CT, is limited and has been particularly exposed in recent research. This paper examines the challenges LLMs face in processing medical codes due to the shortcomings of current tokenization methods. As a result, we introduce the UniStruct architecture to design a multimodal medical foundation model of unstructured text and structured data, which addresses these challenges by adapting subword tokenization techniques specifically for the structured medical codes. Our approach is validated through model pre-training on both an extensive internal medical database and a public repository of structured medical records. Trained on over 1 billion tokens on the internal medical database, the proposed model achieves up to a 23% improvement in evaluation metrics, with around 2% gain attributed to our proposed tokenization. Additionally, when evaluated on the EHRSHOT public benchmark with a 1/1000 fraction of the pre-training data, the UniStruct model improves performance on over 42% of the downstream tasks. Our approach not only enhances the representation and generalization capabilities of patient-centric models but also bridges a critical gap in representation learning models' ability to handle complex structured medical data, alongside unstructured text., Comment: NeurIPS 2024 Workshop on Unifying Representations in Neural Models (UniReps 2024)

Despite their impressive success, training foundation models remains computationally costly. This paper investigates how to efficiently train speech foundation models with self-supervised learning (SSL) under a limited compute budget. We examine critical factors in SSL that impact the budget, including model architecture, model size, and data size. Our goal is to make analytical steps toward understanding the training dynamics of speech foundation models. We benchmark SSL objectives in an entirely comparable setting and find that other factors contribute more significantly to the success of SSL. Our results show that slimmer model architectures outperform common small architectures under the same compute and parameter budget. We demonstrate that the size of the pre-training data remains crucial, even with data augmentation during SSL training, as performance suffers when iterating over limited data. Finally, we identify a trade-off between model size and data size, highlighting an optimal model size for a given compute budget., Comment: Accepted to IEEE SLT 2024

State-of-the-art Large Language Models (LLMs) are accredited with an increasing number of different capabilities, ranging from reading comprehension, over advanced mathematical and reasoning skills to possessing scientific knowledge. In this paper we focus on their multi-hop reasoning capability: the ability to identify and integrate information from multiple textual sources. Given the concerns with the presence of simplifying cues in existing multi-hop reasoning benchmarks, which allow models to circumvent the reasoning requirement, we set out to investigate, whether LLMs are prone to exploiting such simplifying cues. We find evidence that they indeed circumvent the requirement to perform multi-hop reasoning, but they do so in more subtle ways than what was reported about their fine-tuned pre-trained language model (PLM) predecessors. Motivated by this finding, we propose a challenging multi-hop reasoning benchmark, by generating seemingly plausible multi-hop reasoning chains, which ultimately lead to incorrect answers. We evaluate multiple open and proprietary state-of-the-art LLMs, and find that their performance to perform multi-hop reasoning is affected, as indicated by up to 45% relative decrease in F1 score when presented with such seemingly plausible alternatives. We conduct a deeper analysis and find evidence that while LLMs tend to ignore misleading lexical cues, misleading reasoning paths indeed present a significant challenge., Comment: 15 pages, 3 figures, EMNLP 2024 Main Conference

Automatic Speech Recognition (ASR) systems are pivotal in transcribing speech into text, yet the errors they introduce can significantly degrade the performance of downstream tasks like summarization. This issue is particularly pronounced in clinical dialogue summarization, a low-resource domain where supervised data for fine-tuning is scarce, necessitating the use of ASR models as black-box solutions. Employing conventional data augmentation for enhancing the noise robustness of summarization models is not feasible either due to the unavailability of sufficient medical dialogue audio recordings and corresponding ASR transcripts. To address this challenge, we propose MEDSAGE, an approach for generating synthetic samples for data augmentation using Large Language Models (LLMs). Specifically, we leverage the in-context learning capabilities of LLMs and instruct them to generate ASR-like errors based on a few available medical dialogue examples with audio recordings. Experimental results show that LLMs can effectively model ASR noise, and incorporating this noisy data into the training process significantly improves the robustness and accuracy of medical dialogue summarization systems. This approach addresses the challenges of noisy ASR outputs in critical applications, offering a robust solution to enhance the reliability of clinical dialogue summarization., Comment: Accepted by the Thirty-Ninth AAAI Conference on Artificial Intelligence (AAAI-25)

