Search for: All records

We measure the performance of separately characterized machine learning-based EDFA models for predicting the optical power spectrum evolution in a 5-span system with six ROADM nodes deployed in the COSMOS testbed, which achieve a mean absolute error of 0.6–0.7 dB after 10 EDFAs under varying channel loading configurations. 

While current vision algorithms excel at many challenging tasks, it is unclear how well they understand the physical dynamics of real-world environments. Here we introduce Physion, a dataset and benchmark for rigorously evaluating the ability to predict how physical scenarios will evolve over time. Our dataset features realistic simulations of a wide range of physical phenomena, including rigid and soft-body collisions, stable multi-object configurations, rolling, sliding, and projectile motion, thus providing a more comprehensive challenge than previous benchmarks. We used Physion to benchmark a suite of models varying in their architecture, learning objective, input-output structure, and training data. In parallel, we obtained precise measurements of human prediction behavior on the same set of scenarios, allowing us to directly evaluate how well any model could approximate human behavior. We found that vision algorithms that learn object-centric representations generally outperform those that do not, yet still fall far short of human performance. On the other hand, graph neural networks with direct access to physical state information both perform substantially better and make predictions that are more similar to those made by humans. These results suggest that extracting physical representations of scenes is the main bottleneck to achieving human-level and human-like physical understanding in vision algorithms. We have publicly released all data and code to facilitate the use of Physion to benchmark additional models in a fully reproducible manner, enabling systematic evaluation of progress towards vision algorithms that understand physical environments as robustly as people do. 

We present an all-sky search for long-duration gravitational waves (GWs) from the first part of the LIGO-Virgo-KAGRA fourth observing run (O4), called O4a and comprising data taken between May 24, 2023, and January 16, 2024. The GW signals targeted by this search are the so-called “long-duration” ( $\gtrsim 1\mathrm{s}$) transients expected from a variety of astrophysical processes, including nonaxisymmetric deformations in magnetars or eccentric binary coalescences. We make minimal assumptions on the emitted GW waveforms in terms of morphologies and durations. Overall, our search targets signals with durations of $\sim 1–1000\mathrm{s}$and frequency content in the range 16–2048 Hz. In the absence of significant detections, we report the sensitivity limits of our search in terms of root-sum-square signal amplitude ( $h_{\mathrm{rss}}$) of reference waveforms. These limits improve upon the results from the third LIGO-Virgo-KAGRA observing run (O3) by about 30% on average. Moreover, this analysis demonstrates substantial progress in our ability to search for long-duration GW signals owing to enhancements in pipeline detection efficiencies. As detector sensitivities continue to advance and observational runs grow longer, unmodeled long-duration searches will increasingly be able to explore a range of compelling astrophysical scenarios involving neutron stars and black holes.