Search

In this work we search for best practices in pre-processing of Electrocardiogram (ECG) signals in order to train better classifiers for the diagnosis of heart conditions. State of the art machine learning algorithms have achieved remarkable results in classification of some heart conditions using ECG data, yet there appears to be no consensus on pre-processing best practices. Is this lack of consensus due to different conditions and architectures requiring different processing steps for optimal performance? Is it possible that state of the art deep-learning models have rendered pre-processing unnecessary? In this work we apply down-sampling, normalization, and filtering functions to 3 different multi-label ECG datasets and measure their effects on 3 different high-performing time-series classifiers. We find that sampling rates as low as 50Hz can yield comparable results to the commonly used 500Hz. This is significant as smaller sampling rates will result in smaller datasets and models, which require less time and resources to train. Additionally, despite their common usage, we found min-max normalization to be slightly detrimental overall, and band-passing to make no measurable difference. We found the blind approach to pre-processing of ECGs for multi-label classification to be ineffective, with the exception of sample rate reduction which reliably reduces computational resources, but does not increase accuracy.

Thorough analysis of local droplet-level interactions is crucial to better understand the microphysical processes in clouds and their effect on the global climate. High-accuracy simulations of relevant droplet size distributions from Large Eddy Simulations (LES) of bin microphysics challenge current analysis techniques due to their high dimensionality involving three spatial dimensions, time, and a continuous range of droplet sizes. Utilizing the compact latent representations from Variational Autoencoders (VAEs), we produce novel and intuitive visualizations for the organization of droplet sizes and their evolution over time beyond what is possible with clustering techniques. This greatly improves interpretation and allows us to examine aerosol-cloud interactions by contrasting simulations with different aerosol concentrations. We find that the evolution of the droplet spectrum is similar across aerosol levels but occurs at different paces. This similarity suggests that precipitation initiation processes are alike despite variations in onset times., Comment: 4 pages, 3 figures, accepted at NeurIPS 2023 (Machine Learning and the Physical Sciences Workshop)

We introduce SAGE; a Generative LLM for inferring attribute values for products across world-wide e-Commerce catalogs. We introduce a novel formulation of the attribute-value prediction problem as a Seq2Seq summarization task, across languages, product types and target attributes. Our novel modeling approach lifts the restriction of predicting attribute values within a pre-specified set of choices, as well as, the requirement that the sought attribute values need to be explicitly mentioned in the text. SAGE can infer attribute values even when such values are mentioned implicitly using periphrastic language, or not-at-all-as is the case for common-sense defaults. Additionally, SAGE is capable of predicting whether an attribute is inapplicable for the product at hand, or non-obtainable from the available information. SAGE is the first method able to tackle all aspects of the attribute-value-prediction task as they arise in practical settings in e-Commerce catalogs. A comprehensive set of experiments demonstrates the effectiveness of the proposed approach, as well as, its superiority against state-of-the-art competing alternatives. Moreover, our experiments highlight SAGE's ability to tackle the task of predicting attribute values in zero-shot setting; thereby, opening up opportunities for significantly reducing the overall number of labeled examples required for training., Comment: (17 pages)

Recently, the fundamental problem of unsupervised domain adaptation (UDA) on 3D point clouds has been motivated by a wide variety of applications in robotics, virtual reality, and scene understanding, to name a few. The point cloud data acquisition procedures manifest themselves as significant domain discrepancies and geometric variations among both similar and dissimilar classes. The standard domain adaptation methods developed for images do not directly translate to point cloud data because of their complex geometric nature. To address this challenge, we leverage the idea of multimodality and alignment between distributions. We propose a new UDA architecture for point cloud classification that benefits from multimodal contrastive learning to get better class separation in both domains individually. Further, the use of optimal transport (OT) aims at learning source and target data distributions jointly to reduce the cross-domain shift and provide a better alignment. We conduct a comprehensive empirical study on PointDA-10 and GraspNetPC-10 and show that our method achieves state-of-the-art performance on GraspNetPC-10 (with approx 4-12% margin) and best average performance on PointDA-10. Our ablation studies and decision boundary analysis also validate the significance of our contrastive learning module and OT alignment.

We consider selfish sources that send updates to a monitor over a shared wireless access. The sources would like to minimize the age of their information at the monitor. Our goal is to devise strategies that incentivize such sources to use the shared spectrum cooperatively. Earlier work has modeled such a setting using a non-cooperative one-shot game, played over a single access slot, and has shown that under certain access settings the dominant strategy of each source is to transmit in any slot, resulting in packet collisions between the sources' transmissions and causing all of them to be decoded in error at the monitor. We capture the interaction of the sources over an infinitely many medium access slots using infinitely repeated games. We investigate strategies that enable cooperation resulting in an efficient use of the wireless access, while disincentivizing any source from unilaterally deviating from the strategy. Formally, we are interested in strategies that are a subgame perfect Nash equilibrium (SPNE). We begin by investigating the properties of the one-stage (slot) optimal and access-fair correlated strategies. We then consider their many-slot variants, the age-fair and access-fair strategies, in the infinitely repeated game model. We prove that the access-fair and age-fair strategies are SPNEs for when collision slots are longer than successful transmission slots. Otherwise, neither is a SPNE. We end with simulations that shed light on a possible SPNE for the latter case., Comment: Accepted for publication in IEEE Global Communications Conference (GLOBECOM) 2023

