Explore the vast range of titles available.

Sampling is ubiquitous in machine learning methodologies. Due to the growth of large datasets and model complexity, we want to learn and adapt the sampling process while training a representation. Towards achieving this grand goal, a variety of sampling techniques have been proposed. However, most of them either use a fixed sampling scheme or adjust the sampling scheme based on simple heuristics. They cannot choose the best sample for model training in different stages. Inspired by \"Think, Fast and Slow\" (System 1 and System 2) in cognitive science, we propose a reward-guided sampling strategy called Adaptive Sample with Reward (ASR) to tackle this challenge. To the best of our knowledge, this is the first work utilizing reinforcement learning (RL) to address the sampling problem in representation learning. Our approach optimally adjusts the sampling process to achieve optimal performance. We explore geographical relationships among samples by distance-based sampling to maximize overall cumulative reward. We apply ASR to the long-standing sampling problems in similarity-based loss functions. Empirical results in information retrieval and clustering demonstrate ASR's superb performance across different datasets. We also discuss an engrossing phenomenon which we name as \"ASR gravity well\" in experiments.

Of late, many synthesis processes have been studied to develop irregular nano-rnorphologies of gold nanostructures for biomedical applications in order to increase the efficacy of nanoparticle theranostics, tune the plasmonic absorbance spectra, and increase the sensitivity of biomolecule detection through surface enhanced Raman spectroscopy. Here we report, a novel, non-seed mediated versatile single pot synthesis method capable of producing hyperbranched gold ＂nano-polyvilli＂ with more than 50-90 branching nanowires propagating from a single origin within each structure. The technique was capable of achieving precise tuning of the branch propagation where the branching could be controlled by varying the duration of incubation, temperature, and hydrogen ion concentration.

In a Human-in-the-Loop paradigm, a robotic agent is able to act mostly autonomously in solving a task, but can request help from an external expert when needed. However, knowing when to request such assistance is critical: too few requests can lead to the robot making mistakes, but too many requests can overload the expert. In this paper, we present a Reinforcement Learning based approach to this problem, where a semi-autonomous agent asks for external assistance when it has low confidence in the eventual success of the task. The confidence level is computed by estimating the variance of the return from the current state. We show that this estimate can be iteratively improved during training using a Bellman-like recursion. On discrete navigation problems with both fully- and partially-observable state information, we show that our method makes effective use of a limited budget of expert calls at run-time, despite having no access to the expert at training time.

We present two new classes of algorithms for efficient field integration on graphs encoding point clouds. The first class, SeparatorFactorization(SF), leverages the bounded genus of point cloud mesh graphs, while the second class, RFDiffusion(RFD), uses popular epsilon-nearest-neighbor graph representations for point clouds. Both can be viewed as providing the functionality of Fast Multipole Methods (FMMs), which have had a tremendous impact on efficient integration, but for non-Euclidean spaces. We focus on geometries induced by distributions of walk lengths between points (e.g., shortest-path distance). We provide an extensive theoretical analysis of our algorithms, obtaining new results in structural graph theory as a byproduct. We also perform exhaustive empirical evaluation, including on-surface interpolation for rigid and deformable objects (particularly for mesh-dynamics modeling), Wasserstein distance computations for point clouds, and the Gromov-Wasserstein variant.

