The past years have seen a drastic rise in studies devoted to the investigation of colexification patterns in individual languages families in particular and the languages of the world in specific. Specifically computational studies have profited from the fact that colexification as a scientific construct is easy to operationalize, enabling scholars to infer colexification patterns for large collections of cross-linguistic data. Studies devoted to partial colexifications -- colexification patterns that do not involve entire words, but rather various parts of words--, however, have been rarely conducted so far. This is not surprising, since partial colexifications are less easy to deal with in computational approaches and may easily suffer from all kinds of noise resulting from false positive matches. In order to address this problem, this study proposes new approaches to the handling of partial colexifications by (1) proposing new models with which partial colexification patterns can be represented, (2) developing new efficient methods and workflows which help to infer various types of partial colexification patterns from multilingual wordlists, and (3) illustrating how inferred patterns of partial colexifications can be computationally analyzed and interactively visualized.

With news and information being as easy to access as they currently are, it is more important than ever to ensure that people are not mislead by what they read. Recently, the rise of neural fake news (AI-generated fake news) and its demonstrated effectiveness at fooling humans has prompted the development of models to detect it. One such model is the Grover model, which can both detect neural fake news to prevent it, and generate it to demonstrate how a model could be misused to fool human readers. In this work we explore the Grover model's fake news detection capabilities by performing targeted attacks through perturbations on input news articles. Through this we test Grover's resilience to these adversarial attacks and expose some potential vulnerabilities which should be addressed in further iterations to ensure it can detect all types of fake news accurately.

Neural networks-based machine learning methods are rapidly developed for various numerical problems, such as physics-informed neural networks (PINNs) [10, 11, 12], the deep Ritz method [2], and the deep Galerkin method [15]. One of the advantages of these approaches is that they show the possibility for solving high-dimensional problems. In [2], the deep learning techniques as well as the Monte-Carlo integration are used to solve eigenvalue problems, which provides a feasible strategy for high-dimensional cases. For the same eigenvalue problems, [16] applies a neural network-based low-rank tensor model, i.e. the tensor neural network (TNN), with a quadrature scheme to perform efficient numerical integration, and thus it achieves a much better result than [2]. Furthermore, [17] employs the TNN to solve the manybody Schrödinger equation, which emerges the practical value of such low-rank approximation method.

This study investigates the use of Visually Grounded Speech (VGS) models for keyword localisation in speech. The study focusses on two main research questions: (1) Is keyword localisation possible with VGS models and (2) Can keyword localisation be done cross-lingually in a real low-resource setting? Four methods for localisation are proposed and evaluated on an English dataset, with the best-performing method achieving an accuracy of 57%. A new dataset containing spoken captions in Yoruba language is also collected and released for cross-lingual keyword localisation. The cross-lingual model obtains a precision of 16% in actual keyword localisation and this performance can be improved by initialising from a model pretrained on English data. The study presents a detailed analysis of the model's success and failure modes and highlights the challenges of using VGS models for keyword localisation in low-resource settings.

The Robert C. Byrd Green Bank Telescope in West Virginia is one of several helping to search for alien civilizations.Credit: Jim West/Alamy From the hills of West Virginia to the flats of rural Australia, some of the world's largest telescopes are listening for signals from distant alien civilizations. The search for extraterrestrial intelligence, known as SETI, is an effort to find artificial-looking electromagnetic radiation that might have come from a technologically advanced civilization in a far-away solar system. A study published today1 describes one of several efforts to use machine learning, a subset of artificial intelligence (AI), to help astronomers sift quickly through the reams of data such searches yield. As AI reshapes many scientific fields, what promise does it hold for the search for life beyond Earth? "It is a new era for SETI research that is opening up thanks to machine learning technology," says Franck Marchis, a planetary astronomer at the SETI Institute in Mountain View, California.

Human syntactic structures are usually represented as graphs. Much research has focused on the mapping between such graphs and linguistic sequences, but less attention has been paid to the shapes of the graphs themselves: their topologies. This study investigates how the topologies of syntactic graphs reveal traces of the processes that led to their emergence. I report a new universal regularity in syntactic structures: Their topology is communicatively efficient above chance. The pattern holds, without exception, for all 124 languages studied, across linguistic families and modalities (spoken, written, and signed). This pattern can arise from a process optimizing for communicative efficiency or, alternatively, by construction, as a by-effect of a sublinear preferential attachment process reflecting language production mechanisms known from psycholinguistics. This dual explanation shows how communicative efficiency, per se, does not require optimization. Among the two options, efficiency without optimization offers the better explanation for the new pattern.

