Search

Cadences are complex structures that have been driving music from the beginning of contrapuntal polyphony until today. Detecting such structures is vital for numerous MIR tasks such as musicological analysis, key detection, or music segmentation. However, automatic cadence detection remains challenging mainly because it involves a combination of high-level musical elements like harmony, voice leading, and rhythm. In this work, we present a graph representation of symbolic scores as an intermediate means to solve the cadence detection task. We approach cadence detection as an imbalanced node classification problem using a Graph Convolutional Network. We obtain results that are roughly on par with the state of the art, and we present a model capable of making predictions at multiple levels of granularity, from individual notes to beats, thanks to the fine-grained, note-by-note representation. Moreover, our experiments suggest that graph convolution can learn non-local features that assist in cadence detection, freeing us from the need of having to devise specialized features that encode non-local context. We argue that this general approach to modeling musical scores and classification tasks has a number of potential advantages, beyond the specific recognition task presented here., Comment: In proceedings of the International Society for Music Information Retrieval Conference 2022 (ISMIR)

Current approaches for explaining deep learning systems applied to musical data provide results in a low-level feature space, e.g., by highlighting potentially relevant time-frequency bins in a spectrogram or time-pitch bins in a piano roll. This can be difficult to understand, particularly for musicologists without technical knowledge. To address this issue, we focus on more human-friendly explanations based on high-level musical concepts. Our research targets trained systems (post-hoc explanations) and explores two approaches: a supervised one, where the user can define a musical concept and test if it is relevant to the system; and an unsupervised one, where musical excerpts containing relevant concepts are automatically selected and given to the user for interpretation. We demonstrate both techniques on an existing symbolic composer classification system, showcase their potential, and highlight their intrinsic limitations., Comment: In Proceedings of the 23rd International Society for Music Information Retrieval Conference (ISMIR 2022), Bengaluru, India

The absence of large labeled datasets remains a significant challenge in many application areas of deep learning. Researchers and practitioners typically resort to transfer learning and data augmentation to alleviate this issue. We study these strategies in the context of audio retrieval with natural language queries (Task 6b of the DCASE 2022 Challenge). Our proposed system uses pre-trained embedding models to project recordings and textual descriptions into a shared audio-caption space in which related examples from different modalities are close. We employ various data augmentation techniques on audio and text inputs and systematically tune their corresponding hyperparameters with sequential model-based optimization. Our results show that the used augmentations strategies reduce overfitting and improve retrieval performance., Comment: accepted at DCASE Workshop 2022

Standard machine learning models for tagging and classifying acoustic signals cannot handle classes that were not seen during training. Zero-Shot (ZS) learning overcomes this restriction by predicting classes based on adaptable class descriptions. This study sets out to investigate the effectiveness of self-attention-based audio embedding architectures for ZS learning. To this end, we compare the very recent patchout spectrogram transformer with two classic convolutional architectures. We evaluate these three architectures on three tasks and on three different benchmark datasets: general-purpose tagging on AudioSet, environmental sound classification on ESC-50, and instrument tagging on OpenMIC. Our results show that the self-attention-based embedding methods outperform both compared convolutional architectures in all of these settings. By designing training and test data accordingly, we observe that prediction performance suffers significantly when the `semantic distance' between training and new test classes is large, an effect that will deserve more detailed investigations., Comment: published in EUSIPCO 2022

Algorithms for finding the sink in Unique Sink Orientations (USOs) of the hypercube can be used to solve many algebraic and geometric problems, most importantly including the P-Matrix Linear Complementarity Problem and Linear Programming. The realizable USOs are those that arise from the reductions of these problems to the USO sink-finding problem. Finding the sink of realizable USOs is thus highly practically relevant, yet it is unknown whether realizability can be exploited algorithmically to find the sink more quickly. However, all (non-trivial) known unconditional lower bounds for sink-finding make use of USOs that are provably not realizable. This indicates that the sink-finding problem might indeed be strictly easier on realizable USOs. In this paper we show that this is true for a subclass of all USOs. We consider the class of Matou\v{s}ek-type USOs, which are a translation of Matou\v{s}ek's LP-type problems into the language of USOs. We show a query complexity gap between sink-finding in all, and sink-finding in only the realizable $n$-dimensional Matou\v{s}ek-type USOs. We provide concrete deterministic algorithms and lower bounds for both cases, and show that in the realizable case $O(log^2 n)$ vertex evaluation queries suffice, while in general exactly $n$ queries are needed. The Matou\v{s}ek-type USOs are the first USO class found to admit such a gap., Comment: 18 pages, 5 figures