Large Language Models (LLMs) are increasingly adopted for applications in healthcare, reaching the performance of domain experts on tasks such as question answering and document summarisation. Despite their success on these tasks, it is unclear how well LLMs perform on tasks that are traditionally pursued in the biomedical domain, such as structured information extration. To breach this gap, in this paper, we systematically benchmark LLM performance in Medical Classification and Named Entity Recognition (NER) tasks. We aim to disentangle the contribution of different factors to the performance, particularly the impact of LLMs' task knowledge and reasoning capabilities, their (parametric) domain knowledge, and addition of external knowledge. To this end we evaluate various open LLMs -- including BioMistral and Llama-2 models -- on a diverse set of biomedical datasets, using standard prompting, Chain-of-Thought (CoT) and Self-Consistency based reasoning as well as Retrieval-Augmented Generation (RAG) with PubMed and Wikipedia corpora. Counter-intuitively, our results reveal that standard prompting consistently outperforms more complex techniques across both tasks, laying bare the limitations in the current application of CoT, self-consistency and RAG in the biomedical domain. Our findings suggest that advanced prompting methods developed for knowledge- or reasoning-intensive tasks, such as CoT or RAG, are not easily portable to biomedical tasks where precise structured outputs are required. This highlights the need for more effective integration of external knowledge and reasoning mechanisms in LLMs to enhance their performance in real-world biomedical applications., Comment: 11 pages

There is vivid research on adapting Large Language Models (LLMs) to perform a variety of tasks in high-stakes domains such as healthcare. Despite their popularity, there is a lack of understanding of the extent and contributing factors that allow LLMs to recall relevant knowledge and combine it with presented information in the clinical and biomedical domain: a fundamental pre-requisite for success on down-stream tasks. Addressing this gap, we use Multiple Choice and Abstractive Question Answering to conduct a large-scale empirical study on 22 datasets in three generalist and three specialist biomedical sub-domains. Our multifaceted analysis of the performance of 15 LLMs, further broken down by sub-domain, source of knowledge and model architecture, uncovers success factors such as instruction tuning that lead to improved recall and comprehension. We further show that while recently proposed domain-adapted models may lack adequate knowledge, directly fine-tuning on our collected medical knowledge datasets shows encouraging results, even generalising to unseen specialist sub-domains. We complement the quantitative results with a skill-oriented manual error analysis, which reveals a significant gap between the models' capabilities to simply recall necessary knowledge and to integrate it with the presented context. To foster research and collaboration in this field we share M-QALM, our resources, standardised methodology, and evaluation results, with the research community to facilitate further advancements in clinical knowledge representation learning within language models., Comment: Accepted at ACL 2024 (Findings)

Computerised clinical coding approaches aim to automate the process of assigning a set of codes to medical records. While there is active research pushing the state of the art on clinical coding for hospitalized patients, the outpatient setting -- where doctors tend to non-hospitalised patients -- is overlooked. Although both settings can be formalised as a multi-label classification task, they present unique and distinct challenges, which raises the question of whether the success of inpatient clinical coding approaches translates to the outpatient setting. This paper is the first to investigate how well state-of-the-art deep learning-based clinical coding approaches work in the outpatient setting at hospital scale. To this end, we collect a large outpatient dataset comprising over 7 million notes documenting over half a million patients. We adapt four state-of-the-art clinical coding approaches to this setting and evaluate their potential to assist coders. We find evidence that clinical coding in outpatient settings can benefit from more innovations in popular inpatient coding benchmarks. A deeper analysis of the factors contributing to the success -- amount and form of data and choice of document representation -- reveals the presence of easy-to-solve examples, the coding of which can be completely automated with a low error rate., Comment: 9 pages, preprint under review

