Search

Current instruction-tuned language models are exclusively trained with textual preference data and thus are often not aligned with the unique requirements of other modalities, such as speech. To better align language models with the speech domain, we explore (i) prompting strategies grounded in radio-industry best practices and (ii) preference learning using a novel speech-based preference data of 20K samples, generated with a wide spectrum of prompts that induce varying dimensions of speech-suitability and labeled by annotators who listen to response pairs. Both human and automatic evaluation show that both prompting and preference learning increase the speech-suitability of popular instruction-tuned LLMs. Interestingly, we find that prompting and preference learning can be additive; combining them achieves the best win rates in head-to-head comparison, resulting in responses that are preferred or tied to the base model in 76.2% of comparisons on average. Lastly, we share lexical, syntactical, and qualitative analyses to showcase how each method contributes to improving the speech-suitability of generated responses., Comment: EMNLP2024

The progenitor system(s) as well as the explosion mechanism(s) of thermonuclear (Type Ia) supernovae are long-standing issues in astrophysics. Here we present ejecta masses and other physical parameters for 28 recent Type Ia supernovae inferred from multiband photometric and optical spectroscopic data. Our results confirm that the majority of SNe Ia show {\it observable} ejecta masses below the Chandrasekhar-limit (having a mean $M_{\rm ej} \approx 1.1 \pm 0.3$ M$_\odot$), consistent with the predictions of recent sub-M$_{\rm Ch}$ explosion models. They are compatible with models assuming either single- or double-degenerate progenitor configurations. We also recover a sub-sample of supernovae within $1.2 $ M$_\odot$ $< M_{\rm {ej}} < 1.5$ M$_\odot$ that are consistent with near-Chandrasekhar explosions. Taking into account the uncertainties of the inferred ejecta masses, about half of our SNe are compatible with both explosion models. We compare our results with those in previous studies, and discuss the caveats and concerns regarding the applied methodology.

Converting input symbols to output audio in TTS requires modelling the durations of speech sounds. Leading non-autoregressive (NAR) TTS models treat duration modelling as a regression problem. The same utterance is then spoken with identical timings every time, unlike when a human speaks. Probabilistic models of duration have been proposed, but there is mixed evidence of their benefits. However, prior studies generally only consider speech read aloud, and ignore spontaneous speech, despite the latter being both a more common and a more variable mode of speaking. We compare the effect of conventional deterministic duration modelling to durations sampled from a powerful probabilistic model based on conditional flow matching (OT-CFM), in three different NAR TTS approaches: regression-based, deep generative, and end-to-end. Across four different corpora, stochastic duration modelling improves probabilistic NAR TTS approaches, especially for spontaneous speech. Please see https://shivammehta25.github.io/prob_dur/ for audio and resources., Comment: 5 pages, 2 figures. Final version, accepted to Interspeech 2024

There is a growing number of peculiar events that cannot be assigned to any of the main supernova (SN) classes. SN 1987A and a handful of similar objects, thought to be explosive outcomes of blue supergiant stars, belong to them: while their spectra closely resemble those of H-rich (IIP) SNe, their light-curve (LC) evolution is very different. Here we present the detailed photometric and spectroscopic analysis of SN 2021aatd, a peculiar Type II explosion: while its early-time evolution resembles that of the slowly evolving, double-peaked SN 2020faa (however, at a lower luminosity scale), after $\sim$40 days, its LC shape becomes similar to that of SN 1987A-like explosions. Beyond comparing LCs, color curves, and spectra of SN 2021aatd to that of SNe 2020faa, 1987A, and of other objects, we compare the observed spectra with our own SYN++ models and with the outputs of published radiative transfer models. We also modeled the pseudo-bolometric LCs of SNe 2021aatd and 1987A assuming a two-component (core+shell) ejecta, and involving the rotational energy of a newborn magnetar in addition to radioactive decay. We find that both the photometric and spectroscopic evolution of SN 2021aatd can be well described with the explosion of a $\sim$15 $M_\odot$ blue supergiant star. Nevertheless, SN 2021aatd shows higher temperatures and weaker Na ID and Ba II 6142 A lines than SN 1987A, which is reminiscent of rather to IIP-like atmospheres. With the applied two-component ejecta model (counting with both decay and magnetar energy), we can successfully describe the bolometric LC of SN 2021aatd, including the first $\sim$40-day long phase showing an excess compared to 87A-like SNe but being strikingly similar to that of the long-lived SN 2020faa. Nevertheless, finding a unified model that also explains the LCs of more luminous events (like SN 2020faa) is still a matter of concern., Comment: 24 pages, 18 figures, 10 tables; accepted for publication in A&A