DP-coloring (also called correspondence coloring) of graphs is a generalization of list coloring that has been widely studied since its introduction by Dvo\v{r}\'{a}k and Postle in $2015$. Intuitively, DP-coloring generalizes list coloring by allowing the colors that are identified as the same to vary from edge to edge. Formally, DP-coloring of a graph $G$ is equivalent to an independent transversal in an auxiliary structure called a DP-cover of $G$. In this paper, we introduce the notion of random DP-covers and study the behavior of DP-coloring from such random covers. We prove a series of results about the probability that a graph is or is not DP-colorable from a random cover. These results support the following threshold behavior on random $k$-fold DP-covers as $\rho\to\infty$ where $\rho$ is the maximum density of a graph: graphs are non-DP-colorable with high probability when $k$ is sufficiently smaller than $\rho/\ln\rho$, and graphs are DP-colorable with high probability when $k$ is sufficiently larger than $\rho/\ln\rho$. Our results depend on $\rho$ growing fast enough and imply a sharp threshold for dense enough graphs. For sparser graphs, we analyze DP-colorability in terms of degeneracy. We also prove fractional DP-coloring analogs to these results., Comment: 19 pages

Our knowledge of the organisation of the human brain at the population-level is yet to translate into power to predict functional differences at the individual-level, limiting clinical applications, and casting doubt on the generalisability of inferred mechanisms. It remains unknown whether the difficulty arises from the absence of individuating biological patterns within the brain, or from limited power to access them with the models and compute at our disposal. Here we comprehensively investigate the resolvability of such patterns with data and compute at unprecedented scale. Across 23 810 unique participants from UK Biobank, we systematically evaluate the predictability of 25 individual biological characteristics, from all available combinations of structural and functional neuroimaging data. Over 4526 GPU hours of computation, we train, optimize, and evaluate out-of-sample 700 individual predictive models, including fully-connected feed-forward neural networks of demographic, psychological, serological, chronic disease, and functional connectivity characteristics, and both uni- and multi-modal 3D convolutional neural network models of macro- and micro-structural brain imaging. We find a marked discrepancy between the high predictability of sex (balanced accuracy 99.7%), age (mean absolute error 2.048 years, R2 0.859), and weight (mean absolute error 2.609Kg, R2 0.625), for which we set new state-of-the-art performance, and the surprisingly low predictability of other characteristics. Neither structural nor functional imaging predicted psychology better than the coincidence of chronic disease (p<0.05). Serology predicted chronic disease (p<0.05) and was best predicted by it (p<0.001), followed by structural neuroimaging (p<0.05). Our findings suggest either more informative imaging or more powerful models are needed to decipher individual level characteristics from the human brain., Comment: 38 pages, 6 figures, 1 table, 2 supplementary figures, 1 supplementary table

We adopt magnetohydrodynamics (MHD) simulations that model the formation of filamentary molecular clouds via the collision-induced magnetic reconnection (CMR) mechanism under varying physical conditions. We conduct radiative transfer using RADMC-3D to generate synthetic dust emission of CMR filaments. We use the previously developed machine learning technique CASI-2D along with the diffusion model to identify the location of CMR filaments in dust emission. Both models showed a high level of accuracy in identifying CMR filaments in the test dataset, with detection rates of over 80% and 70%, respectively, at a false detection rate of 5%. We then apply the models to real Herschel dust observations of different molecular clouds, successfully identifying several high-confidence CMR filament candidates. Notably, the models are able to detect high-confidence CMR filament candidates in Orion A from dust emission, which have previously been identified using molecular line emission., Comment: ApJ accepted

The emergence of quantum computing raises the question of how to identify (security-relevant) programming errors during development. However, current static code analysis tools fail to model information specific to quantum computing. In this paper, we identify this information and propose to extend classical code analysis tools accordingly. Among such tools, we identify the Code Property Graph to be very well suited for this task as it can be easily extended with quantum computing specific information. For our proof of concept, we implemented a tool which includes information from the quantum world in the graph and demonstrate its ability to analyze source code written in Qiskit and OpenQASM. Our tool brings together the information from the classical and quantum world, enabling analysis across both domains. By combining all relevant information into a single detailed analysis, this powerful tool can facilitate tackling future quantum source code analysis challenges., Comment: 2023 IEEE International Conference on Quantum Computing and Engineering (QCE)