We present a new loss function that addresses the out-of-distribution (OOD) calibration problem. While many objective functions have been proposed to effectively calibrate models in-distribution, our findings show that they do not always fare well OOD. Based on the Principle of Maximum Entropy, we incorporate helpful statistical constraints observed during training, delivering better model calibration without sacrificing accuracy. We provide theoretical analysis and show empirically that our method works well in practice, achieving state-of-the-art calibration on both synthetic and real-world benchmarks., Comment: 38th AAAI Conference on Artificial Intelligence, AAAI24 (Oral)

This paper describes PULSAR, our system submission at the ImageClef 2023 MediQA-Sum task on summarising patient-doctor dialogues into clinical records. The proposed framework relies on domain-specific pre-training, to produce a specialised language model which is trained on task-specific natural data augmented by synthetic data generated by a black-box LLM. We find limited evidence towards the efficacy of domain-specific pre-training and data augmentation, while scaling up the language model yields the best performance gains. Our approach was ranked second and third among 13 submissions on task B of the challenge. Our code is available at https://github.com/yuping-wu/PULSAR., Comment: 8 pages. ImageClef 2023 MediQA-Sum

Medical progress notes play a crucial role in documenting a patient's hospital journey, including his or her condition, treatment plan, and any updates for healthcare providers. Automatic summarisation of a patient's problems in the form of a problem list can aid stakeholders in understanding a patient's condition, reducing workload and cognitive bias. BioNLP 2023 Shared Task 1A focuses on generating a list of diagnoses and problems from the provider's progress notes during hospitalisation. In this paper, we introduce our proposed approach to this task, which integrates two complementary components. One component employs large language models (LLMs) for data augmentation; the other is an abstractive summarisation LLM with a novel pre-training objective for generating the patients' problems summarised as a list. Our approach was ranked second among all submissions to the shared task. The performance of our model on the development and test datasets shows that our approach is more robust on unknown data, with an improvement of up to 3.1 points over the same size of the larger model., Comment: Accepted by ACL 2023's workshop BioNLP 2023

Clinical notes in healthcare facilities are tagged with the International Classification of Diseases (ICD) code; a list of classification codes for medical diagnoses and procedures. ICD coding is a challenging multilabel text classification problem due to noisy clinical document inputs and long-tailed label distribution. Recent automated ICD coding efforts improve performance by encoding medical notes and codes with additional data and knowledge bases. However, most of them do not reflect how human coders generate the code: first, the coders select general code categories and then look for specific subcategories that are relevant to a patient's condition. Inspired by this, we propose a two-stage decoding mechanism to predict ICD codes. Our model uses the hierarchical properties of the codes to split the prediction into two steps: At first, we predict the parent code and then predict the child code based on the previous prediction. Experiments on the public MIMIC-III data set show that our model performs well in single-model settings without external data or knowledge., Comment: Accepted to ACL'23

Sentence representations are a critical component in NLP applications such as retrieval, question answering, and text classification. They capture the meaning of a sentence, enabling machines to understand and reason over human language. In recent years, significant progress has been made in developing methods for learning sentence representations, including unsupervised, supervised, and transfer learning approaches. However there is no literature review on sentence representations till now. In this paper, we provide an overview of the different methods for sentence representation learning, focusing mostly on deep learning models. We provide a systematic organization of the literature, highlighting the key contributions and challenges in this area. Overall, our review highlights the importance of this area in natural language processing, the progress made in sentence representation learning, and the challenges that remain. We conclude with directions for future research, suggesting potential avenues for improving the quality and efficiency of sentence representations., Comment: Accepted to EACL'24

