Sean Moran

Maxime Labonne, Sean Moran

This paper investigates the effectiveness of large language models (LLMs) in email spam detection by comparing prominent models from three distinct families: BERT-like, Sentence Transformers, and Seq2Seq. Additionally, we examine well-established machine learning techniques for spam detection, such as Naïve Bayes and LightGBM, as baseline methods. We assess the performance of these models across four public datasets, utilizing different numbers of training samples (full training set and few-shot settings). Our findings reveal that, in the majority of cases, LLMs surpass the performance of the popular baseline techniques, particularly in few-shot scenarios. This adaptability renders LLMs uniquely suited to spam detection tasks, where labeled samples are limited in number and models require frequent updates. Additionally, we introduce Spam-T5, a Flan-T5 model that has been specifically adapted and fine-tuned for the purpose of detecting email spam. Our results demonstrate that Spam-T5 surpasses baseline models and other LLMs in the majority of scenarios, particularly when there are a limited number of training samples available. Our code is publicly available at https://github.com/jpmorganchase/emailspamdetection.

Agathe Lherondelle, Varun Babbar, Yash Satsangi et al.

This paper presents Topical, a novel deep neural network for repository level embeddings. Existing methods, reliant on natural language documentation or naive aggregation techniques, are outperformed by Topical's utilization of an attention mechanism. This mechanism generates repository-level representations from source code, full dependency graphs, and script level textual data. Trained on publicly accessible GitHub repositories, Topical surpasses multiple baselines in tasks such as repository auto-tagging, highlighting the attention mechanism's efficacy over traditional aggregation methods. Topical also demonstrates scalability and efficiency, making it a valuable contribution to repository-level representation computation. For further research, the accompanying tools, code, and training dataset are provided at: https://github.com/jpmorganchase/topical.

Sae Young Moon, Gregor Kerr, Fran Silavong et al.

When designing a new API for a large project, developers need to make smart design choices so that their code base can grow sustainably. To ensure that new API components are well designed, developers can learn from existing API components. However, the lack of standardized methods for comparing API designs makes this learning process time-consuming and difficult. To address this gap we developed API-Miner, to the best of our knowledge, one of the first API-to-API specification recommendation engines. API-Miner retrieves relevant specification components written in OpenAPI (a widely adopted language used to describe web APIs). API-miner presents several significant contributions, including: (1) novel methods of processing and extracting key information from OpenAPI specifications, (2) innovative feature extraction techniques that are optimized for the highly technical API specification domain, and (3) a novel log-linear probabilistic model that combines multiple signals to retrieve relevant and high quality OpenAPI specification components given a query specification. We evaluate API-Miner in both quantitative and qualitative tasks and achieve an overall of 91.7% recall@1 and 56.2% F1, which surpasses baseline performance by 15.4% in recall@1 and 3.2% in F1. Overall, API-Miner will allow developers to retrieve relevant OpenAPI specification components from a public or internal database in the early stages of the API development cycle, so that they can learn from existing established examples and potentially identify redundancies in their work. It provides the guidance developers need to accelerate development process and contribute thoughtfully designed APIs that promote code maintainability and quality. Code is available on GitHub at https://github.com/jpmorganchase/api-miner.

Lili Tao, Alexandru-Petre Cazan, Senad Ibraimoski et al.

The use of packaged libraries can significantly shorten the software development cycle by improving the quality and readability of code. In this paper, we present a recommendation engine called Librarian for open source libraries. A candidate library package is recommended for a given context if: 1) it has been frequently used with the imported libraries in the program; 2) it has similar functionality to the imported libraries in the program; 3) it has similar functionality to the developer's implementation, and 4) it can be used efficiently in the context of the provided code. We apply the state-of-the-art CodeBERT-based model for analysing the context of the source code to deliver relevant library recommendations to users.

Antonios Georgiadis, Varun Babbar, Fran Silavong et al.