Turbulent plasmas notably self-organize to higher energy states upon application of additional free energy sources or modification of edge operating conditions. Mechanisms whereby such bifurcations occur have been actively debated for decades. Enhanced confinement occurs at the plasma edge, where a shortfall of predicted turbulence intensity has been puzzling scientists for decades. We show, from the primitive kinetic equations that both problems are connected and that interplay of confined plasma turbulence with its material boundaries is essential to curing the shortfall of predicted turbulence and to triggering spontaneous transport barrier onset at the plasma edge. Both problems determine access to improved confinement and are central to fusion research. A comprehensive discussion of the underlying mechanisms is proposed. These results, highly relevant to the quest for magnetic fusion may also be generic to many problems in fluids and plasmas where turbulence self-advection is active.

This paper presents the specifications of match: a file format that extends a MIDI human performance with note-, beat-, and downbeat-level alignments to a corresponding musical score. This enables advanced analyses of the performance that are relevant for various tasks, such as expressive performance modeling, score following, music transcription, and performer classification. The match file includes a set of score-related descriptors that makes it usable also as a bare-bones score representation. For applications that require the use of structural score elements (e.g., voices, parts, beams, slurs), the match file can be easily combined with the symbolic score. To support the practical application of our work, we release a corrected and upgraded version of the Vienna4x22 dataset of scores and performances aligned with match files.

Partitura is a lightweight Python package for handling symbolic musical information. It provides easy access to features commonly used in music information retrieval tasks, like note arrays (lists of timed pitched events) and 2D piano roll matrices, as well as other score elements such as time and key signatures, performance directives, and repeat structures. Partitura can load musical scores (in MEI, MusicXML, Kern, and MIDI formats), MIDI performances, and score-to-performance alignments. The package includes some tools for music analysis, such as automatic pitch spelling, key signature identification, and voice separation. Partitura is an open-source project and is available at https://github.com/CPJKU/partitura/.

Adversarial attacks can drastically degrade performance of recommenders and other machine learning systems, resulting in an increased demand for defence mechanisms. We present a new line of defence against attacks which exploit a vulnerability of recommenders that operate in high dimensional data spaces (the so-called hubness problem). We use a global data scaling method, namely Mutual Proximity (MP), to defend a real-world music recommender which previously was susceptible to attacks that inflated the number of times a particular song was recommended. We find that using MP as a defence greatly increases robustness of the recommender against a range of attacks, with success rates of attacks around 44% (before defence) dropping to less than 6% (after defence). Additionally, adversarial examples still able to fool the defended system do so at the price of noticeably lower audio quality as shown by a decreased average SNR., Comment: 6 pages, to be published in Proceedings of the 19th Sound and Music Computing Conference 2022 (SMC-22)

A well established method to detect and classify human movements using Millimeter-Wave ( mmWave) devices is the time-frequency analysis of the small-scale Doppler effect (termed micro-Doppler) of the different body parts, which requires a regularly spaced and dense sampling of the Channel Impulse Response ( CIR). This is currently done in the literature either using special-purpose radar sensors, or interrupting communications to transmit dedicated sensing waveforms, entailing high overhead and channel utilization. In this work we present SPARCS, an integrated human sensing and communication solution for mmWave systems. SPARCS is the first method that reconstructs high quality signatures of human movement from irregular and sparse CIR samples, such as the ones obtained during communication traffic patterns. To accomplish this, we formulate the micro-Doppler extraction as a sparse recovery problem, which is critical to enable a smooth integration between communication and sensing. Moreover, if needed, our system can seamlessly inject short CIR estimation fields into the channel whenever communication traffic is absent or insufficient for the micro-Doppler extraction. SPARCS effectively leverages the intrinsic sparsity of the mmWave channel, thus drastically reducing the sensing overhead with respect to available approaches. We implemented SPARCS on an IEEE 802.11ay Software Defined Radio (SDR) platform working in the 60 GHz band, collecting standard-compliant CIR traces matching the traffic patterns of real WiFi access points. Our results show that the micro-Doppler signatures obtained by SPARCS enable a typical downstream application such as human activity recognition with more than 7 times lower overhead with respect to existing methods, while achieving better recognition performance., Comment: 13 pages, 9 figures, 2 tables