Clinical notes are assigned ICD codes - sets of codes for diagnoses and procedures. In the recent years, predictive machine learning models have been built for automatic ICD coding. However, there is a lack of widely accepted benchmarks for automated ICD coding models based on large-scale public EHR data. This paper proposes a public benchmark suite for ICD-10 coding using a large EHR dataset derived from MIMIC-IV, the most recent public EHR dataset. We implement and compare several popular methods for ICD coding prediction tasks to standardize data preprocessing and establish a comprehensive ICD coding benchmark dataset. This approach fosters reproducibility and model comparison, accelerating progress toward employing automated ICD coding in future studies. Furthermore, we create a new ICD-9 benchmark using MIMIC-IV data, providing more data points and a higher number of ICD codes than MIMIC-III. Our open-source code offers easy access to data processing steps, benchmark creation, and experiment replication for those with MIMIC-IV access, providing insights, guidance, and protocols to efficiently develop ICD coding models., Comment: Benchmark, Multilabel, Classification

Facial expression analysis has long been an active research area of computer vision. Traditional methods mainly analyse images for prototypical discrete emotions; as a result, they do not provide an accurate depiction of the complex emotional states in humans. Furthermore, illumination variance remains a challenge for face analysis in the visible light spectrum. To address these issues, we propose using a dimensional model based on valence and arousal to represent a wider range of emotions, in combination with near infra-red (NIR) imagery, which is more robust to illumination changes. Since there are no existing NIR facial expression datasets with valence-arousal labels available, we present two complementary data augmentation methods (face morphing and CycleGAN approach) to create NIR image datasets with dimensional emotion labels from existing categorical and/or visible-light datasets. Our experiments show that these generated NIR datasets are comparable to existing datasets in terms of data quality and baseline prediction performance.

The development of powerful deep learning technologies has brought about some negative effects to both society and individuals. One such issue is the emergence of fake media. To tackle the issue, we have organized the Trusted Media Challenge (TMC) to explore how Artificial Intelligence (AI) technologies could be leveraged to combat fake media. To enable further research, we are releasing the dataset that we had prepared from the TMC challenge, consisting of 4,380 fake and 2,563 real videos, with various video and/or audio manipulation methods employed to produce different types of fake media. All the videos in the TMC dataset are accompanied with audios and have a minimum resolution of 360p. The videos have various durations, background, illumination, and may contain perturbations that mimic transmission errors and compression. We have also carried out a user study to demonstrate the quality of the TMC dataset and to compare the performance of humans and AI models. The results showed that the TMC dataset can fool human participants in many cases, and the winning AI models of the Trusted Media Challenge outperformed humans. The TMC dataset is available for research purpose upon request via tmc-dataset@aisingapore.org.

Ground-based whole sky imagers (WSIs) are being used by researchers in various fields to study the atmospheric events. These ground-based sky cameras capture visible-light images of the sky at regular intervals of time. Owing to the atmospheric interference and camera sensor noise, the captured images often exhibit noise and blur. This may pose a problem in subsequent image processing stages. Therefore, it is important to accurately identify the blurred images. This is a difficult task, as clouds have varying shapes, textures, and soft edges whereas the sky acts as a homogeneous and uniform background. In this paper, we propose an efficient framework that can identify the blurred sky/cloud images. Using a static external marker, our proposed methodology has a detection accuracy of 94\%. To the best of our knowledge, our approach is the first of its kind in the automatic identification of blurred images for ground-based sky/cloud images., Comment: Accepted in Proc. IEEE AP-S Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting, 2021

Despite their continued popularity, categorical approaches to affect recognition have limitations, especially in real-life situations. Dimensional models of affect offer important advantages for the recognition of subtle expressions and more fine-grained analysis. We introduce a simple but effective facial expression analysis (FEA) system for dimensional affect, solely based on geometric features and Partial Least Squares (PLS) regression. The system jointly learns to estimate Arousal and Valence ratings from a set of facial images. The proposed approach is robust, efficient, and exhibits comparable performance to contemporary deep learning models, while requiring a fraction of the computational resources.