Chest Computational Tomography (CT) scans present low cost, speed and objectivity for COVID-19 diagnosis and deep learning methods have shown great promise in assisting the analysis and interpretation of these images. Most hospitals or countries can train their own models using in-house data, however empirical evidence shows that those models perform poorly when tested on new unseen cases, surfacing the need for coordinated global collaboration. Due to privacy regulations, medical data sharing between hospitals and nations is extremely difficult. We propose a GAN-augmented federated learning model, dubbed ST-FL (Style Transfer Federated Learning), for COVID-19 image segmentation. Federated learning (FL) permits a centralised model to be learned in a secure manner from heterogeneous datasets located in disparate private data silos. We demonstrate that the widely varying data quality on FL client nodes leads to a sub-optimal centralised FL model for COVID-19 chest CT image segmentation. ST-FL is a novel FL framework that is robust in the face of highly variable data quality at client nodes. The robustness is achieved by a denoising CycleGAN model at each client of the federation that maps arbitrary quality images into the same target quality, counteracting the severe data variability evident in real-world FL use-cases. Each client is provided with the target style, which is the same for all clients, and trains their own denoiser. Our qualitative and quantitative results suggest that this FL model performs comparably to, and in some cases better than, a model that has centralised access to all the training data.

Jiaeli Shi, Najah Ghalyan, Kostis Gourgoulias et al.

Machine learning models trained on sensitive or private data can inadvertently memorize and leak that information. Machine unlearning seeks to retroactively remove such details from model weights to protect privacy. We contribute a lightweight unlearning algorithm that leverages the Fisher Information Matrix (FIM) for selective forgetting. Prior work in this area requires full retraining or large matrix inversions, which are computationally expensive. Our key insight is that the diagonal elements of the FIM, which measure the sensitivity of log-likelihood to changes in weights, contain sufficient information for effective forgetting. Specifically, we compute the FIM diagonal over two subsets -- the data to retain and forget -- for all trainable weights. This diagonal representation approximates the complete FIM while dramatically reducing computation. We then use it to selectively update weights to maximize forgetting of the sensitive subset while minimizing impact on the retained subset. Experiments show that our algorithm can successfully forget any randomly selected subsets of training data across neural network architectures. By leveraging the FIM diagonal, our approach provides an interpretable, lightweight, and efficient solution for machine unlearning with practical privacy benefits.

Georgios Papadopoulos, Fran Silavong, Sean Moran

Finding relevant and high-quality datasets to train machine learning models is a major bottleneck for practitioners. Furthermore, to address ambitious real-world use-cases there is usually the requirement that the data come labelled with high-quality annotations that can facilitate the training of a supervised model. Manually labelling data with high-quality labels is generally a time-consuming and challenging task and often this turns out to be the bottleneck in a machine learning project. Weak Supervised Learning (WSL) approaches have been developed to alleviate the annotation burden by offering an automatic way of assigning approximate labels (pseudo-labels) to unlabelled data based on heuristics, distant supervision and knowledge bases. We apply probabilistic generative latent variable models (PLVMs), trained on heuristic labelling representations of the original dataset, as an accurate, fast and cost-effective way to generate pseudo-labels. We show that the PLVMs achieve state-of-the-art performance across four datasets. For example, they achieve 22% points higher F1 score than Snorkel in the class-imbalanced Spouse dataset. PLVMs are plug-and-playable and are a drop-in replacement to existing WSL frameworks (e.g. Snorkel) or they can be used as benchmark models for more complicated algorithms, giving practitioners a compelling accuracy boost.

