Search

LLMs are predominantly trained on English data, which leads to a significant drop in performance on low-resource languages. Understanding how LLMs handle these languages is crucial for improving their effectiveness. This study focuses on Urdu as a use case for exploring the challenges faced by LLMs in processing low-resource languages. LLMs primarily reason in English when prompted in another language, with the final layers acting as translators to convert the English response into the target language. This study finds that even for low-resource languages, the internal latent response of LLMs in English is quite coherent; however, the translation features are lossy and result in poor translations, leading to reduced performance. By mechanistically removing these translation features and using a separate translation model to translate the internal latent response of LLM, the performance of LLMs improves significantly while also preserving the cultural nuances of the input in low-resource languages.

The majority of genetic variants identified in genome-wide association studies of complex traits are non-coding, and characterizing their function remains an important challenge in human genetics. Genomic deep learning models have emerged as a promising approach to enable in silico prediction of variant effects. These include supervised sequence-to-activity models, which predict genome-wide chromatin states or gene expression levels directly from DNA sequence, and self-supervised genomic language models. Here, we review progress in leveraging these models for non-coding variant effect prediction. We describe practical considerations for making such predictions and categorize the types of ground truth data that have been used to evaluate deep learning-based variant effect predictions, providing insight into the settings in which current models are most useful. We also discuss downstream applications of such models to understanding disease-relevant non-coding variants. Our review highlights key considerations for practitioners and opportunities for future improvements in model development and evaluation.

This paper investigates the evolving relationship between protecting Intellectual Property Rights (IPRs) and blockchain technology. We conducted a comprehensive literature review, supplemented by case study analyses and research paper reviews, to understand the scope and implications of blockchain about intellectual property rights. Our study demonstrates how applying blockchain technology for IPR could revolutionize transparency, security, and operational efficiency. It also identifies the primary challenges and openings in this area. We provide an extensive framework for integrating blockchain technology with intellectual property rights and other technical components (some of which already exist or are resolved by blockchain; some might need attention), drawing on current research and best practices. This framework has the potential to give a new perspective in a structured manner for the intellectual property landscape by providing 360-degree coverage across different layers of operation.

Deep learning is actively being used in biometrics to develop efficient identification and verification systems. Handwritten signatures are a common subset of biometric data for authentication purposes. Generative adversarial networks (GANs) learn from original and forged signatures to generate forged signatures. While most GAN techniques create a strong signature verifier, which is the discriminator, there is a need to focus more on the quality of forgeries generated by the generator model. This work focuses on creating a generator that produces forged samples that achieve a benchmark in spoofing signature verification systems. We use CycleGANs infused with Inception model-like blocks with attention heads as the generator and a variation of the SigCNN model as the base Discriminator. We train our model with a new technique that results in 80% to 100% success in signature spoofing. Additionally, we create a custom evaluation technique to act as a goodness measure of the generated forgeries. Our work advocates generator-focused GAN architectures for spoofing data quality that aid in a better understanding of biometric data generation and evaluation.

The internet is a common place for businesses to collect and store as much client data as possible and computer storage capacity has increased exponentially due to this trend. Businesses utilize this data to enhance customer satisfaction, generate revenue, boost sales, and increase profile. However, the emerging sector of data brokers is plagued with legal challenges. In part I, we will look at what a data broker is, how it collects information, the data industry, and some of the difficulties it encounters. In Part II, we will look at potential options for regulating data brokers. All options are provided in light of the EU General Data Protection Regulation (GDPR). In Part III, we shall present our analysis and findings.

Large language models (LLM) have achieved impressive performance on medical question-answering benchmarks. However, high benchmark accuracy does not imply that the performance generalizes to real-world clinical settings. Medical question-answering benchmarks rely on assumptions consistent with quantifying LLM performance but that may not hold in the open world of the clinic. Yet LLMs learn broad knowledge that can help the LLM generalize to practical conditions regardless of unrealistic assumptions in celebrated benchmarks. We seek to quantify how well LLM medical question-answering benchmark performance generalizes when benchmark assumptions are violated. Specifically, we present an adversarial method that we call MedFuzz (for medical fuzzing). MedFuzz attempts to modify benchmark questions in ways aimed at confounding the LLM. We demonstrate the approach by targeting strong assumptions about patient characteristics presented in the MedQA benchmark. Successful "attacks" modify a benchmark item in ways that would be unlikely to fool a medical expert but nonetheless "trick" the LLM into changing from a correct to an incorrect answer. Further, we present a permutation test technique that can ensure a successful attack is statistically significant. We show how to use performance on a "MedFuzzed" benchmark, as well as individual successful attacks. The methods show promise at providing insights into the ability of an LLM to operate robustly in more realistic settings., Comment: 9 pages, 3 figures, 2 algorithms, appendix