Recent advances in generative language modeling applied to discrete speech tokens presented a new avenue for text-to-speech (TTS) synthesis. These speech language models (SLMs), similarly to their textual counterparts, are scalable, probabilistic, and context-aware. While they can produce diverse and natural outputs, they sometimes face issues such as unintelligibility and the inclusion of non-speech noises or hallucination. As the adoption of this innovative paradigm in speech synthesis increases, there is a clear need for an in-depth evaluation of its capabilities and limitations. In this paper, we evaluate TTS from a discrete token-based SLM, through both automatic metrics and listening tests. We examine five key dimensions: speaking style, intelligibility, speaker consistency, prosodic variation, spontaneous behaviour. Our results highlight the model's strength in generating varied prosody and spontaneous outputs. It is also rated higher in naturalness and context appropriateness in listening tests compared to a conventional TTS. However, the model's performance in intelligibility and speaker consistency lags behind traditional TTS. Additionally, we show that increasing the scale of SLMs offers a modest boost in robustness. Our findings aim to serve as a benchmark for future advancements in generative SLMs for speech synthesis., Comment: 11 pages, 4 figures. Language Resources and Evaluation Conference (LREC) 2024. demo: https://swatsw.github.io/lrec24_eval_slm/

In this paper three unrelated problems will be discussed. What connects them is the rich methodology of classical probability theory. In the first two problems we have a complete answer to the problems raised; in the third case, what we call the Hungarian roulette problem, we only have a conjecture with heuristic justification., Comment: 13 pages, 1 figure. To appear in Annales Univ. Sci. Budapest., Sec. Comp., vol. 56

Non-verbal signals in speech are encoded by prosody and carry information that ranges from conversation action to attitude and emotion. Despite its importance, the principles that govern prosodic structure are not yet adequately understood. This paper offers an analytical schema and a technological proof-of-concept for the categorization of prosodic signals and their association with meaning. The schema interprets surface-representations of multi-layered prosodic events. As a first step towards implementation, we present a classification process that disentangles prosodic phenomena of three orders. It relies on fine-tuning a pre-trained speech recognition model, enabling the simultaneous multi-class/multi-label detection. It generalizes over a large variety of spontaneous data, performing on a par with, or superior to, human annotation. In addition to a standardized formalization of prosody, disentangling prosodic patterns can direct a theory of communication and speech organization. A welcome by-product is an interpretation of prosody that will enhance speech- and language-related technologies.

We introduce the Visual Experience Dataset (VEDB), a compilation of over 240 hours of egocentric video combined with gaze- and head-tracking data that offers an unprecedented view of the visual world as experienced by human observers. The dataset consists of 717 sessions, recorded by 58 observers ranging from 6-49 years old. This paper outlines the data collection, processing, and labeling protocols undertaken to ensure a representative sample and discusses the potential sources of error or bias within the dataset. The VEDB's potential applications are vast, including improving gaze tracking methodologies, assessing spatiotemporal image statistics, and refining deep neural networks for scene and activity recognition. The VEDB is accessible through established open science platforms and is intended to be a living dataset with plans for expansion and community contributions. It is released with an emphasis on ethical considerations, such as participant privacy and the mitigation of potential biases. By providing a dataset grounded in real-world experiences and accompanied by extensive metadata and supporting code, the authors invite the research community to utilize and contribute to the VEDB, facilitating a richer understanding of visual perception and behavior in naturalistic settings., Comment: 40 pages, 1 table, 9 figures

The manual modeling of complex systems is a daunting task; and although a plethora of methods exist that mitigate this issue, the problem remains very difficult. Recent advances in generative AI have allowed the creation of general-purpose chatbots, capable of assisting software engineers in various modeling tasks. However, these chatbots are often inaccurate, and an unstructured use thereof could result in erroneous system models. In this paper, we outline a method for the safer and more structured use of chatbots as part of the modeling process. To streamline this integration, we propose leveraging scenario-based modeling techniques, which are known to facilitate the automated analysis of models. We argue that through iterative invocations of the chatbot and the manual and automatic inspection of the resulting models, a more accurate system model can eventually be obtained. We describe favorable preliminary results, which highlight the potential of this approach., Comment: This is a preprint version of a paper that will appear at Modelsward 2024