The development and approval of new treatments generates large volumes of results, such as summaries of efficacy and safety. However, it is commonly overlooked that analyzing clinical study data also produces data in the form of results. For example, descriptive statistics and model predictions are data. Although integrating and putting findings into context is a cornerstone of scientific work, analysis results are often neglected as a data source. Results end up stored as "data products" such as PDF documents that are not machine readable or amenable to future analysis. We propose a solution to "calculate once, use many times" by combining analysis results standards with a common data model. This analysis results data model re-frames the target of analyses from static representations of the results (e.g., tables and figures) to a data model with applications in various contexts, including knowledge discovery. Further, we provide a working proof of concept detailing how to approach analyses standardization and construct a schema to store and query analysis results., Comment: 11 pages, 4 figures, 1 table, submitted to the journal Scientific Data

Global simulations of electromagnetic turbulence, collisionless tearing modes, and Alfven Eigenmodes in the presence of fast particles are carried out using the gyrokinetic particle-in-cell codes ORB5 (E. Lanti et al, Comp. Phys. Comm., ${\bf 251}$, 107072 (2020)) and EUTERPE (V. Kornilov et al, Phys. Plasmas, ${\bf 11}$, 3196 (2004)) in tokamak and stellarator geometries. Computational feasibility of simulating such complex coupled systems is demonstrated.

This paper examines the diffusion of media slant, specifically how partisan content from national cable news affects local newspapers in the U.S., 2005-2008. We use a text-based measure of cable news slant trained on content from Fox News Channel (FNC), CNN, and MSNBC to analyze how local newspapers adopt FNC's slant over CNN/MSNBC's. Our findings show that local news becomes more similar to FNC content in response to an exogenous increase in local FNC viewership. This shift is not limited to borrowing from cable news, but rather, local newspapers' own content changes. Further, cable TV slant polarizes local news content.

The electronic, optical and magnetic properties of graphene nanoribbons (GNRs) can be engineered by controlling their edge structure and width with atomic precision through bottom-up fabrication based on molecular precursors. This approach offers a unique platform for all-carbon electronic devices but requires careful optimization of the growth conditions to match structural requirements for successful device integration, with GNR length being the most critical parameter. In this work, we study the growth, characterization, and device integration of 5-atom wide armchair GNRs (5-AGNRs), which are expected to have an optimal band gap as active material in switching devices. 5-AGNRs are obtained via on-surface synthesis under ultra-high vacuum conditions from Br- and I-substituted precursors. We show that the use of I-substituted precursors and the optimization of the initial precursor coverage quintupled the average 5-AGNR length. This significant length increase allowed us to integrate 5-AGNRs into devices and to realize the first field-effect transistor based on narrow bandgap AGNRs that shows switching behavior at room temperature. Our study highlights that optimized growth protocols can successfully bridge between the sub-nanometer scale, where atomic precision is needed to control the electronic properties, and the scale of tens of nanometers relevant for successful device integration of GNRs.

Primordial black holes, with masses comparable to asteroids, are an attractive possibility for dark matter. In addition, other forms of dark matter could form compact dark objects (CDO). We search for small tidal accelerations from low mass black holes or CDOs orbiting near the Earth, and find none. Using about 10 years of data from the superconducting gravimeters in the Black Forest Observatory in South-Western Germany and at Djougou, Northern Benin in Western Africa we set an upper limit on the maximum mass of any dark object orbiting the Earth as a function of orbital radius. For semi-major axis less than two earth radii we exclude all black holes or CDOs with masses larger than 6.7x10^{13} kg. Lower mass primordial black holes may be strongly constrained by Hawking radiation. We conclude that near Earth black holes are extremely unlikely., Comment: six pages, 6 figures