Abstract

Purpose : This study aims to develop an artificial intelligence (AI)-based system for accurately grading arrested retinal development using optical coherence tomography (OCT) across various manufacturers. We employ deep learning techniques, specifically Unsupervised Domain Adaptation (UDA), to create a device-agnostic classification model distinguishing normal and abnormal retinal development. Methods : Foveal scans from three OCT manufacturers (TM-OCT1, HH-OCT1, TM-OCT2, TM-OCT3) were collected and annotated from datasets exceeding 20,000 OCT scans. The dataset was divided into training (80%) and testing (20%) sets. We utilised Convolutional Neural Networks (CNN) with ResNet50 backbone, assessing each pair as the source (for supervised training) and target (for unsupervised training). The diagnostic accuracy of the AI models was compared to seven clinician graders with varying experience (1 to 10 years), evaluating sensitivity, specificity, and overall accuracy. Results : The cross-domain binary classification demonstrated exceptional diagnostic accuracy ranging from 87.75% to 96.06%. Sensitivity and specificity metrics further validated the robustness of our AI system (sensitivity: 88.68% to 98.38%, specificity: 78.00% to 98.02%). Notably, the model trained on TM-OCT1 and HH-OCT1 achieved 96.06% accuracy, 98.38% sensitivity, and 89.11% specificity on the HH-OCT1 test set, with an Area Under the Curve (AUC) of 99.4 (95% CI). These outcomes highlight the system's ability to distinguish normal and abnormal retinal development across diverse OCT devices. Conclusions : Our study demonstrates, for the first time, the feasibility of employing a device-agnostic AI system in paediatric OCT interpretation without additional labelled data. The AI system's diagnostic performance is comparable to a clinician with over 10 years of experience, showcasing its potential to reduce inter-examiner variability and enhance clinical care pathways. With integration of paediatric OCT into routine clinical assessment, our AI system serves as a robust foundation for the development of a real-time, frontline diagnostic tool for retinal developmental disorders.

Abstract

The focus of recent research has shifted from merely improving the metrics based performance of Deep Neural Networks (DNNs) to DNNs which are more interpretable to humans. The field of eXplainable Artificial Intelligence (XAI) has observed various techniques, including saliency-based and concept-based approaches. These approaches explain the model's decisions in simple human understandable terms called Concepts. Concepts are known to be the thinking ground of humans}. Explanations in terms of concepts enable detecting spurious correlations, inherent biases, or clever-hans. With the advent of concept-based explanations, a range of concept representation methods and automatic concept discovery algorithms have been introduced. Some recent works also use concepts for model improvement in terms of interpretability and generalization. We provide a systematic review and taxonomy of various concept representations and their discovery algorithms in DNNs, specifically in vision. We also provide details on concept-based model improvement literature marking the first comprehensive survey of these methods.

Abstract

Autism Spectrum Disorder (ASD) presents challenges in communication, social interaction, and sensory processing. This pilot study introduces a GOAL driven Recommendation Framework for personalized music interventions, leveraging music's therapeutic functions to address emotional, cognitive, and social needs in individuals with ASD was tested on an Indian population. Through participants music playlist compilation and data collected from online sources, a diverse dataset is curated. The recommendation system employs similarity metric to recommend music aligned with specific therapeutic goals. Evaluation metrics and behavioural changes indicate significant improvements in mood management, sleep regularization, focus, and sensory regulation. However, enhancements in communication and social skills are less conclusive. This study underscores the feasibility and effectiveness of personalized music recommendations in addressing therapeutic goals for individuals with ASD, emphasizing the need for further research to explore longitudinal outcomes to assess its impact on the quality of life of individuals with ASD.

Abstract

We introduce StethoSpeech, a silent speech interface that transforms flesh-conducted vibrations behind the ear into speech. This innovation is designed to improve social interactions for those with voice disorders, and furthermore enable discreet public communication. Unlike prior efforts, StethoSpeech does not require (a) paired-speech data for recorded vibrations and (b) a specialized device for recording vibrations, as it can work with an off-the-shelf clinical stethoscope. The novelty of our framework lies in the overall design, simulation of the ground-truth speech, and a sequence-to-sequence translation network, which works in the latent space. We present comprehensive experiments on the existing CSTR NAM TIMIT Plus corpus and our proposed StethoText: a large-scale synchronized database of non-audible murmur and text for speech research. Our results show that StethoSpeech provides natural-sounding and intelligible speech, significantly outperforming existing methods on several quantitative and qualitative metrics. Additionally, we showcase its capacity to extend its application to speakers not encountered during training and its effectiveness in challenging, noisy environments. Speech samples are available at https://stethospeech.github.io/StethoSpeech/.