Emotion recognition and understanding is a vital component in human-machine interaction. Dimensional models of affect such as those using valence and arousal have advantages over traditional categorical ones due to the complexity of emotional states in humans. However, dimensional emotion annotations are difficult and expensive to collect, therefore they are not as prevalent in the affective computing community. To address these issues, we propose a method to generate synthetic images from existing categorical emotion datasets using face morphing as well as dimensional labels in the circumplex space with full control over the resulting sample distribution, while achieving augmentation factors of at least 20x or more., Comment: in Proc IEEE International Conference on Image Processing (ICIP), Anchorage, Sep.2021

Long-Short-Term-Memory (LSTM) networks have been used extensively for time series forecasting in recent years due to their ability of learning patterns over different periods of time. In this paper, this ability is applied to learning the pattern of Global Positioning System (GPS)-based Precipitable Water Vapor (PWV) measurements over a period of 4 hours. The trained model was evaluated on more than 1500 hours of recorded data. It achieves a root mean square error (RMSE) of 0.098 mm for a forecasting interval of 5 minutes in the future, and outperforms the naive approach for a lead-time of up to 40 minutes., Comment: Published in Proc. IEEE AP-S Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting, 2020

We examine two key questions in GAN training, namely overfitting and mode drop, from an empirical perspective. We show that when stochasticity is removed from the training procedure, GANs can overfit and exhibit almost no mode drop. Our results shed light on important characteristics of the GAN training procedure. They also provide evidence against prevailing intuitions that GANs do not memorize the training set, and that mode dropping is mainly due to properties of the GAN objective rather than how it is optimized during training., Comment: To appear in ICIP2020

Image quality assessment is critical to control and maintain the perceived quality of visual content. Both subjective and objective evaluations can be utilised, however, subjective image quality assessment is currently considered the most reliable approach. Databases containing distorted images and mean opinion scores are needed in the field of atmospheric research with a view to improve the current state-of-the-art methodologies. In this paper, we focus on using ground-based sky camera images to understand the atmospheric events. We present a new image quality assessment dataset containing original and distorted nighttime images of sky/cloud from SWINSEG database. Subjective quality assessment was carried out in controlled conditions, as recommended by the ITU. Statistical analyses of the subjective scores showed the impact of noise type and distortion level on the perceived quality., Comment: Published in Proc. Progress In Electromagnetics Research Symposium (PIERS), 2019

We analyse the time series of solar irradiance measurements using chaos theory. The False Nearest Neighbour method (FNN), one of the most common methods of chaotic analysis is used for the analysis. One year data from the weather station located at Nanyang Technological University (NTU) Singapore with a temporal resolution of $1$ minute is employed for the study. The data is sampled at $60$ minutes interval and $30$ minutes interval for the analysis using the FNN method. Our experiments revealed that the optimum dimension required for solar irradiance is $4$ for both samplings. This indicates that a minimum of $4$ dimensions is required for embedding the data for the best representation of input. This study on obtaining the embedding dimension of solar irradiance measurement will greatly assist in fixing the number of previous data required for solar irradiance forecasting., Comment: Published in Proc. Progress In Electromagnetics Research Symposium (PIERS), 2019

Solar irradiance is the primary input for all solar energy generation systems. The amount of available solar radiation over time under the local weather conditions helps to decide the optimal location, technology and size of a solar energy project. We study the behaviour of incident solar irradiance on the earth's surface using weather sensors. In this paper, we propose a time-series based technique to forecast the solar irradiance values for shorter lead times of upto 15 minutes. Our experiments are conducted in the tropical region viz. Singapore, which receives a large amount of solar irradiance throughout the year. We benchmark our method with two common forecasting techniques, namely persistence model and average model, and we obtain good prediction performance. We report a root mean square of 147 W/m^2 for a lead time of 15 minutes., Comment: Published in Proc. Progress In Electromagnetics Research Symposium (PIERS), 2019