One of the most studied extensions of the famous Traveling Salesperson Problem (TSP) is the {\sc Multiple TSP}: a set of $m\geq 1$ salespersons collectively traverses a set of $n$ cities by $m$ non-trivial tours, to minimize the total length of their tours. This problem can also be considered to be a variant of {\sc Uncapacitated Vehicle Routing} where the objective function is the sum of all tour lengths. When all $m$ tours start from a single common \emph{depot} $v_0$, then the metric {\sc Multiple TSP} can be approximated equally well as the standard metric TSP, as shown by Frieze (1983). The {\sc Multiple TSP} becomes significantly harder to approximate when there is a \emph{set} $D$ of $d \geq 1$ depots that form the starting and end points of the $m$ tours. For this case only a $(2-1/d)$-approximation in polynomial time is known, as well as a $3/2$-approximation for \emph{constant} $d$ which requires a prohibitive run time of $n^{\Theta(d)}$ (Xu and Rodrigues, \emph{INFORMS J. Comput.}, 2015). A recent work of Traub, Vygen and Zenklusen (STOC 2020) gives another approximation algorithm for {\sc Multiple TSP} running in time $n^{\Theta(d)}$ and reducing the problem to approximating TSP. In this paper we overcome the $n^{\Theta(d)}$ time barrier: we give the first efficient approximation algorithm for {\sc Multiple TSP} with a \emph{variable} number $d$ of depots that yields a better-than-2 approximation. Our algorithm runs in time $(1/\varepsilon)^{\mathcal O(d\log d)}\cdot n^{\mathcal O(1)}$, and produces a $(3/2+\varepsilon)$-approximation with constant probability. For the graphic case, we obtain a deterministic $3/2$-approximation in time $2^d\cdot n^{\mathcal O(1)}$.ithm for metric {\sc Multiple TSP} with run time $n^{\Theta(d)}$, which reduces the problem to approximating metric TSP., Comment: To be published at ESA 2023

Magnetic circular dichroism (MCD) spectroscopy was used to study the features of the electronic structure of an epitaxial La0.7Ca0.3MnO3 film in the range of 1.2 - 4 eV. The study of the temperature behavior of the MCD spectra made it possible to establish a correlation between the magnetooptical and transport properties of the sample. The data obtained were analyzed in comparison with MCD data for polycrystalline manganite films of various RE1-xAxMnO3 compositions. The MCD spectra of the films were compared with the spectra of the off-diagonal component of the permittivity tensor calculated from the data of the magneto-optical Kerr effect for films of the same composition. A unified set of ground and excited electronic states characteristic of RE1-xAxMnO3 manganites in the visible and near infrared ranges has been identified. These results are important for a qualitative theoretical description of the electronic structure of strongly correlated magnetic oxides., Comment: 5 figures, 2 tables

Anticipating audience reaction towards a certain text is integral to several facets of society ranging from politics, research, and commercial industries. Sentiment analysis (SA) is a useful natural language processing (NLP) technique that utilizes lexical/statistical and deep learning methods to determine whether different-sized texts exhibit positive, negative, or neutral emotions. Recurrent networks are widely used in machine-learning communities for problems with sequential data. However, a drawback of models based on Long-Short Term Memory networks and Gated Recurrent Units is the significantly high number of parameters, and thus, such models are computationally expensive. This drawback is even more significant when the available data are limited. Also, such models require significant over-parameterization and regularization to achieve optimal performance. Tensorized models represent a potential solution. In this paper, we classify the sentiment of some social media posts. We compare traditional recurrent models with their tensorized version, and we show that with the tensorized models, we reach comparable performances with respect to the traditional models while using fewer resources for the training.

We study the NP-complete Maximum Outerplanar Subgraph problem. The previous best known approximation ratio for this problem is 2/3. We propose a new approximation algorithm which improves the ratio to 7/10.

The goal of this paper is open-vocabulary object detection (OVOD) $\unicode{x2013}$ building a model that can detect objects beyond the set of categories seen at training, thus enabling the user to specify categories of interest at inference without the need for model retraining. We adopt a standard two-stage object detector architecture, and explore three ways for specifying novel categories: via language descriptions, via image exemplars, or via a combination of the two. We make three contributions: first, we prompt a large language model (LLM) to generate informative language descriptions for object classes, and construct powerful text-based classifiers; second, we employ a visual aggregator on image exemplars that can ingest any number of images as input, forming vision-based classifiers; and third, we provide a simple method to fuse information from language descriptions and image exemplars, yielding a multi-modal classifier. When evaluating on the challenging LVIS open-vocabulary benchmark we demonstrate that: (i) our text-based classifiers outperform all previous OVOD works; (ii) our vision-based classifiers perform as well as text-based classifiers in prior work; (iii) using multi-modal classifiers perform better than either modality alone; and finally, (iv) our text-based and multi-modal classifiers yield better performance than a fully-supervised detector., Comment: ICML 2023, project page: https://www.robots.ox.ac.uk/vgg/research/mm-ovod/