In the field of human-computer interaction (HCI), the usability assessment of m-learning (mobile-learning) applications is a real challenge. Such assessment typically involves extraction of the best features of an application like efficiency, effectiveness, learnability, cognition, memorability, etc., and further ranking of those features for an overall assessment of the quality of the mobile application. In the previous literature, it is found that there is neither any theory nor any tool available to measure or assess a user perception and assessment of usability features of a m-learning application for the sake of ranking the graphical user interface of a mobile application in terms of a user acceptance and satisfaction. In this paper, a novel approach is presented by performing a mobile applications quantitative and qualitative analysis. Based on user requirements and perception, a criterion is defined based on a set of important features. Afterward, for the qualitative analysis, a genetic algorithm (GA) is used to score prescribed features for the usability assessment of a mobile application. The used approach assigns a score to each usability feature according to the user requirement and weight of each feature. GA performs the rank assessment process initially by performing feature selection and scoring the best features of the application. A comparison of assessment analysis of GA and various machine learning models, K-nearest neighbours, Naive Bayes, and Random Forests is performed. It was found that a GA-based support vector machine (SVM) provides more accuracy in the extraction of the best features of a mobile application and further ranking of those features., Comment: 20 pages, 22 Figures, Journal Paper

Recent advances in AI combine large language models (LLMs) with vision encoders that bring forward unprecedented technical capabilities to leverage for a wide range of healthcare applications. Focusing on the domain of radiology, vision-language models (VLMs) achieve good performance results for tasks such as generating radiology findings based on a patient's medical image, or answering visual questions (e.g., 'Where are the nodules in this chest X-ray?'). However, the clinical utility of potential applications of these capabilities is currently underexplored. We engaged in an iterative, multidisciplinary design process to envision clinically relevant VLM interactions, and co-designed four VLM use concepts: Draft Report Generation, Augmented Report Review, Visual Search and Querying, and Patient Imaging History Highlights. We studied these concepts with 13 radiologists and clinicians who assessed the VLM concepts as valuable, yet articulated many design considerations. Reflecting on our findings, we discuss implications for integrating VLM capabilities in radiology, and for healthcare AI more generally., Comment: to appear at CHI 2024

Federated learning is a technique that allows multiple entities to collaboratively train models using their data without compromising data privacy. However, despite its advantages, federated learning can be susceptible to false data injection attacks. In these scenarios, a malicious entity with control over specific agents in the network can manipulate the learning process, leading to a suboptimal model. Consequently, addressing these data injection attacks presents a significant research challenge in federated learning systems. In this paper, we propose a novel technique to detect and mitigate data injection attacks on federated learning systems. Our mitigation method is a local scheme, performed during a single instance of training by the coordinating node, allowing the mitigation during the convergence of the algorithm. Whenever an agent is suspected to be an attacker, its data will be ignored for a certain period, this decision will often be re-evaluated. We prove that with probability 1, after a finite time, all attackers will be ignored while the probability of ignoring a trustful agent becomes 0, provided that there is a majority of truthful agents. Simulations show that when the coordinating node detects and isolates all the attackers, the model recovers and converges to the truthful model., Comment: This work will be presented at IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2024

Software startup companies develop innovative, software-intensive products within limited time frames and with few resources, searching for sustainable and scalable business models. Software startups are quite distinct from traditional mature software companies, but also from micro-, small-, and medium-sized enterprises, introducing new challenges relevant for software engineering research. This paper's research agenda focuses on software engineering in startups, identifying, in particular, 70+ research questions in the areas of supporting startup engineering activities, startup evolution models and patterns, ecosystems and innovation hubs, human aspects in software startups, applying startup concepts in non-startup environments, and methodologies and theories for startup research. We connect and motivate this research agenda with past studies in software startup research, while pointing out possible future directions. While all authors of this research agenda have their main background in Software Engineering or Computer Science, their interest in software startups broadens the perspective to the challenges, but also to the opportunities that emerge from multi-disciplinary research. Our audience is therefore primarily software engineering researchers, even though we aim at stimulating collaborations and research that crosses disciplinary boundaries. We believe that with this research agenda we cover a wide spectrum of the software startup industry current needs.

