Search for: All records

Discrete representation learning has shown promising results across various domains, including generation and understanding in image, speech and language. Inspired by these advances, we propose MuseTok, a tokenization method for symbolic music, and investigate its effectiveness in both music generation and understanding tasks. MuseTok employs the residual vector quantized-variational autoencoder (RQ-VAE) on bar-wise music segments within a Transformerbased encoder-decoder framework, producing music codes that achieve high-fidelity music reconstruction and accurate understanding of music theory. For comprehensive evaluation, we apply MuseTok to music generation and semantic understanding tasks, including melody extraction, chord recognition, and emotion recognition. Models incorporating MuseTok outperform previous representation learning baselines in semantic understanding while maintaining comparable performance in content generation. Furthermore, qualitative analyses on MuseTok codes, using ground-truth categories and synthetic datasets, reveal that MuseTok effectively captures underlying musical concepts from large music collections. 

Abstract MotivationSeeding is an essential preparatory step for many fundamental computational tasks that require large-scale sequence comparison. Substring-based seeding methods such as kmers are ideal for sequences with low error rates but struggle to achieve high sensitivity while maintaining a reasonable precision for error-prone long reads. SubseqHash, a novel subsequence-based seeding method we recently developed, achieves superior accuracy to substring-based methods in seeding sequences with high mutation/error rates, while the only drawback is its computation speed. ResultsWe propose SubseqHash2, an improved algorithm that can compute multiple sets of seeds in one run, by defining k orders over all length-k subsequences and finding the optimal subsequence under each of the k orders in a single dynamic programming framework. The algorithm is further accelerated using single instruction, multiple data instructions for parallel computing. The design of SubseqHash2 also allows it to generate the same sets of seeds for a string and its reverse complement by using symmetric random tables. We demonstrate that SubseqHash2 drastically outperforms popular substring-based methods including kmers, minimizers, syncmers, and Strobemers for three fundamental applications. In read mapping, SubseqHash2 can generate adequate seed matches for aligning hard reads that minimap2 fails on. In sequence alignment, SubseqHash2 achieves high coverage of correct seeds and low coverage of incorrect seeds. In overlap detection, seeds produced by SubseqHash2 lead to more correct overlapping pairs at the same false-positive rate. In all experiments, SubseqHash2 achieves a 10–50× speedup over SubseqHash while maintaining nearly identical high accuracy. With all the algorithmic breakthroughs of SubseqHash2, we clear the path for the wide adoption of subsequence-based seeds in long-read analysis. Availability and implementationSubseqHash2 is available at https://github.com/Shao-Group/SubseqHash2 and have also been archived on Software Heritage (swh:1:dir:86738fc4b919eb6a9a26f7f533c25eb69f9a96d5). 

Abstract Coastal oceans, particularly the Northwest Atlantic continental shelf, are experiencing profound changes driven by global variability and local processes. These changes carry significant implications for marine ecosystems, fisheries, and coastal communities, underscoring the importance of understanding and predicting coastal variability. Koul et al. (2024),https://doi.org/10.1029/2024gl110946contribute to this effort by utilizing a high‐resolution regional ocean model downscaled from a global prediction system to forecast a pause in the rapid warming of the Northwest Atlantic Shelf in the coming decade. This prediction, attributed to internal climate variability linked to a strengthening of Atlantic Meridional Overturning Circulation (AMOC) and a southward Gulf Stream shift, highlights the potential of regional models to improve predictive skill. While the study provides an important contribution, further research is needed to explore alternative modeling frameworks and improve understanding of key processes. Long‐term sustained observational programs remain critical for validating models and uncovering processes, enhancing prediction reliability. 

Abstract The monthly mean sea level along the U.S. Mid‐Atlantic Coast varies seasonally, reaching a minimum in January and a maximum in September during the 1960–2020 period. However, this seasonal cycle has changed significantly on multi‐decadal timescales. In the last two decades, the annual minimum has shifted from January to February. The amplitude of seasonal changes increased by 65% from 14.16 cm in 1980–1999 to 23.16 cm in 2000–2020. Even more concerning, the maximum sea level in September rose by 82%, from 6.81 to 12.38 cm, potentially exacerbating coastal flooding over the past 20 years. A two‐layer ocean model effectively replicates both the phase and magnitude of the observed changes and attributes these shifts to changes in wind stress near the coast, with relatively minor influence from deep ocean forcing. Both alongshore and cross‐shore wind stress changes are found to contribute to changes in the sea level's seasonal cycle. 

Sequence similarity estimation is essential for many bioinformatics tasks, including functional annotation, phylogenetic analysis, and overlap graph construction. Alignment-free methods aim to solve large-scale sequence similarity estimation by mapping sequences to more easily comparable features that can approximate edit distances efficiently. Substrings or k-mers, as the dominant choice of features, face an unavoidable compromise between sensitivity and specificity when selecting the proper k-value. Recently, subsequence-based features have shown improved performance, but they are computationally demanding, and determining the ideal subsequence length remains an intricate art. In this work, we introduce SubseqSketch, a novel alignment-free scheme that maps a sequence to an integer vector, where the entries correspond to dynamic, rather than fixed, lengths of random subsequences. The cosine similarity between these vectors exhibits a strong correlation with the edit similarity between the original sequences. Through experiments on benchmark datasets, we demonstrate that SubseqSketch is both efficient and effective across various alignment-free tasks, including nearest neighbor search and phylogenetic clustering. A C++ implementation of SubseqSketch is openly available at https://github.com/Shao-Group/SubseqSketch. 

We have investigated the terahertz (THz) nonlinear response of the multigap superconductor MgB2, using THz two-dimensional coherent spectroscopy (THz 2DCS). With broadband THz drive fields, we identified a nonlinear response at twice the lower superconducting gap energy 2⁢Δ𝜋 at the lowest temperatures. Using narrow-band THz driving pulses, we observed first (FH) and third harmonic responses. The FH intensity shows a monotonic increase with decreasing temperature when properly normalized by the driving field strength. This is distinct from the single-gap superconductor NbN, where the FH signal exhibited a resonant enhancement at temperatures when twice the gap energy 2⁢Δ was resonant with the driving photon energy, which was interpreted to originate from the superconducting amplitude mode. Our results in MgB2 are consistent with a well-defined amplitude mode only at the lowest temperatures and indicate strong damping as temperature increases. This likely indicates the importance of interband coupling in MgB2 and its influence on the nature of the amplitude mode and its damping 

Abstract Understanding the occurrence of the intrusion of open ocean water onto continental shelves has scientific significance and societal relevance as the intrusion can significantly disrupt the marine ecosystem and fisheries. High-resolution numerical modeling is used to investigate the spatiotemporal occurrence and mechanisms of highly anomalous bottom intrusions on the southern New England shelf. Based on multi-year numerical simulations, this study reveals a hotspot of cross-isobath bottom-intensified intrusions at a topographic trough. Examination of multiple events portrays a robust mechanism of locally enhanced bottom intrusions. Persistent upwelling-favorable winds set up an enhanced pressure gradient field at the topographic trough and drive the intrusion a large-distance onshore. Numerical experiments with and without the topographic trough show that the localized pressure gradient results from a combination of the shelf orientation and local bathymetry. Although highly anomalous waters on the shelf relate to wind forcing, correlations between the wind stress anomaly and bottom salinity anomaly at the location of the enhanced intrusion is modest, implying the need to incorporate other environmental factors to develop more deterministic prediction models for subsurface conditions on the shelf. The results have important implications for marine environment and fisheries management.