Farhad Bazyari, Xianghang Liu, Sean Moran

Osborne and Dredze (2014) reported that Twitter was the timeliest social-media source of breaking news, trailing only newswire. Twelve years on, the platform landscape has shifted - Google+ is gone, X replaced Twitter, Bluesky and Threads have appeared - and platform data now flows almost exclusively through commercial social-listening providers that redact key fields. We revisit the question with two sampling designs run through the same downstream pipeline. Sample A draws N = 50 events from the Wikipedia Current Events Portal (WCEP) ranked by article pageviews. Sample B draws N = 109 events from Polymarket prediction markets ranked by USD trading volume, with each event's news moment pinned to the largest 1-hour trade-volume spike. Both samples are pulled from one commercial provider across nine indexed channels. We report three findings. (1) The X-vs-news direction depends on the sample. News leads X by a median of 21.6 min on Sample A (n = 6 paired); the same comparison is tied at -0.02 min on Sample B (n = 16 paired, X earliest in 38%). (2) The channel ecosystem has diversified. Bluesky, Facebook public, and YouTube together account for 24-32% of earliest channel wins; the 2014 "X versus newswire" framing no longer fits. (3) Coverage gaps are structural. Even with U.S.-relevance filtering and a pageview prior, the provider's index returns no on-topic evidence on 24% of randomly-sampled WCEP events. The paper's contribution is the cross-surface design that exposes the sample dependency in (1).

Rundong Liu, Andre Frade, Amal Vaidya et al.

This paper introduces CodeQUEST, a novel framework leveraging Large Language Models (LLMs) to iteratively evaluate and enhance code quality across multiple dimensions, including readability, maintainability, efficiency, and security. The framework is divided into two main components: an Evaluator that assesses code quality across ten dimensions, providing both quantitative scores and qualitative summaries, and an Optimizer that iteratively improves the code based on the Evaluator's feedback. Our study demonstrates that CodeQUEST can effectively and robustly evaluate code quality, with its assessments aligning closely with established code quality metrics. Through a series of experiments using a curated dataset of Python and JavaScript examples, CodeQUEST demonstrated significant improvements in code quality, achieving a mean relative percentage improvement of 52.6%. The framework's evaluations were validated against a set of proxy metrics comprising of Pylint Score, Radon Maintainability Index, and Bandit output logs, showing a meaningful correlation. This highlights the potential of LLMs in automating code quality evaluation and improvement processes, presenting a significant advancement toward enhancing software development practices. The code implementation of the framework is available at: https://github.com/jpmorganchase/CodeQuest.

Xiaoying Zhi, Varun Babbar, Rundong Liu et al.

The subject of green AI has been gaining attention within the deep learning community given the recent trend of ever larger and more complex neural network models. Existing solutions for reducing the computational load of training at inference time usually involve pruning the network parameters. Pruning schemes often create extra overhead either by iterative training and fine-tuning for static pruning or repeated computation of a dynamic pruning graph. We propose a new parameter pruning strategy for learning a lighter-weight sub-network that minimizes the energy cost while maintaining comparable performance to the fully parameterised network on given downstream tasks. Our proposed pruning scheme is green-oriented, as it only requires a one-off training to discover the optimal static sub-networks by dynamic pruning methods. The pruning scheme consists of a binary gating module and a polarizing loss function to uncover sub-networks with user-defined sparsity. Our method enables pruning and training simultaneously, which saves energy in both the training and inference phases and avoids extra computational overhead from gating modules at inference time. Our results on CIFAR-10, CIFAR-100, and Tiny Imagenet suggest that our scheme can remove 50% of connections in deep networks with <1% reduction in classification accuracy. Compared to other related pruning methods, our method demonstrates a lower drop in accuracy for equivalent reductions in computational cost.

Sean Moran, Steven McDonagh, Gregory Slabaugh

We present a novel approach to adjust global image properties such as colour, saturation, and luminance using human-interpretable image enhancement curves, inspired by the Photoshop curves tool. Our method, dubbed neural CURve Layers (CURL), is designed as a multi-colour space neural retouching block trained jointly in three different colour spaces (HSV, CIELab, RGB) guided by a novel multi-colour space loss. The curves are fully differentiable and are trained end-to-end for different computer vision problems including photo enhancement (RGB-to-RGB) and as part of the image signal processing pipeline for image formation (RAW-to-RGB). To demonstrate the effectiveness of CURL we combine this global image transformation block with a pixel-level (local) image multi-scale encoder-decoder backbone network. In an extensive experimental evaluation we show that CURL produces state-of-the-art image quality versus recently proposed deep learning approaches in both objective and perceptual metrics, setting new state-of-the-art performance on multiple public datasets. Our code is publicly available at: https://github.com/sjmoran/CURL.