Methodologies for development of complex systems and models include external reviews by domain and technology experts. Among others, such reviews can uncover undocumented built-in assumptions that may be critical for correct and safe operation or constrain applicability. Since such assumptions may still escape human-centered processes like reviews, agile development, and risk analyses, here, we contribute toward making this process more methodical and automatable. We first present a blueprint for a taxonomy and formalization of the problem. We then show that a variety of digital artifacts of the system or model can be automatically checked against extensive reference knowledge. Since mimicking the breadth and depth of knowledge and skills of experts may appear unattainable, we illustrate the basic feasibility of automation with rudimentary experiments using OpenAI's ChatGPT. We believe that systematic handling of this aspect of system engineering can contribute significantly to the quality and safety of complex systems and models, and to the efficiency of development projects. We dedicate this work to Werner Damm, whose contributions to modeling and model-based development, in industry and academia, with a special focus on safety, helped establish a solid foundation to our discipline and to the work of many scientists and professionals, including, naturally, the approaches and techniques described here.

Commonly recognized evolutionarily relevant effects of sexual reproduction include increased diversity, accelerated adaptation, and constrained accumulation of deleterious mutations, along with a secondary effect of species genotype homogenization. Still, strong published arguments prioritize the contribution of biological mechanisms underlying bi-parental reproduction to maintaining species identity above their contribution to diversity. Here, we contribute to the latter position. In an initial mathematical analysis and simulation, we show that in an environment where copying is prone to error, quantitative polygenic traits that are shared within a parents' generation are transmitted to future generations under bi-parental reproduction with less deviation than under asexual reproduction. Furthermore, we abstract away many biological details, and show that this trait conservation is a general statistical effect, driven by the very nature of mixing of parental traits, separately from DNA repair and from the reproductive failures, barriers and disadvantages induced by biological mechanisms. Since survival of ecosystem interaction networks depends on the ability of individuals to replace the networked function of failing, dying or absent members of the same species, more faithful inheritance of common traits helps sustain species and ecosystems. This sustaining effect may have contributed to the very evolution of sexual reproduction., Comment: Relative to V4: Misc edits throughout. Added some citations. Changed article format

Neural networks excel at discovering statistical patterns in high-dimensional data sets. In practice, higher-order cumulants, which quantify the non-Gaussian correlations between three or more variables, are particularly important for the performance of neural networks. But how efficient are neural networks at extracting features from higher-order cumulants? We study this question in the spiked cumulant model, where the statistician needs to recover a privileged direction or "spike" from the order-$p\ge 4$ cumulants of $d$-dimensional inputs. Existing literature established the presence of a wide statistical-to-computational gap in this problem. We deepen this line of work by finding an exact formula for the likelihood ratio norm which proves that statistical distinguishability requires $n\gtrsim d$ samples, while distinguishing the two distributions in polynomial time requires $n \gtrsim d^2$ samples for a wide class of algorithms, i.e. those covered by the low-degree conjecture. Numerical experiments show that neural networks do indeed learn to distinguish the two distributions with quadratic sample complexity, while "lazy" methods like random features are not better than random guessing in this regime. Our results show that neural networks extract information from higher-ordercorrelations in the spiked cumulant model efficiently, and reveal a large gap in the amount of data required by neural networks and random features to learn from higher-order cumulants.

Let $S$ be a set of four points chosen independently, uniformly at random from a square. Join every pair of points of $S$ with a straight line segment. Color these edges red if they have positive slope and blue, otherwise. We show that the probability that $S$ defines a pair of crossing edges of the same color is equal to $1/4$. This is connected to a recent result of Aichholzer et al. [GD 2019] who showed that by 2-colouring the edges of a geometric graph and counting monochromatic crossings instead of crossings, the number of crossings can be more than halfed. Our result shows that for the described random drawings, there is a coloring of the edges such that the number of monochromatic crossings is in expectation $\frac{1}{2}-\frac{7}{50}$ of the total number of crossings.

We consider extremal problems related to decks and multidecks of rooted binary trees (a.k.a. rooted phylogenetic tree shapes). Here, the deck (resp. multideck) of a tree $T$ refers to the set (resp. multiset) of leaf induced binary subtrees of $T$. On the one hand, we consider the reconstruction of trees from their (multi)decks. We give lower and upper bounds on the minimum (multi)deck size required to uniquely encode a rooted binary tree on $n$ leaves. On the other hand, we consider problems related to deck cardinalities. In particular, we characterize trees with minimum-size as well as maximum-size decks. Finally, we present some exhaustive computations for $k$-universal trees, i.e., rooted binary trees that contain all $k$-leaf rooted binary trees as induced subtrees., Comment: 18 pages, 14 figures