Electrocatalytic reduction of waste nitrates (NO 3 − ) enables the synthesis of ammonia (NH 3 ) in a carbon neutral and decentralized manner. Atomically dispersed metal-nitrogen-carbon (M-N-C) catalysts demonstrate a high catalytic activity and uniquely favor mono-nitrogen products. However, the reaction fundamentals remain largely underexplored. Herein, we report a set of 14; 3 d -, 4 d -, 5 d - and f -block M-N-C catalysts. The selectivity and activity of NO 3 − reduction to NH 3 in neutral media, with a specific focus on deciphering the role of the NO 2 − intermediate in the reaction cascade, reveals strong correlations (R=0.9) between the NO 2 − reduction activity and NO 3 − reduction selectivity for NH 3 . Moreover, theoretical computations reveal the associative/dissociative adsorption pathways for NO 2 − evolution, over the normal M-N 4 sites and their oxo-form (O-M-N 4 ) for oxyphilic metals. This work provides a platform for designing multi-element NO 3 RR cascades with single-atom sites or their hybridization with extended catalytic surfaces. Understanding of atomically dispersed metal-nitrogen-carbon (M-N-C) for electrochemical nitrate reduction is important and of high interest. Here, the authors employ a series of M-N-C catalysts to investigate selectivity correlation between nitrate and nitrite reduction.

ObjectiveSqualene epoxidase (SQLE) promotes metabolic dysfunction-associated steatohepatitis-associated hepatocellular carcinoma (MASH-HCC), but its role in modulating the tumour immune microenvironment in MASH-HCC remains unclear.DesignWe established hepatocyte-specific Sqle transgenic (tg) and knockout mice, which were subjected to a choline-deficient high-fat diet plus diethylnitrosamine to induce MASH-HCC. SQLE function was also determined in orthotopic and humanised mice. Immune landscape alterations of MASH-HCC mediated by SQLE were profiled by single-cell RNA sequencing and flow cytometry.ResultsHepatocyte-specific Sqle tg mice exhibited a marked increase in MASH-HCC burden compared with wild-type littermates, together with decreased tumour-infiltrating functional IFN-γ+ and Granzyme B+ CD8+ T cells while enriching Arg-1+ myeloid-derived suppressor cells (MDSCs). Conversely, hepatocyte-specific Sqle knockout suppressed tumour growth with increased cytotoxic CD8+ T cells and reduced Arg-1+ MDSCs, inferring that SQLE promotes immunosuppression in MASH-HCC. Mechanistically, SQLE-driven cholesterol accumulation in tumour microenvironment underlies its effect on CD8+ T cells and MDSCs. SQLE and its metabolite, cholesterol, impaired CD8+ T cell activity by inducing mitochondrial dysfunction. Cholesterol depletion in vitro abolished the effect of SQLE-overexpressing MASH-HCC cell supernatant on CD8+ T cell suppression and MDSC activation, whereas cholesterol supplementation had contrasting functions on CD8+ T cells and MDSCs treated with SQLE-knockout supernatant. Targeting SQLE with genetic ablation or pharmacological inhibitor, terbinafine, rescued the efficacy of anti-PD-1 treatment in MASH-HCC models.ConclusionSQLE induces an impaired antitumour response in MASH-HCC via attenuating CD8+ T cell function and augmenting immunosuppressive MDSCs. SQLE is a promising target in boosting anti-PD-1 immunotherapy for MASH-HCC.

Li- and Mn-rich (LMR) cathode materials that utilize both cation and anion redox can yield substantial increases in battery energy density 1 – 3 . However, although voltage decay issues cause continuous energy loss and impede commercialization, the prerequisite driving force for this phenomenon remains a mystery 3 – 6 Here, with in situ nanoscale sensitive coherent X-ray diffraction imaging techniques, we reveal that nanostrain and lattice displacement accumulate continuously during operation of the cell. Evidence shows that this effect is the driving force for both structure degradation and oxygen loss, which trigger the well-known rapid voltage decay in LMR cathodes. By carrying out micro- to macro-length characterizations that span atomic structure, the primary particle, multiparticle and electrode levels, we demonstrate that the heterogeneous nature of LMR cathodes inevitably causes pernicious phase displacement/strain, which cannot be eliminated by conventional doping or coating methods. We therefore propose mesostructural design as a strategy to mitigate lattice displacement and inhomogeneous electrochemical/structural evolutions, thereby achieving stable voltage and capacity profiles. These findings highlight the significance of lattice strain/displacement in causing voltage decay and will inspire a wave of efforts to unlock the potential of the broad-scale commercialization of LMR cathode materials. Diffractive imaging of an important class of battery electrodes during cycling shows that lattice strain is a crucial yet overlooked factor that contributes to voltage fade over time.

