On Edge Reweighting for Link Prediction with Graph Auto-Encoders

<i>During the last five years, graph auto-encoders became popular unsupervised methods, based on graph neural networks, to learn a node embedding from a graph. Researchers train graph auto-encoders by optimizing reconstruction losses that are computed from the connected node pairs (the edges) and non-connected node pairs of the graph. Many graphs being sparse, researchers often positively reweight the edges in these reconstruction losses. In this paper, we report an analysis of the effect of edge reweighting on the node embedding. We show that, on a link prediction problem, results are quite insensitive to edge reweighting, with the exception of very unbalanced reconstruction losses. We also discuss whether training models from perfectly balanced reconstruction losses is optimal or suboptimal, in terms of average scores and of standard deviations.</i>

On Edge Reweighting for Link Prediction with Graph Auto-Encoders

<i>During the last five years, graph auto-encoders became popular unsupervised methods, based on graph neural networks, to learn a node embedding from a graph. Researchers train graph auto-encoders by optimizing reconstruction losses that are computed from the connected node pairs (the edges) and non-connected node pairs of the graph. Many graphs being sparse, researchers often positively reweight the edges in these reconstruction losses. In this paper, we report an analysis of the effect of edge reweighting on the node embedding. We show that, on a link prediction problem, results are quite insensitive to edge reweighting, with the exception of very unbalanced reconstruction losses. We also discuss whether training models from perfectly balanced reconstruction losses is optimal or suboptimal, in terms of average scores and of standard deviations.</i>

Three-Dimensional Triptycene-Based Covalent Organic Frameworks with ceq or acs Topology

<a>The growth of three-dimensional covalent organic frameworks (3D COFs) with new topologies is still considered as a great challenge due to limited availability of high-connectivity building units. Here we report the design and synthesis of novel 3D triptycene-based COFs, </a><a></a><a>termed</a> JUC-568 and JUC-569, following the deliberate symmetry-guided design principle. By combining a triangular prism (6-connected) node with a planar triangle (3-connected) or another triangular prism node, the targeted COFs adopt unreported <b>ceq </b>or non-interpenetrated <b>acs</b> topology, respectively. <a>Both materials</a> show permanent porosity and impressive performance <a>in the adsorption of CO₂</a> (~ 98 cm³/g at 273 K and 1 bar), CH₄ (~ 48 cm³/g at 273 K and 1 bar), and especially H₂ (up to 274 cm³/g or 2.45 wt% at 77 K and 1 bar), which is <a>highest </a>among <a>porous organic materials</a> reported to date. This research thus provides a promising strategy for diversifying 3D COFs based on complex building blocks and promotes their <a></a><a>potential applications</a> <a>in</a><a></a><a> energy storage and environment-related field</a>s.

Three-Dimensional Triptycene-Based Covalent Organic Frameworks with ceq or acs Topology

<a>The growth of three-dimensional covalent organic frameworks (3D COFs) with new topologies is still considered as a great challenge due to limited availability of high-connectivity building units. Here we report the design and synthesis of novel 3D triptycene-based COFs, </a><a></a><a>termed</a> JUC-568 and JUC-569, following the deliberate symmetry-guided design principle. By combining a triangular prism (6-connected) node with a planar triangle (3-connected) or another triangular prism node, the targeted COFs adopt unreported <b>ceq </b>or non-interpenetrated <b>acs</b> topology, respectively. <a>Both materials</a> show permanent porosity and impressive performance <a>in the adsorption of CO₂</a> (~ 98 cm³/g at 273 K and 1 bar), CH₄ (~ 48 cm³/g at 273 K and 1 bar), and especially H₂ (up to 274 cm³/g or 2.45 wt% at 77 K and 1 bar), which is <a>highest </a>among <a>porous organic materials</a> reported to date. This research thus provides a promising strategy for diversifying 3D COFs based on complex building blocks and promotes their <a></a><a>potential applications</a> <a>in</a><a></a><a> energy storage and environment-related field</a>s.

