With the launch of NOAA-15 satellite in May 1998, a new generation passive microwave sounders was initiated. The Advanced Microwave Sounding Unit (AMSU), with 20 channels spanning frequency range from 23-183 GHz, offers enhanced temperature and moisture sounding capability well beyond its predecessor, (MSU). In addition, by utilizing number window on AMSU, National Oceanic Atmospheric Administration (NOAA) expanded AMSU this original purpose developed suite products that are generated...

This paper proposes an image-to-painting translation method that generates vivid and realistic painting artworks with controllable styles. Different from previous image-to-image methods formulate the as pixel-wise prediction, we deal such artistic creation process in a vectorized environment produce sequence of physically meaningful stroke parameters can be further used for rendering. Since typical vector render is not differentiable, design novel neural renderer which imitates behavior then...

Federated learning has become increasingly important for modern machine learning, especially data privacy-sensitive scenarios. Existing federated mostly adopts the central server-based architecture or centralized architecture. However, in many social network scenarios, is not applicable (e.g., a agent server connecting all users may exist, communication cost to affordable). In this paper, we consider generic setting: 1) and 2) unidirectional of single-sided trust (i.e., user A trusts B but...

To examine the utility of using satellite passive microwave observations to measure soil moisture over large regions, we conducted a pilot study scanning multichannel radiometer (SMMR) on Nimbus‐7, which operated from 1978 1987, and actual in situ state Illinois, United States, began 1981. We examined SMMR midnight brightness temperatures 0.5° × grid, compared them with direct measurements at 14 sites Illinois for period 1982–1987. The results suggest that both polarization difference...

Rapid Serial Visual Presentation (RSVP)-based Brain-Computer Interfaces (BCIs) facilitate high-throughput target image detection by identifying event-related potentials (ERPs) evoked in EEG signals. The RSVP-BCI systems effectively detect single-class targets within a stream of images but have limited applicability scenarios that require detecting multiple categories. Multi-class address this limitation simultaneously the presence and distinguishing its category. However, existing...

Diffusion models have achieved remarkable success in sequential decision-making by leveraging the highly expressive model capabilities policy learning. A central problem for learning diffusion policies is to align output with human intents various tasks. To achieve this, previous methods conduct return-conditioned generation or Reinforcement Learning (RL)-based optimization, while they both rely on pre-defined reward functions. In this work, we propose a novel framework, Forward KL...

Actor-critic algorithm and their extensions have made great achievements in real-world decision-making problems. In contrast to its empirical success, the theoretical understanding of actor-critic seems unsatisfactory. Most existing results only show asymptotic convergence, which is developed mainly based on approximating dynamic system actor critic using ordinary differential equations. However, finite-time convergence analysis remains be explored. The main challenges lie nonconvexity...

By analyzing in situ soil moisture data, we show that variability consists of two components, one which is related to large‐scale atmospheric forcing, and the other small‐scale land surface hydrologic processes. We use empirically estimated spatial autocorrelation functions for Illinois estimate errors averaging observations, using method statistically optimal meteorological fields. The dependence root‐mean‐square on station network density can be used analyze existing observational networks...

We propose a vision-based framework for dynamic sky replacement and harmonization in videos. Different from previous editing methods that either focus on static photos or require real-time pose signal the camera's inertial measurement units, our method is purely vision-based, without any requirements capturing devices, can be well applied to online offline processing scenarios. Our runs free of manual interactions. decompose video into several proxy tasks, including motion estimation,...

Recently, solar energy has been gaining attention as one of the best promising renewable sources. Accurate PV power prediction models can solve impact on system due to non-linearity and randomness generation play a crucial role in operation scheduling plants. This paper proposes novel machine learning network framework predict short-term time-series manner. The combination nonlinear auto-regressive neural networks with exogenous input (NARX), long short term memory (LSTM) network, light...

Functional annotation of human genes is fundamentally important for understanding the molecular basis various genetic diseases. A major challenge in determining functions lies functional diversity proteins, that is, a gene can perform different as it may consist multiple protein coding isoforms (PCIs). Therefore, differentiating PCIs significantly deepen our genes. However, due to lack isoform-level gold-standards (ground-truth annotation), many existing approaches are developed at...

Euwallacea fornicatus (Eichhoff) is an important forest pest that has caused serious damage in America and Vietnam. In 2014, it attacked forests of Acer trialatum the Yunnan province China, creating concern China's Forestry Bureau. We used CLIMEX model to predict compare potential distribution for E. fornicates China under current (1981-2010) projected climate conditions (2011-2040) using one scenario (RCP8.5) global (GCM), CSIRO-Mk3-6-0. Under both future conditions, predicted be mainly...

The genome-wide association study (GWAS) is a popular approach to identify disease-associated genetic factors for Alzhemer's Disease (AD). However, it remains challenging because of the small number samples, very high feature dimensionality and complex structures. To accurately risk AD, we propose novel method based on an in-depth exploration hierarchical structure among features commonality across related tasks. Specifically, first extract encode tree hierarchy features; then, integrate...

Decentralized Online Learning (online learning in decentralized networks) attracts more and attention, since it is believed that can help the data providers cooperatively better solve their online problems without sharing private to a third party or other providers. Typically, cooperation achieved by letting exchange models between neighbors, e.g., recommendation model. However, best regret bound for algorithm $\Ocal{n\sqrt{T}}$, where $n$ number of nodes (or users) $T$ iterations. This...

Mapping vegetation cover is critical for understanding and monitoring ecosystem functions in semi-arid biomes. As existing estimates tend to underestimate the woody areas with dry deciduous shrubland woodland, we present an approach improve regional estimation of herbaceous fractional East Asia steppe. This developed uses Random Forest models by combining multiple remote sensing data—training samples derived from high-resolution image a tailored spatial sampling model inputs composed...

Abstract Silicon (Si) not only plays an important role in plant growth but also contributes significantly to the long‐term terrestrial carbon sink form of phytoliths. This study investigated Si content 184 species meadow steppe and typical northern China examine influential factors distribution evaluate potential phytolith sequestration these grasslands. Our results indicated that average generally decreased following order Equisetopsida > Monocotyledoneae Dicotyledoneae. Within...

Temporal-Difference (TD) learning with nonlinear smooth function approximation for policy evaluation has achieved great success in modern reinforcement learning. It is shown that such a problem can be reformulated as stochastic nonconvex-strongly-concave optimization problem, which challenging naive gradient descent-ascent algorithm suffers from slow convergence. Existing approaches this are based on two-timescale or double-loop algorithms, may also require sampling large-batch data....

We study the robust one-bit compressed sensing problem whose goal is to design an algorithm that faithfully recovers any sparse target vector $θ_0\in\mathbb{R}^d$ \textit{uniformly} via $m$ quantized noisy measurements. Specifically, we consider a new framework for this where sparsity implicitly enforced mapping low dimensional representation $x_0 \in \mathbb{R}^k$ through known $n$-layer ReLU generative network $G:\mathbb{R}^k\rightarrow\mathbb{R}^d$ such $θ_0 = G(x_0)$. Such poses...

This paper focuses on generating multi-hop reasoning questions from the raw text in a low resource circumstance. Such have to be syntactically valid and need logically correlate with answers by deducing over multiple relations several sentences text. Specifically, we first build generation model guide it satisfy logical rationality chain extracted given Since labeled data is limited insufficient for training, propose learn help of large scale unlabeled that much easier obtain. contains rich...