Historically long accepted to be the singular root cause of capacity fading, transition metal dissolution has been reported to severely degrade the anode. However, its impact on the cathode behavior remains poorly understood. Here we show the correlation between capacity fading and phase/surface stability of an LiMn 2 O 4 cathode. It is revealed that a combination of structural transformation and transition metal dissolution dominates the cathode capacity fading. LiMn 2 O 4 exhibits irreversible phase transitions driven by manganese(III) disproportionation and Jahn-Teller distortion, which in conjunction with particle cracks results in serious manganese dissolution. Meanwhile, fast manganese dissolution in turn triggers irreversible structural evolution, and as such, forms a detrimental cycle constantly consuming active cathode components. Furthermore, lithium-rich LiMn 2 O 4 with lithium/manganese disorder and surface reconstruction could effectively suppress the irreversible phase transition and manganese dissolution. These findings close the loop of understanding capacity fading mechanisms and allow for development of longer life batteries. To unlock the potential of Mn-based cathode materials, the fast capacity fading process has to be first understood. Here the authors utilize advanced characterization techniques to look at a spinel LiMn 2 O 4 system, revealing that a combination of irreversible structural transformations and Mn dissolution takes responsibility.

Fibroblast growth factors (FGFs) and their receptors are significant components during fundamental cellular processes. FGF18 plays a distinctive role in modulating the activity of both tumor cells and tumor microenvironment. This study aims to comprehensively investigate the expression and functional role of FGF18 in gastric cancer (GC) and elucidate its regulatory mechanisms. The upregulation of FGF18 was detected in seven out of eleven (63.6%) GC cell lines. In primary GC samples, FGF18 was overexpressed in genomically stable and chromosomal instability subtypes of GC and its overexpression was associated with poor survival. Knocking down FGF18 inhibited tumor formation abilities, induced G1 phase cell cycle arrest and enhanced anti-cancer drug sensitivity. Expression microarray profiling revealed that silencing of FGF18 activated ATM pathway but quenched TGF-β pathway. The key factors that altered in the related signaling were validated by western blot and immunofluorescence. Meanwhile, treating GC cells with human recombinant FGF18 or FGF18-conditioned medium accelerated tumor growth through activation of ERK-MAPK signaling. FGF18 was further confirmed to be a direct target of tumor suppressor, miR-590-5p. Their expressions showed a negative correlation in primary GC samples and more importantly, re-overexpression of FGF18 partly abolished the tumor-suppressive effect of miR-590-5p. Our study not only identified that FGF18 serves as a novel prognostic marker and a therapeutic target in GC but also enriched the knowledge of FGF-FGFR signaling during gastric tumorigenesis.

The protein STK3/MST2 is a serine/threonine-protein kinase, a homologue of the Hippo protein in Drosophila, which plays an essential role in the Hippo signaling pathway. Traditionally, the activated STK3 kinase will undergo dimerization and negatively regulate the yes-associated protein 1 (YAP1) and WW Domain-Containing Transcription Regulator Protein 1 (TAZ), which trigger the expression of proliferation genes and inhibit apoptosis [3]. Strikingly, the cell populations with high STK3 expression demonstrated extremely high activities in cell cycle regulation, Ras signaling pathway, and MAPK signaling pathway, reinforcing the linkage between STK3 activation and tumor progression (Fig. 3d). More importantly, in the early state of tumor development, STK3 demonstrated high expression level, and the proliferation-associated genes, such as E2F1, CCND1, CDK4, and PCNA, followed similar kinetic trends with STK3, which was revealed by functional pseudotime analysis from a scRNA-seq dataset (Fig. 3g).