Search for: All records

Deep learning’s performance has been extensively recognized recently. Graph neural networks (GNNs) are designed to deal with graph-structural data that classical deep learning does not easily manage. Since most GNNs were created using distinct theories, direct comparisons are impossible. Prior research has primarily concentrated on categorizing existing models, with little attention paid to their intrinsic connections. The purpose of this study is to establish a unified framework that integrates GNNs based on spectral graph and approximation theory. The framework incorporates a strong integration between spatial- and spectral-based GNNs while tightly associating approaches that exist within each respective domain.

 

Top-k frequent items detection is a fundamental task in data stream mining. Many promising solutions are proposed to improve memory efficiency while still maintaining high accuracy for detecting the Top-k items. Despite the memory efficiency concern, the users could suffer from privacy loss if participating in the task without proper protection, since their contributed local data streams may continually leak sensitive individual information. However, most existing works solely focus on addressing either the memory-efficiency problem or the privacy concerns but seldom jointly, which cannot achieve a satisfactory tradeoff between memory efficiency, privacy protection, and detection accuracy.

In this paper, we present a novel framework HG-LDP to achieve accurate Top-k item detection at bounded memory expense, while providing rigorous local differential privacy (LDP) protection. Specifically, we identify two key challenges naturally arising in the task, which reveal that directly applying existing LDP techniques will lead to an inferior accuracy-privacy-memory efficiency tradeoff. Therefore, we instantiate three advanced schemes under the framework by designing novel LDP randomization methods, which address the hurdles caused by the large size of the item domain and by the limited space of the memory. We conduct comprehensive experiments on both synthetic and real-world datasets to show that the proposed advanced schemes achieve a superior accuracy-privacy-memory efficiency tradeoff, saving 2300× memory over baseline methods when the item domain size is 41,270. Our code is anonymously open-sourced via the link.

 

It has long been known that the alteration of protein side chains that occlude or expose the heme cofactor to water can greatly affect the stability of the oxyferrous heme state. Here, we demonstrate that the rate of dynamically driven water penetration into the core of an artificial oxygen transport protein also correlates with oxyferrous state lifetime by reducing global dynamics, without altering the structure of the active site, via the simple linking of the two monomers in a homodimeric artificial oxygen transport protein using a glycine-rich loop. The tethering of these two helices does not significantly affect the active site structure, pentacoordinate heme-binding affinity, reduction potential, or gaseous ligand affinity. It does, however, significantly reduce the hydration of the protein core, as demonstrated by resonance Raman spectroscopy, backbone amide hydrogen exchange, and pKa shifts in buried histidine side chains. This further destabilizes the charge-buried entatic state and nearly triples the oxyferrous state lifetime. These data are the first direct evidence that dynamically driven water penetration is a rate-limiting step in the oxidation of these complexes. It furthermore demonstrates that structural rigidity that limits water penetration is a critical design feature in metalloenzyme construction and provides an explanation for both the failures and successes of earlier attempts to create oxygen-binding proteins. 

The bottom of the lithosphere is characterized by a thermally controlled transition from brittle to ductile deformation. While the mechanical behavior of rocks firmly within the brittle and ductile regimes is relatively well understood, how the transition operates remains elusive. Here, we study the mechanical properties of pure olivine gouge from 100 to 500°C under 100 MPa pore‐fluid pressure in a triaxial deformation apparatus as a proxy for the mechanical properties of the upper mantle across the brittle‐ductile transition. We describe the mechanical data with a rate‐, state‐, and temperature‐dependent constitutive law with multiple thermally activated deformation mechanisms. The stress power exponents decrease from 70 ± 10 in the brittle regime to 17 ± 3 and 4 ± 2 in the semi‐brittle and ductile regimes, respectively. The mechanical model consistently explains the mechanical behavior of olivine gouge across the brittle‐ductile transition, capturing the gradual evolution from cataclasis to crystal plasticity.

 

Finding Nash equilibrial policies for two-player differential games requires solving Hamilton-Jacobi-Isaacs (HJI) PDEs. Self-supervised learning has been used to approximate solutions of such PDEs while circumventing the curse of dimensionality. However, this method fails to learn discontinuous PDE solutions due to its sampling nature, leading to poor safety performance of the resulting controllers in robotics applications when player rewards are discontinuous. This paper investigates two potential solutions to this problem: a hybrid method that leverages both supervised Nash equilibria and the HJI PDE, and a value-hardening method where a sequence of HJIs are solved with a gradually hardening reward. We compare these solutions using the resulting generalization and safety performance in two vehicle interaction simulation studies with 5D and 9D state spaces, respectively. Results show that with informative supervision (e.g., collision and near-collision demonstrations) and the low cost of self-supervised learning, the hybrid method achieves better safety performance than the supervised, self-supervised, and value hardening approaches on equal computational budget. Value hardening fails to generalize in the higher-dimensional case without informative supervision. Lastly, we show that the neural activation function needs to be continuously differentiable for learning PDEs and its choice can be case dependent.