Continuous graph neural networks (CGNNs) have garnered significant attention
due to their ability to generalize existing discrete graph neural networks
(GNNs) by introducing continuous dynamics. They typically draw inspiration from
diffusion-based methods to introduce a novel propagation scheme, which is
analyzed using ordinary differential equations (ODE). However, the
implementation of CGNNs requires significant computational power, making them
challenging to deploy on battery-powered devices. Inspired by recent spiking
neural networks (SNNs), which emulate a biological inference process and
provide an energy-efficient neural architecture, we incorporate the SNNs with
CGNNs in a unified framework, named Continuous Spiking Graph Neural Networks
(COS-GNN). We employ SNNs for graph node representation at each time step,
which are further integrated into the ODE process along with time. To enhance
information preservation and mitigate information loss in SNNs, we introduce
the high-order structure of COS-GNN, which utilizes the second-order ODE for
spiking representation and continuous propagation. Moreover, we provide the
theoretical proof that COS-GNN effectively mitigates the issues of exploding
and vanishing gradients, enabling us to capture long-range dependencies between
nodes. Experimental results on graph-based learning tasks demonstrate the
effectiveness of the proposed COS-GNN over competitive baselines.

连续脉冲图神经网络（COS-GNN）是一个统一框架，将脉冲神经网络与连续图神经网络相结合，利用高阶结构和脉冲表示进行连续传播，以增强信息保留并捕捉节点之间的长程依赖。实验结果表明，COS-GNN 在图学习任务中比竞争对手具有更好的效果。

连续脉冲图神经网络

Continuous Spiking Graph Neural Networks

Thermodynamics and information have intricate inter-relations. The
justification of the fact that information is physical, is done by
inter-linking information and thermodynamics - through Landauer's principle.
This modern approach towards information recently has improved our
understanding of thermodynamics, both in classical and quantum domains. Here we
show thermodynamics as a consequence of information conservation. Our approach
can be applied to most general situations, where systems and thermal-baths
could be quantum, of arbitrary sizes and even could posses inter-system
correlations. The approach does not rely on an a priori predetermined
temperature associated to a thermal bath, which is not meaningful for
finite-size cases. Hence, the thermal-baths and systems are not different,
rather both are treated on an equal footing. This results in a
"temperature"-independent formulation of thermodynamics. We exploit the fact
that, for a fix amount of coarse-grained information, measured by the von
Neumann entropy, any system can be transformed to a state that possesses
minimal energy, without changing its entropy. This state is known as a
completely passive state, which assumes Boltzmann-Gibb's canonical form with an
intrinsic temperature. This leads us to introduce the notions of bound and free
energy, which we further use to quantify heat and work respectively. With this
guiding principle of information conservation, we develop universal notions of
equilibrium, heat and work, Landauer's principle and also universal fundamental
laws of thermodynamics. We show that the maximum efficiency of a quantum
engine, equipped with a finite baths, is in general lower than that of an ideal
Carnot's engine. We also introduce a resource theoretic framework for
intrinsic-temperature based thermodynamics, within which we address the problem
of work extraction and state transformations.

本文通过 Landauer 原理将热力学与信息学相互关联，从而提出信息保存作为热力学的基本原则，建立了温度无关的热力学基础理论，并通过该理论研究了量子引擎最大效率和状态转换等问题。

信息熵视角下的热力学

Thermodynamics from information

Statistical methods protecting sensitive information or the identity of the
data owner have become critical to ensure privacy of individuals as well as of
organizations. This paper investigates anonymization methods based on
representation learning and deep neural networks, and motivated by novel
information theoretical bounds. We introduce a novel training objective for
simultaneously training a predictor over target variables of interest (the
regular labels) while preventing an intermediate representation to be
predictive of the private labels. The architecture is based on three
sub-networks: one going from input to representation, one from representation
to predicted regular labels, and one from representation to predicted private
labels. The training procedure aims at learning representations that preserve
the relevant part of the information (about regular labels) while dismissing
information about the private labels which correspond to the identity of a
person. We demonstrate the success of this approach for two distinct
classification versus anonymization tasks (handwritten digits and sentiment
analysis).

本研究探讨基于表示学习和深度神经网络的匿名化方法，介绍了一种新的训练目标，同时训练一个预测器和防止中间表征成为私有标签的预测值。该三子网络的基架对应输入到表征、表征到预测的常规标签和表征到预测的私有标签。在保留有关常规标签的信息的同时，摒弃与身份有关的私人标签信息的学习表征，从而实现了有关分类与匿名化任务（手写数字和情感分析）的成功演示。