Chun-Fu Chen, Bill Moriarty, Shaohan Hu et al.

The recent rapid advancements in both sensing and machine learning technologies have given rise to the universal collection and utilization of people's biometrics, such as fingerprints, voices, retina/facial scans, or gait/motion/gestures data, enabling a wide range of applications including authentication, health monitoring, or much more sophisticated analytics. While providing better user experiences and deeper business insights, the use of biometrics has raised serious privacy concerns due to their intrinsic sensitive nature and the accompanying high risk of leaking sensitive information such as identity or medical conditions. In this paper, we propose a novel modality-agnostic data transformation framework that is capable of anonymizing biometric data by suppressing its sensitive attributes and retaining features relevant to downstream machine learning-based analyses that are of research and business values. We carried out a thorough experimental evaluation using publicly available facial, voice, and motion datasets. Results show that our proposed framework can achieve a \highlight{high suppression level for sensitive information}, while at the same time retain underlying data utility such that subsequent analyses on the anonymized biometric data could still be carried out to yield satisfactory accuracy.

Sean Moran

Approximate Nearest Neighbour (ANN) search is a fundamental problem in information retrieval, underpinning large-scale applications in computer vision, natural language processing, and cross-modal search. Hashing-based methods provide an efficient solution by mapping high-dimensional data into compact binary codes that enable fast similarity computations in Hamming space. Over the past two decades, a substantial body of work has explored learning to hash, where projection and quantisation functions are optimised from data rather than chosen at random. This article offers a foundational survey of early learning-based hashing methods, with an emphasis on the core ideas that shaped the field. We review supervised, unsupervised, and semi-supervised approaches, highlighting how projection functions are designed to generate meaningful embeddings and how quantisation strategies convert these embeddings into binary codes. We also examine extensions to multi-bit and multi-threshold models, as well as early advances in cross-modal retrieval. Rather than providing an exhaustive account of the most recent methods, our goal is to introduce the conceptual foundations of learning-based hashing for ANN search. By situating these early models in their historical context, we aim to equip readers with a structured understanding of the principles, trade-offs, and open challenges that continue to inform current research in this area.

Gregor Kerr, David Algorry, Senad Ibraimoski et al.

We introduce a new challenge to the software development community: 1) leveraging AI to accurately detect and flag up secrets in code and on popular document sharing platforms that frequently used by developers, such as Confluence and 2) automatically remediating the detections (e.g. by suggesting password vault functionality). This is a challenging, and mostly unaddressed task. Existing methods leverage heuristics and regular expressions, that can be very noisy, and therefore increase toil on developers. The next step - modifying code itself - to automatically remediate a detection, is a complex task. We introduce two baseline AI models that have good detection performance and propose an automatic mechanism for remediating secrets found in code, opening up the study of this task to the wider community.

Amal Vaidya, Mohan Krishna Vankayalapati, Jacky Chan et al.

We present a tool that leverages generative AI to accelerate the migration of on-premises applications to the cloud. The Cloud Migration LLM accepts input from the user specifying the parameters of their migration, and outputs a migration strategy with an architecture diagram. A user study suggests that the migration LLM can assist inexperienced users in finding the right cloud migration profile, while avoiding complexities of a manual approach.

Kostis Gourgoulias, Najah Ghalyan, Maxime Labonne et al.

This paper introduces an unsupervised method to estimate the class separability of text datasets from a topological point of view. Using persistent homology, we demonstrate how tracking the evolution of embedding manifolds during training can inform about class separability. More specifically, we show how this technique can be applied to detect when the training process stops improving the separability of the embeddings. Our results, validated across binary and multi-class text classification tasks, show that the proposed method's estimates of class separability align with those obtained from supervised methods. This approach offers a novel perspective on monitoring and improving the fine-tuning of sentence transformers for classification tasks, particularly in scenarios where labeled data is scarce. We also discuss how tracking these quantities can provide additional insights into the properties of the trained classifier.