Recent works have shown that imposing tensor structures on the coefficient tensor in regression problems can lead to more reliable parameter estimation and lower sample complexity compared to vector-based methods. This work investigates a new low-rank tensor model, called Low Separation Rank (LSR), in Generalized Linear Model (GLM) problems. The LSR model -- which generalizes the well-known Tucker and CANDECOMP/PARAFAC (CP) models, and is a special case of the Block Tensor Decomposition (BTD) model -- is imposed onto the coefficient tensor in the GLM model. This work proposes a block coordinate descent algorithm for parameter estimation in LSR-structured tensor GLMs. Most importantly, it derives a minimax lower bound on the error threshold on estimating the coefficient tensor in LSR tensor GLM problems. The minimax bound is proportional to the intrinsic degrees of freedom in the LSR tensor GLM problem, suggesting that its sample complexity may be significantly lower than that of vectorized GLMs. This result can also be specialised to lower bound the estimation error in CP and Tucker-structured GLMs. The derived bounds are comparable to tight bounds in the literature for Tucker linear regression, and the tightness of the minimax lower bound is further assessed numerically. Finally, numerical experiments on synthetic datasets demonstrate the efficacy of the proposed LSR tensor model for three regression types (linear, logistic and Poisson). Experiments on a collection of medical imaging datasets demonstrate the usefulness of the LSR model over other tensor models (Tucker and CP) on real, imbalanced data with limited available samples., Comment: 43 pages; published in Transactions on Machine Learning Research (08/2023)

This paper considers the problem of understanding the behavior of a general class of accelerated gradient methods on smooth nonconvex functions. Motivated by some recent works that have proposed effective algorithms, based on Polyak's heavy ball method and the Nesterov accelerated gradient method, to achieve convergence to a local minimum of nonconvex functions, this work proposes a broad class of Nesterov-type accelerated methods and puts forth a rigorous study of these methods encompassing the escape from saddle-points and convergence to local minima through a both asymptotic and a non-asymptotic analysis. In the asymptotic regime, this paper answers an open question of whether Nesterov's accelerated gradient method (NAG) with variable momentum parameter avoids strict saddle points almost surely. This work also develops two metrics of asymptotic rate of convergence and divergence, and evaluates these two metrics for several popular standard accelerated methods such as the NAG, and Nesterov's accelerated gradient with constant momentum (NCM) near strict saddle points. In the local regime, this work provides an analysis that leads to the "linear" exit time estimates from strict saddle neighborhoods for trajectories of these accelerated methods as well the necessary conditions for the existence of such trajectories. Finally, this work studies a sub-class of accelerated methods that can converge in convex neighborhoods of nonconvex functions with a near optimal rate to a local minima and at the same time this sub-class offers superior saddle-escape behavior compared to that of NAG., Comment: 107 pages, 10 figures; pre-print of a journal submission

A hyperproperty relates executions of a program and is used to formalize security objectives such as confidentiality, non-interference, privacy, and anonymity. Formally, a hyperproperty is a collection of allowable sets of executions. A program violates a hyperproperty if the set of its executions is not in the collection specified by the hyperproperty. The logic HyperCTL^* has been proposed in the literature to formally specify and verify hyperproperties. The problem of checking whether a finite-state program satisfies a HyperCTL^* formula is known to be decidable. However, the problem turns out to be undecidable for procedural (recursive) programs. Surprisingly, we show that decidability can be restored if we consider restricted classes of hyperproperties, namely those that relate only those executions of a program which have the same call-stack access pattern. We call such hyperproperties, \emph{stack-aware hyperproperties.} Our decision procedure can be used as a proof method for establishing security objectives such as noninference for recursive programs, and also for refuting security objectives such as observational determinism. Further, if the call stack size is observable to the attacker, the decision procedure provides exact verification.