Search for: All records

Molecular Representation Learning (MRL) has proven impactful in numerous biochemical applications such as drug discovery and enzyme design. While Graph Neural Networks (GNNs) are effective at learning molecular representations from a 2D molecular graph or a single 3D structure, existing works often overlook the flexible nature of molecules, which continuously interconvert across conformations via chemical bond rotations and minor vibrational perturbations. To better account for molecular flexibility, some recent works formulate MRL as an ensemble learning problem, focusing on explicitly learning from a set of conformer structures. However, most of these studies have limited datasets, tasks, and models. In this work, we introduce the first MoleculAR Conformer Ensemble Learning (MARCEL) benchmark to thoroughly evaluate the potential of learning on con- former ensembles and suggest promising research directions. MARCEL includes four datasets covering diverse molecule- and reaction-level properties of chemically diverse molecules including organocatalysts and transition-metal catalysts, extending beyond the scope of common GNN benchmarks that are confined to drug-like molecules. In addition, we conduct a comprehensive empirical study, which benchmarks representative 1D, 2D, and 3D MRL models, along with two strategies that explicitly incorporate conformer ensembles into 3D models. Our findings reveal that direct learning from an accessible conformer space can improve performance on a variety of tasks and models. 

Recent multi-point measurements, in particular from the Magnetospheric Multiscale (MMS) spacecraft, have advanced the understanding of micro-scale aspects of magnetic reconnection. In addition, the MMS mission, as part of the Heliospheric System Observatory, combined with recent advances in global magnetospheric modeling, have furthered the understanding of meso- and global-scale structure and consequences of reconnection. Magnetic reconnection at the dayside magnetopause and in the magnetotail are the drivers of the global Dungey cycle, a classical picture of global magnetospheric circulation. Some recent advances in the global structure and consequences of reconnection that are addressed here include a detailed understanding of the location and steadiness of reconnection at the dayside magnetopause, the importance of multiple plasma sources in the global circulation, and reconnection consequences in the magnetotail. These advances notwithstanding, there are important questions about global reconnection that remain. These questions focus on how multiple reconnection and reconnection variability fit into and complicate the Dungey Cycle picture of global magnetospheric circulation.

 

While proving, and more broadly conceived “reasoning and sense-making,” have received a great deal of attention in mathematics education research over the past three decades, recently scholars have argued for the importance of justification as a learning and teaching practice. As teachers work toward realizing goals for more equitable classroom environments, little is known about whether teachers’ conceptions about mathematical practices, such as justification, reflect an understanding of how students’ engagement in those practices can support more than just mathematical achievement. In this paper, we present findings from our analysis of interviews with 10 secondary mathematics teachers engaged in participatory action research to explore connections, and potential disconnections, between teachers’ conceptions of justification and their visions for equitable instruction. 

As 6G wireless communications push the operation frequency above 110 GHz, it is critical to have low-loss interconnects that can be accurately tested. To this end, D-band (110 GHz to 170 GHz) substrate-integrated waveguides (SIWs) are designed on a 100-μm-thick SiC substrate. The fabricated SIWs are probed on-wafer in a single sweep from 70 kHz to 220 GHz with their input/output transitioned to grounded coplanar waveguides (GCPWs). From CPW-probed scattering parameters, two-tier calibration is used to de-embed the SIW-GCPW transitions and to extract the intrinsic SIW characteristics. In general, the record low loss measured agrees with that obtained from finite-element full-wave electromagnetic simulation. For example, across the D band, the average insertion loss is approximately 0.2 dB/mm, which is several times better than that of coplanar or microstrip transmission lines fabricated on the same substrate. A 3-pole filter exhibits a 1-dB insertion loss at 135 GHz with 20-dB selectivity and 11% bandwidth, which is order-of-magnitude better than typical on-chip filters. These results underscore the potential of using SIWs to interconnect transistors, filters, antennas, and other circuit elements on the same monolithically integrated chip. 

A D-band (110‒170 GHz) SiC substrate-integrated waveguide (SIW) is characterized on-wafer by two different vector network analyzers (VNAs): a 220-GHz single-sweep VNA and an 110-GHz VNA with WR8 (90‒140 GHz) and WR5 (140‒220 GHz) frequency extenders. To facilitate probing, the SIW input and output are transitioned to grounded coplanar waveguides (GCPWs). Two-tier calibration is used to de-embed the SIW-GCPW transitions as well as to extract the intrinsic SIW characteristics. In general, the two VNAs are in agreement and both result in an ultra-low insertion loss of approximately 0.2 dB/mm for the same SIW, despite stitching errors at band edges.