Abstract

We present ParrotTTS, a modularized text-tospeech synthesis model leveraging disentangled self supervised speech representations. It can train a multi-speaker variant effectively using transcripts from a single speaker. ParrotTTS adapts to a new language in low resource setup and generalizes to languages not seen while training the self-supervised backbone. Moreover, without training on bilingual or parallel examples, ParrotTTS can transfer voices across languages while preserving the speaker-specific characteristics, e.g., synthesizing fluent Hindi speech using a French speaker’s voice and accent. We present extensive results in monolingual and multi-lingual scenarios. ParrotTTS outperforms state-of-theart multi-lingual text-to-speech (TTS) models using only a fraction of paired data as latter. Speech samples from ParrotTTS and code can be found at https://parrot-tts. github.io/tts/

Abstract

Lyrics play a crucial role in affecting and reinforcing emotional states by providing meaning and emotional connotations that interact with the acoustic properties of the music. Specific lyrical themes and emotions may intensify existing negative states in listeners and may lead to undesirable outcomes, especially in listeners with mood disorders such as depression. Hence, it is important for such individuals to be mindful of their listening strategies. In this study, we examine online music consumption of individuals at risk of depression in light of lyrical themes and emotions. Lyrics obtained from the listening histories of 541 Last.fm users, divided into At-Risk and No-Risk based on their mental well-being scores, were analyzed using natural language processing techniques. Statistical analyses of the results revealed that individuals at risk for depression prefer songs with lyrics associated with low valence and low arousal. Additionally, lyrics associated with themes of denial, self-reference, and ambivalence were preferred. In contrast, themes such as liberation, familiarity, and activity are not as favored. This study opens up the possibility of an approach to assessing depression risk from the digital footprint of individuals and potentially developing personalized recommendation systems.

Abstract

Continual learning has primarily focused on the issue of catastrophic forgetting and the associated stability-plasticity tradeoffs. However, little attention has been paid to the efficacy of continually learned representations, as representations are learned alongside classifiers throughout the learning process. Our primary contribution is empirically demonstrating that existing online continually trained deep networks produce inferior representations compared to a simple pre-defined random transforms. Our approach embeds raw pixels using a fixed random transform, approximating an RBF-Kernel initialized before any data is seen. We then train a simple linear classifier on top without storing any exemplars, processing one sample at a time in an online continual learning setting. This method, called RanDumb, significantly outperforms state-of-the-art continually learned representations across all standard online continual learning benchmarks. Our study reveals the significant limitations of representation learning, particularly in low-exemplar and online continual learning scenarios. Extending our investigation to popular exemplar-free scenarios with pretrained models, we find that training only a linear classifier on top of pretrained representations surpasses most continual fine-tuning and prompt-tuning strategies. Overall, our investigation challenges the prevailing assumptions about effective representation learning in online continual learning.

Abstract

To implement a blockchain, we need a blockchain protocol for all the nodes to follow. To design a blockchain protocol, we need a block publisher selection mechanism and a chain selection rule. In Proof-of-Stake (PoS) based blockchain protocols, block publisher selection mechanism selects the node to publish the next block based on the relative stake held by the node. However, PoS protocols may face vulnerability to fully adaptive corruptions. In literature, researchers address this issue at the cost of performance. In this paper, we propose a novel PoS-based blockchain protocol, QuickSync, to achieve security against fully adaptive corruptions without compromising on performance. We propose a metric called block power, a value defined for each block, derived from the output of the verifiable random function based on the digital signature of the block publisher. With this metric, we compute chain power, the sum of block powers of all the blocks comprising the chain, for all the valid chains. These metrics are a function of the block publisher's stake to enable the PoS aspect of the protocol. The chain selection rule selects the chain with the highest chain power as the one to extend. This chain selection rule hence determines the selected block publisher of the previous block. When we use metrics to define the chain selection rule, it may lead to vulnerabilities against Sybil attacks. QuickSync uses a Sybil attack resistant function implemented using histogram matching. We prove that QuickSync satisfies common prefix, chain growth, and chain quality properties and hence it is secure. We also show that it is resilient to different types of adversarial …

Abstract

One of the widely used peak reduction methods in smart grids is demand response, where one analyzes the shift in customers’ (agents’) usage patterns in response to the signal from the dis- tribution company. Often, these signals are in the form of incentives offered to agents. This work studies the effect of incentives on the prob- abilities of accepting such offers in a real-world smart grid simulator, PowerTAC. We first show that there exists a function that depicts the prob- ability of an agent reducing its load as a func- tion of the discounts offered to them. We call it reduction probability (RP). RP function is fur- ther parametrized by the rate of reduction (RR), which can differ for each agent. We provide an optimal algorithm, MJS–EXPRESPONSE, that outputs the discounts to each agent by maximiz- ing the expected reduction under a budget con- straint. When RRs are unknown, we propose a Multi-Armed Bandit (MAB) based online algo- rithm, namely MJSUCB–EXPRESPONSE, to learn RRs. Experimentally we show that it exhibits sub- linear regret. Finally, we showcase the efficacy of the proposed algorithm in mitigating demand peaks in a real-world smart grid system using the Power- TAC simulator as a test bed.

Abstract

The introduction of emojis (or emoticons) in social media platforms has given the users an increased potential for expression. We propose a novel method called Classification of Emojis using Siamese Network Architecture (CESNA) to learn emoji-based representations of resource-poor languages by jointly training them with resource-rich languages using a siamese network. CESNA model consists of twin Bi-directional Long Short-Term Memory Recurrent Neural Networks (Bi-LSTM RNN) with shared parameters joined by a contrastive loss function based on a similarity metric. The model learns the representations of resource-poor and resource-rich language in a common emoji space by using a similarity metric based on the emojis present in sentences from both languages. The model, hence, projects sentences with similar emojis closer to each other and the sentences with different emojis farther from one another. Experiments on large-scale Twitter datasets of resource-rich languages-English and Spanish and resource-poor languages-Hindi and Telugu reveal that CESNA outperforms the state-of-the-art emoji prediction approaches based on distributional semantics, semantic rules, lexicon lists and deep neural network representations without shared parameters.