Ground-based whole sky cameras are extensively used for localized monitoring of clouds nowadays. They capture hemispherical images of the sky at regular intervals using a fisheye lens. In this paper, we propose a framework for estimating solar irradiance from pictures taken by those imagers. Unlike pyranometers, such sky images contain information about cloud coverage and can be used to derive cloud movement. An accurate estimation of solar irradiance using solely those images is thus a first step towards short-term forecasting of solar energy generation based on cloud movement. We derive and validate our model using pyranometers co-located with our whole sky imagers. We achieve a better performance in estimating solar irradiance and in particular its short-term variations as compared to other related methods using ground-based observations., Comment: Published in Atmospheric Measurement Techniques (AMT), 2019

In recent years, there has been growing interest in using Precipitable Water Vapor (PWV) derived from Global Positioning System (GPS) signal delays to predict rainfall. However, the occurrence of rainfall is dependent on a myriad of atmospheric parameters. This paper proposes a systematic approach to analyze various parameters that affect precipitation in the atmosphere. Different ground-based weather features like Temperature, Relative Humidity, Dew Point, Solar Radiation, PWV along with Seasonal and Diurnal variables are identified, and a detailed feature correlation study is presented. While all features play a significant role in rainfall classification, only a few of them, such as PWV, Solar Radiation, Seasonal and Diurnal features, stand out for rainfall prediction. Based on these findings, an optimum set of features are used in a data-driven machine learning algorithm for rainfall prediction. The experimental evaluation using a four-year (2012-2015) database shows a true detection rate of 80.4%, a false alarm rate of 20.3%, and an overall accuracy of 79.6%. Compared to the existing literature, our method significantly reduces the false alarm rates., Comment: Published in IEEE Transactions on Geoscience and Remote Sensing, 2019

We analyze clouds in the earth's atmosphere using ground-based sky cameras. An accurate segmentation of clouds in the captured sky/cloud image is difficult, owing to the fuzzy boundaries of clouds. Several techniques have been proposed that use color as the discriminatory feature for cloud detection. In the existing literature, however, analysis of daytime and nighttime images is considered separately, mainly because of differences in image characteristics and applications. In this paper, we propose a light-weight deep-learning architecture called CloudSegNet. It is the first that integrates daytime and nighttime (also known as nychthemeron) image segmentation in a single framework, and achieves state-of-the-art results on public databases., Comment: Published in IEEE Geoscience and Remote Sensing Letters, 2019

Advertisements (ads) often contain strong affective content to capture viewer attention and convey an effective message to the audience. However, most computational affect recognition (AR) approaches examine ads via the text modality, and only limited work has been devoted to decoding ad emotions from audiovisual or user cues. This work (1) compiles an affective ad dataset capable of evoking coherent emotions across users; (2) explores the efficacy of content-centric convolutional neural network (CNN) features for AR vis-\~a-vis handcrafted audio-visual descriptors; (3) examines user-centric ad AR from Electroencephalogram (EEG) responses acquired during ad-viewing, and (4) demonstrates how better affect predictions facilitate effective computational advertising as determined by a study involving 18 users. Experiments reveal that (a) CNN features outperform audiovisual descriptors for content-centric AR; (b) EEG features are able to encode ad-induced emotions better than content-based features; (c) Multi-task learning performs best among a slew of classification algorithms to achieve optimal AR, and (d) Pursuant to (b), EEG features also enable optimized ad insertion onto streamed video, as compared to content-based or manual insertion techniques in terms of ad memorability and overall user experience., Comment: Under consideration for publication in IEEE Trans. Affective Computing. arXiv admin note: text overlap with arXiv:1709.01684

Different empirical models have been developed for cloud detection. There is a growing interest in using the ground-based sky/cloud images for this purpose. Several methods exist that perform binary segmentation of clouds. In this paper, we propose to use a deep learning architecture (U-Net) to perform multi-label sky/cloud image segmentation. The proposed approach outperforms recent literature by a large margin.

Global Positioning System (GPS) derived precipitable water vapor (PWV) is extensively being used in atmospheric remote sensing for applications like rainfall prediction. Many applications require PWV values with good resolution and without any missing values. In this paper, we implement an exponential smoothing method to accurately predict the missing PWV values. The method shows good performance in terms of capturing the seasonal variability of PWV values. We report a root mean square error of 0.1~mm for a lead time of 15 minutes, using past data of 30 hours measured at 5-minute intervals.

