The exceptional resolution achieved by the EHT is made possible by an array of telescopes spanning the Earth and operating as a very long baseline interferometer (VLBI; Event Horizon Telescope Collaboration et al. 2019b, 2019c). Despite this global reach, the sparse interferometric coverage of the EHT array (especially during the 2017 observations that have been used for all of the publications to date) makes the already complex problem of interferometric image reconstruction particularly challenging. In such situations, special care is needed to assess the impact of imaging algorithms and sparse interferometric data on the final set of images that can be reconstructed from it. A cornerstone of the EHT data analysis strategy was the use of several independent analysis methods, each with different priorities, assumptions, and choices, to ensure that the EHT results were robust to these differences. The use of several general-purpose imaging algorithms, for example, was motivated by a desire to reconstruct an image that was consistent with the EHT data while remaining model-agnostic.

In unveiling the nature of the first stars, the main astronomical clue is the elemental compositions of the second generation of stars, observed as extremely metal-poor (EMP) stars, in the Milky Way. However, no observational constraint was available on their multiplicity, which is crucial for understanding early phases of galaxy formation. We develop a new data-driven method to classify observed EMP stars into mono- or multi-enriched stars with support vector machines. We also use our own nucleosynthesis yields of core-collapse supernovae with mixing fallback that can explain many of the observed EMP stars. Our method predicts, for the first time, that 31.8%

We study on-line gradient-descent learning in multilayer networks analytically and numerically. The training is based on randomly drawn inputs and their corresponding outputs as defined by a target rule. In the thermodynamic limit we derive deterministic differential equations for the order parameters of the problem which allow an exact calculation of the evolution of the generalization error. First we consider a single-layer perceptron with sigmoidal activation function learning a target rule defined by a network of the same architecture. For this model the generalization error decays exponentially with the number of training examples if the learning rate is sufficiently small.

Most of the human activities in daily life are communicating with others. Navigation, manipulation and gesture are some of the basic interactions. Perhaps people are supported by ground to navigate complex situations and avoid obstacles [1]. They always use their hands and fingers in a large number of tasks to communicate with others (through communication gestures) or with the physical world around them. This recommends that movement and communication with the environment are strongly intertwined [2].

Wearable devices equipped with plethysmography (PPG) sensors provided a low-cost, long-term solution to early diagnosis and continuous screening of heart conditions. However PPG signals collected from such devices often suffer from corruption caused by artifacts. The objective of this study is to develop an effective supervised algorithm to locate the regions of artifacts within PPG signals. We treat artifact detection as a 1D segmentation problem. We solve it via a novel combination of an active-contour-based loss and an adapted U-Net architecture.

Space weather phenomena such as solar flares have a massive destructive power when they reach a certain magnitude. Here, we explore the deep-learning approach in order to build a solar flare-forecasting model, while examining its limitations and feature-extraction ability based on the available Geostationary Operational Environmental Satellite (GOES) X-ray time-series data. We present a multilayer 1D convolutional neural network to forecast the solar flare event probability occurrence of M- and X-class flares at 1, 3, 6, 12, 24, 48, 72, and 96 hr time frames. The forecasting models were trained and evaluated in two different scenarios: (1) random selection and (2) chronological selection, which were compared afterward in terms of common score metrics. Additionally, we also compared our results to state-of-the-art flare-forecasting models.

The following article is Open access.

The development of machine learning methods based on deep learning boosted the field of artificial intelligence towards unprecedented achievements and application in several fields. Such prominent results were made in parallel with the first successful demonstrations of fault tolerant hardware for quantum information processing. To which extent deep learning can take advantage of the existence of a hardware based on qubits behaving as a universal quantum computer is an open question under investigation. Here I review the convergence between the two fields towards implementation of advanced quantum algorithms, including quantum deep learning.

In stochastic systems, numerically sampling the relevant trajectories for the estimation of the large deviation statistics of time-extensive observables requires overcoming their exponential (in space and time) scarcity. The optimal way to access these rare events is by means of an auxiliary dynamics obtained from the original one through the so-called'generalised Doob transformation'. While this optimal dynamics is guaranteed to exist its use is often impractical, as to define it requires the often impossible task of diagonalising a (tilted) dynamical generator. While approximate schemes have been devised to overcome this issue they are difficult to automate as they tend to require knowledge of the systems under study. Here we address this problem from the perspective of deep learning.

McCarthy et al [1] organized the Dartmouth workshop in 1956 to initiate artificial intelligence (AI) as a research field with a lofty goal to simulate, enhance, or even surpass human intelligence. Given the tremendous potentials and challenges, the excitements and frustrations are equally remarkable. Their interactions lead to alterations of AI springs and winters, through which the AI field has been developed step by step, and elevated to today's level, and we believe that this field will have an even brighter future. Currently, AI is in a new spring, especially its sub-field machine learning (ML) which enjoys rapid development and constant innovations featured by deep neural networks, also known as deep learning. On August 30, 2019, the White House issued a memorandum on the Fiscal Year 2021 Administration Research and Development Budget Priorities [2], underlining that'departments and agencies should prioritize basic and applied research investments that are consistent with the 2019 Executive Order on Maintaining American Leadership in Artificial Intelligence and the eight strategies detailed in the 2019 update of the National Artificial Intelligence Research and Development Strategic Plan.'