Search for: All records

The problem of few-shot graph classification targets at assigning class labels for graph samples, where only limited labeled graphs are provided for each class. To solve the problem brought by label scarcity, recent studies have proposed to adopt the prevalent few-shot learning framework to achieve fast adaptations to graph classes with limited labeled graphs. In particular, these studies typically propose to accumulate meta-knowledge across a large number of meta-training tasks, and then generalize such meta-knowledge to meta-test tasks sampled from a disjoint class set. Nevertheless, existing studies generally ignore the crucial task correlations among meta-training tasks and treat them independently. In fact, such task correlations can help promote the model generalization to meta-test tasks and result in better classification performance. On the other hand, it remains challenging to capture and utilize task correlations due to the complex components and interactions in meta-training tasks. To deal with this, we propose a novel few-shot graph classification framework FAITH to capture task correlations via learning a hierarchical task structure at different granularities. We further propose a task-specific classifier to incorporate the learned task correlations into the few-shot graph classification process. Moreover, we derive FAITH+, a variant of FAITH that can improve the sampling process for the hierarchical task structure. The extensive experiments on four prevalent graph datasets further demonstrate the superiority of FAITH and FAITH+ over other state-of-the-art baselines.

Federated learning (FL) has been widely studied recently due to its property to collaboratively train data from different devices without sharing the raw data. Nevertheless, recent studies show that an adversary can still be possible to infer private information about devices' data, e.g., sensitive attributes such as income, race, and sexual orientation. To mitigate the attribute inference attacks, various existing privacy-preserving FL methods can be adopted/adapted. However, all these existing methods have key limitations: they need to know the FL task in advance, or have intolerable computational overheads or utility losses, or do not have provable privacy guarantees. We address these issues and design a task-agnostic privacy-preserving presentation learning method for FL (TAPPFL) against attribute inference attacks. TAPPFL is formulated via information theory. Specifically, TAPPFL has two mutual information goals, where one goal learns task-agnostic data representations that contain the least information about the private attribute in each device's data, and the other goal ensures the learnt data representations include as much information as possible about the device data to maintain FL utility. We also derive privacy guarantees of TAPPFL against worst-case attribute inference attacks, as well as the inherent tradeoff between utility preservation and privacy protection. Extensive results on multiple datasets and applications validate the effectiveness of TAPPFL to protect data privacy, maintain the FL utility, and be efficient as well. Experimental results also show that TAPPFL outperforms the existing defenses.

Stellar-mass binary black holes (BBHs) embedded in active galactic nucleus (AGN) discs offer a distinct dynamical channel to produce black hole mergers detected in gravitational waves by LIGO/Virgo. To understand their orbital evolution through interactions with the disc gas, we perform a suite of two-dimensional high-resolution, local shearing box, viscous hydrodynamical simulations of equal-mass binaries. We find that viscosity not only smooths the flow structure around prograde circular binaries,but also greatly raises their accretion rates. The torque associated with accretion may be overwhelmingly positive and dominate over the gravitational torque at a high accretion rate. However, the accreted angular momentum per unit mass decreases with increasing viscosity, making it easier to shrink the binary orbit. In addition, retrograde binaries still experience rapid orbital decay, and prograde eccentric binaries still experience eccentricity damping. Our numerical experiments further show that prograde binaries are more likely to be hardened if the physical sizes of the accretors are sufficiently small such that the accretion rate is reduced. The dependence of the binary accretion rate on the accretor size can be weaken through boosted accretion either due to a high viscosity or a more isothermal-like equation of state. Our results widen the explored parameter space for the hydrodynamics of embedded BBHs and demonstrate that their orbital evolution in AGN discs is a complex, multifaceted problem.

Models of long-term product innovation depict the trajectory of products through an evolutionary selection metaphor in which product designs converge toward a dominant design. The product innovation literature favors trajectory descriptions based on the physical architecture of products while neglecting to account for the functional architecture. This paper offers a new way to explain the life cycle of product innovation by identifying motifs that describe a product’s functions. Functional motifs are recurrent function blocks across multiple generations of designs for a product. A collection of functional motifs defines the functional architecture of the product. Using some key examples from innovations in sewing machines, the paper illustrates the occurrence of motifs as the basis for detecting the emergence of a dominant design. Patents related to the sewing machine over 177 years are analyzed to identify functional motifs characterizing the evolution and convergence toward a dominant design. Results show that motifs do not change over long periods once a dominant design emerges, even though components continue to change. This observation confirms a view of dominant designs as a technological frame but refutes the notion that design no longer matters in the era of incremental change. These motifs refine our understanding of how designs evolve along a particular path over the course of product innovation.