Fran Silavong, Sean Moran, Antonios Georgiadis et al.

Machine learning on source code (MLOnCode) is a popular research field that has been driven by the availability of large-scale code repositories and the development of powerful probabilistic and deep learning models for mining source code. Code-to-code recommendation is a task in MLOnCode that aims to recommend relevant, diverse and concise code snippets that usefully extend the code currently being written by a developer in their development environment (IDE). Code-to-code recommendation engines hold the promise of increasing developer productivity by reducing context switching from the IDE and increasing code-reuse. Existing code-to-code recommendation engines do not scale gracefully to large codebases, exhibiting a linear growth in query time as the code repository increases in size. In addition, existing code-to-code recommendation engines fail to account for the global statistics of code repositories in the ranking function, such as the distribution of code snippet lengths, leading to sub-optimal retrieval results. We address both of these weaknesses with \emph{Senatus}, a new code-to-code recommendation engine. At the core of Senatus is \emph{De-Skew} LSH a new locality sensitive hashing (LSH) algorithm that indexes the data for fast (sub-linear time) retrieval while also counteracting the skewness in the snippet length distribution using novel abstract syntax tree-based feature scoring and selection algorithms. We evaluate Senatus and find the recommendations to be of higher quality than competing baselines, while achieving faster search. For example on the CodeSearchNet dataset Senatus improves performance by 31.21\% F1 and 147.9\emph{x} faster query time compared to Facebook Aroma. Senatus also outperforms standard MinHash LSH by 29.2\% F1 and 51.02\emph{x} faster query time.

Sean Moran, Pierre Marza, Steven McDonagh et al.

Digital artists often improve the aesthetic quality of digital photographs through manual retouching. Beyond global adjustments, professional image editing programs provide local adjustment tools operating on specific parts of an image. Options include parametric (graduated, radial filters) and unconstrained brush tools. These highly expressive tools enable a diverse set of local image enhancements. However, their use can be time consuming, and requires artistic capability. State-of-the-art automated image enhancement approaches typically focus on learning pixel-level or global enhancements. The former can be noisy and lack interpretability, while the latter can fail to capture fine-grained adjustments. In this paper, we introduce a novel approach to automatically enhance images using learned spatially local filters of three different types (Elliptical Filter, Graduated Filter, Polynomial Filter). We introduce a deep neural network, dubbed Deep Local Parametric Filters (DeepLPF), which regresses the parameters of these spatially localized filters that are then automatically applied to enhance the image. DeepLPF provides a natural form of model regularization and enables interpretable, intuitive adjustments that lead to visually pleasing results. We report on multiple benchmarks and show that DeepLPF produces state-of-the-art performance on two variants of the MIT-Adobe-5K dataset, often using a fraction of the parameters required for competing methods.

Hao Guan, Liu Liu, Sean Moran et al.

Denoising extreme low light images is a challenging task due to the high noise level. When the illumination is low, digital cameras increase the ISO (electronic gain) to amplify the brightness of captured data. However, this in turn amplifies the noise, arising from read, shot, and defective pixel sources. In the raw domain, read and shot noise are effectively modelled using Gaussian and Poisson distributions respectively, whereas defective pixels can be modeled with impulsive noise. In extreme low light imaging, noise removal becomes a critical challenge to produce a high quality, detailed image with low noise. In this paper, we propose a multi-task deep neural network called Noise Decomposition (NODE) that explicitly and separately estimates defective pixel noise, in conjunction with Gaussian and Poisson noise, to denoise an extreme low light image. Our network is purposely designed to work with raw data, for which the noise is more easily modeled before going through non-linear transformations in the image signal processing (ISP) pipeline. Quantitative and qualitative evaluation show the proposed method to be more effective at denoising real raw images than state-of-the-art techniques.