On Edge Reweighting for Link Prediction with Graph Auto-Encoders

<i>During the last five years, graph auto-encoders became popular unsupervised methods, based on graph neural networks, to learn a node embedding from a graph. Researchers train graph auto-encoders by optimizing reconstruction losses that are computed from the connected node pairs (the edges) and non-connected node pairs of the graph. Many graphs being sparse, researchers often positively reweight the edges in these reconstruction losses. In this paper, we report an analysis of the effect of edge reweighting on the node embedding. We show that, on a link prediction problem, results are quite insensitive to edge reweighting, with the exception of very unbalanced reconstruction losses. We also discuss whether training models from perfectly balanced reconstruction losses is optimal or suboptimal, in terms of average scores and of standard deviations.</i>

The effect of functionally-guided-connectivity-based rTMS on amygdala activation

Repetitive transcranial magnetic stimulation (rTMS) has fundamentally transformed how we treat psychiatric disorders, but is still in need of innovation to optimally correct dysregulation that occurs throughout the fronto-limbic network. rTMS is often applied over the prefrontal cortex, a central node in this network, but less attention is given to subcortical areas because they lie at depths beyond the electric field penetration of rTMS. Recent studies have demonstrated that the effectiveness of rTMS is dependent on the functional connectivity between deep subcortical areas and superficial targets, indicating that leveraging such connectivity may improve dosing approaches for rTMS interventions. The current preliminary study, therefore, sought to test whether task-related, fMRI-connectivity-based rTMS could be used to modulate amygdala activation through its connectivity with the medial prefrontal cortex (mPFC). For this purpose, fMRI was collected on participants to identify a node in the mPFC that showed the strongest negative connectivity with right amygdala, as defined by psychophysiological interaction analysis. To promote long-lasting Hebbian-like effects, and potentially stronger modulation, 5Hz rTMS was then applied to this target as participants viewed frightening video-clips that engaged the fronto-limbic network. Post-rTMS fMRI results revealed promising increases in both the left mPFC and right amygdala, for active rTMS compared to sham. While these modulatory findings are promising, they differ from the a priori expectation that excitatory 5Hz rTMS over a negatively connected node would reduce amygdala activity. As such, further research is needed to better understand how connectivity influences TMS effects on distal structures, and to leverage this information to improve therapeutic applications.

A MOF based on a lead(II) 2-oxido-6-methylpyridine-4-carboxylate network

A new metal-organic framework, [Pb(hmpcaH)2]n (1), has been hydrothermally synthesized from Pb(OAc)2 · 3H2O and 2-hydroxy-6-methylpyridine-4-carboxylic acid (hmpcaH2; 2), and characterized by IR spectroscopy, elemental and thermogravimetric analysis, and single-crystal X-ray diffraction. In complex 1, each hmpcaH− ligand represents a three-connected node to combine with the hexacoordinated Pb(II) ions, generating a 3D binodal (3,6)-connected ant network. The crystal structure of 2 was determined. The solid-state fluorescence properties of 1 and 2 were investigated.

Computing the statistical significance of optimized communities in networks

In scientific problems involving systems that can be modeled as a network (or “graph”), it is often of interest to find network communities - strongly connected node subsets - for unsupervised learning, feature discovery, anomaly detection, or scientific study. The vast majority of community detection methods proceed via optimization of a quality function, which is possible even on random networks without communities. Therefore there is usually not an easy way to tell if a community is “significant”, in this context meaning more internally connected than would be expected under a random graph model without communities. This paper generalizes existing null models and statistical tests for this purpose to bipartite graphs, and introduces a new significance scoring algorithm called Fast Optimized Community Significance (FOCS) that is highly scalable and agnostic to the type of graph. Compared with existing methods on unipartite graphs, FOCS is more numerically stable and better balances the trade-off between detection power and false positives. On a large-scale bipartite graph derived from the Internet Movie Database (IMDB), the significance scores provided by FOCS correlate strongly with meaningful actor/director collaborations on serial cinematic projects.

A Versatile Porous Silver-Coordinated Material for the Heterogeneous Catalysis of Chemical Conversion with Propargylic Alcohols and CO2

The efficient transformation of carbon dioxide into useful chemical feedstock is of great significance, attracting intense research interest. The widely studied porous-coordinated polymers possess large pores to adsorb guest molecules and further allow the contact and to transfer the substrate molecule within their microenvironment. Here we present the synthesis of a silver-based metal-organic frameworks (MOFs) material with a three-dimensional structure by incorporating a tetraphenyl-ethylene moiety as the four-point connected node via the solvothermal method. This polymer exhibits as an efficient heterogeneous catalyst for the carboxylative cyclization of CO2 to α-methylene cyclic carbonates in excellent yields. Moreover, the introduction of silver (Ag (I)) chains in this framework shows the specific alkynophilicity to activate C≡C bonds in propargylic alcohols to greatly accelerate the efficient conversion, and the large pores in the catalyst exhibit a size-selective catalytic performance.