Tuberculosis (TB) is still considered a leading cause of death and a substantial threat to global child health. Both TB infection and disease are curable using antibiotics. However, most children who die of TB are never diagnosed or treated. In clinical practice, experienced physicians assess TB by examining chest X-rays (CXR). Pediatric CXR has specific challenges compared to adult CXR, which makes TB diagnosis in children more difficult. Computer-aided diagnosis systems supported by Artificial Intelligence have shown performance comparable to experienced radiologist TB readings, which could ease mass TB screening and reduce clinical burden. We propose a multi-view deep learning-based solution which, by following a proposed template, aims to automatically regionalize and extract lung and mediastinal regions of interest from pediatric CXR images where key TB findings may be present. Experimental results have shown accurate region extraction, which can be used for further analysis to confirm TB finding presence and severity assessment.

In the last decades, due to the huge technological growth observed, it has become increasingly common that a collection of temporal data rapidly accumulates in vast amounts. This provides an opportunity for extracting valuable information through the estimation of increasingly precise models. But at the same time it imposes the challenge of continuously updating the models as new data become available. Currently available methods for addressing this problem, the so-called online learning methods, use current parameter estimations and novel data to update the estimators. These approaches avoid using the full raw data and speeding up the computations. In this work we consider three online learning algorithms for parameters estimation in the context of time series models. In particular, the methods implemented are: gradient descent, Newton-step and Kalman filter recursions. These algorithms are applied to the recently developed irregularly observed autoregressive (iAR) model. The estimation accuracy of the proposed methods is assessed by means of Monte Carlo experiments. The results obtained show that the proposed online estimation methods allow for a precise estimation of the parameters that generate the data both for the regularly and irregularly observed time series. These online approaches are numerically efficient, allowing substantial computational time savings. Moreover, we show that the proposed methods are able to adapt the parameter estimates quickly when the time series behavior changes, unlike batch estimation methods.

We propose a novel supervised learning approach for political ideology prediction (PIP) that is capable of predicting out-of-distribution inputs. This problem is motivated by the fact that manual data-labeling is expensive, while self-reported labels are often scarce and exhibit significant selection bias. We propose a novel statistical model that decomposes the document embeddings into a linear superposition of two vectors; a latent neutral \emph{context} vector independent of ideology, and a latent \emph{position} vector aligned with ideology. We train an end-to-end model that has intermediate contextual and positional vectors as outputs. At deployment time, our model predicts labels for input documents by exclusively leveraging the predicted positional vectors. On two benchmark datasets we show that our model is capable of outputting predictions even when trained with as little as 5\% biased data, and is significantly more accurate than the state-of-the-art. Through crowd-sourcing we validate the neutrality of contextual vectors, and show that context filtering results in ideological concentration, allowing for prediction on out-of-distribution examples.

From natural language processing to genome sequencing, large-scale machine learning models are bringing advances to a broad range of fields. Many of these models are too large to be trained on a single machine, and instead must be distributed across multiple devices. This has motivated the research of new compute and network systems capable of handling such tasks. In particular, recent work has focused on developing management schemes which decide how to allocate distributed resources such that some overall objective, such as minimising the job completion time (JCT), is optimised. However, such studies omit explicit consideration of how much a job should be distributed, usually assuming that maximum distribution is desirable. In this work, we show that maximum parallelisation is sub-optimal in relation to user-critical metrics such as throughput and blocking rate. To address this, we propose PAC-ML (partitioning for asynchronous computing with machine learning). PAC-ML leverages a graph neural network and reinforcement learning to learn how much to partition computation graphs such that the number of jobs which meet arbitrary user-defined JCT requirements is maximised. In experiments with five real deep learning computation graphs on a recently proposed optical architecture across four user-defined JCT requirement distributions, we demonstrate PAC-ML achieving up to 56.2% lower blocking rates in dynamic job arrival settings than the canonical maximum parallelisation strategy used by most prior works.