We propose a GAN design which models multiple distributions effectively and discovers their commonalities and particularities. Each data distribution is modeled with a mixture of $K$ generator distributions. As the generators are partially shared between the modeling of different true data distributions, shared ones captures the commonality of the distributions, while non-shared ones capture unique aspects of them. We show the effectiveness of our method on various datasets (MNIST, Fashion MNIST, CIFAR-10, Omniglot, CelebA) with compelling results.

Apparent personality and emotion analysis are both central to affective computing. Existing works solve them individually. In this paper we investigate if such high-level affect traits and their relationship can be jointly learned from face images in the wild. To this end, we introduce PersEmoN, an end-to-end trainable and deep Siamese-like network. It consists of two convolutional network branches, one for emotion and the other for apparent personality. Both networks share their bottom feature extraction module and are optimized within a multi-task learning framework. Emotion and personality networks are dedicated to their own annotated dataset. Furthermore, an adversarial-like loss function is employed to promote representation coherence among heterogeneous dataset sources. Based on this, we also explore the emotion-to-apparent-personality relationship. Extensive experiments demonstrate the effectiveness of PersEmoN., Comment: Accepted to IEEE Transactions on Affective Computing

We examine if EEG-based cognitive load (CL) estimation is generalizable across the character, spatial pattern, bar graph and pie chart-based visualizations for the nback~task. CL is estimated via two recent approaches: (a) Deep convolutional neural network, and (b) Proximal support vector machines. Experiments reveal that CL estimation suffers across visualizations motivating the need for effective machine learning techniques to benchmark visual interface usability for a given analytic task.

Data Visualization has been receiving growing attention recently, with ubiquitous smart devices designed to render information in a variety of ways. However, while evaluations of visual tools for their interpretability and intuitiveness have been commonplace, not much research has been devoted to other forms of data rendering, eg, sonification. This work is the first to automatically estimate the cognitive load induced by different acoustic parameters considered for sonification in prior studies. We examine cognitive load via (a) perceptual data-sound mapping accuracies of users for the different acoustic parameters, (b) cognitive workload impressions explicitly reported by users, and (c) their implicit EEG responses compiled during the mapping task. Our main findings are that (i) low cognitive load-inducing (ie, more intuitive) acoustic parameters correspond to higher mapping accuracies, (ii) EEG spectral power analysis reveals higher $\alpha$ band power for low cognitive load parameters, implying a congruent relationship between explicit and implicit user responses, and (iii) Cognitive load classification with EEG features achieves a peak F1-score of 0.64, confirming that reliable workload estimation is achievable with user EEG data compiled using wearable sensors., Comment: Accepted for publication in the proceedings of the 20th ACM International Conference on Multimodal Interaction, Colorado, USA

We examine two different techniques for parameter averaging in GAN training. Moving Average (MA) computes the time-average of parameters, whereas Exponential Moving Average (EMA) computes an exponentially discounted sum. Whilst MA is known to lead to convergence in bilinear settings, we provide the -- to our knowledge -- first theoretical arguments in support of EMA. We show that EMA converges to limit cycles around the equilibrium with vanishing amplitude as the discount parameter approaches one for simple bilinear games and also enhances the stability of general GAN training. We establish experimentally that both techniques are strikingly effective in the non-convex-concave GAN setting as well. Both improve inception and FID scores on different architectures and for different GAN objectives. We provide comprehensive experimental results across a range of datasets -- mixture of Gaussians, CIFAR-10, STL-10, CelebA and ImageNet -- to demonstrate its effectiveness. We achieve state-of-the-art results on CIFAR-10 and produce clean CelebA face images.\footnote{~The code is available at \url{https://github.com/yasinyazici/EMA_GAN}}, Comment: Published as a conference paper at ICLR 2019