As text-to-speech technologies achieve remarkable naturalness in read-aloud tasks, there is growing interest in multimodal synthesis of verbal and non-verbal communicative behaviour, such as spontaneous speech and associated body gestures. This paper presents a novel, unified architecture for jointly synthesising speech acoustics and skeleton-based 3D gesture motion from text, trained using optimal-transport conditional flow matching (OT-CFM). The proposed architecture is simpler than the previous state of the art, has a smaller memory footprint, and can capture the joint distribution of speech and gestures, generating both modalities together in one single process. The new training regime, meanwhile, enables better synthesis quality in much fewer steps (network evaluations) than before. Uni- and multimodal subjective tests demonstrate improved speech naturalness, gesture human-likeness, and cross-modal appropriateness compared to existing benchmarks. Please see https://shivammehta25.github.io/Match-TTSG/ for video examples and code., Comment: 5 pages, 1 figure. Final version, accepted to IEEE ICASSP 2024

We present extensive ultraviolet (UV) and optical photometric and optical spectroscopic follow-up of supernova (SN)~2021gno by the "Precision Observations of Infant Supernova Explosions" (POISE) project, starting less than two days after the explosion. Given its intermediate luminosity, fast photometric evolution, and quick transition to the nebular phase with spectra dominated by [Ca~II] lines, SN~2021gno belongs to the small family of Calcium-rich transients. Moreover, it shows double-peaked light curves, a phenomenon shared with only four other Calcium-rich events. The projected distance from the center of the host galaxy is not as large as other objects in this family. The initial optical light-curve peaks coincide with a very quick decline of the UV flux, indicating a fast initial cooling phase. Through hydrodynamical modelling of the bolometric light curve and line velocity evolution, we found that the observations are compatible with the explosion of a highly-stripped massive star with an ejecta mass of $0.8\,M_\odot$ and a $^{56}$Ni mass of $0.024~M_{\odot}$. The initial cooling phase (first light curve peak) is explained by the presence of an extended circumstellar material comprising $\sim$$10^{-2}\,M_{\odot}$ with an extension of $1100\,R_{\odot}$. We discuss if hydrogen features are present in both maximum-light and nebular spectra, and its implications in terms of the proposed progenitor scenarios for Calcium-rich transients., Comment: 21 pages, 13 figures, accepted for publication in MNRAS

We introduce Matcha-TTS, a new encoder-decoder architecture for speedy TTS acoustic modelling, trained using optimal-transport conditional flow matching (OT-CFM). This yields an ODE-based decoder capable of high output quality in fewer synthesis steps than models trained using score matching. Careful design choices additionally ensure each synthesis step is fast to run. The method is probabilistic, non-autoregressive, and learns to speak from scratch without external alignments. Compared to strong pre-trained baseline models, the Matcha-TTS system has the smallest memory footprint, rivals the speed of the fastest models on long utterances, and attains the highest mean opinion score in a listening test. Please see https://shivammehta25.github.io/Matcha-TTS/ for audio examples, code, and pre-trained models., Comment: 5 pages, 3 figures. Final version, accepted to IEEE ICASSP 2024

We present the photometric and spectroscopic evolution of SN 2022oqm, a nearby multi-peaked hydrogen- and helium-weak calcium-rich transient (CaRT). SN 2022oqm was detected 13.1 kpc from its host galaxy, the face-on spiral galaxy NGC 5875. Extensive spectroscopic coverage reveals an early hot (T >= 40,000 K) continuum and carbon features observed $\sim$1~day after discovery, SN Ic-like photospheric-phase spectra, and strong forbidden calcium emission starting 38 days after discovery. SN 2022oqm has a relatively high peak luminosity (MB = -17 mag) for (CaRTs), making it an outlier in the population. We determine that three power sources are necessary to explain the light curve (LC), with each corresponding to a distinct peak. The first peak is powered by an expanding blackbody with a power law luminosity, suggesting shock cooling by circumstellar material (CSM). Subsequent LC evolution is powered by a double radioactive decay model, consistent with two sources of photons diffusing through optically thick ejecta. From the LC, we derive an ejecta mass and 56Ni mass of ~0.6 solar masses and ~0.09 solar masses. Spectroscopic modeling suggests 0.6 solar masses of ejecta, and with well-mixed Fe-peak elements throughout. We discuss several physical origins for SN 2022oqm and find either a surprisingly massive white dwarf progenitor or a peculiar stripped envelope model could explain SN 2022oqm. A stripped envelope explosion inside a dense, hydrogen- and helium-poor CSM, akin to SNe Icn, but with a large 56Ni mass and small CSM mass could explain SN 2022oqm. Alternatively, helium detonation on an unexpectedly massive white dwarf could also explain SN 2022oqm., Comment: 35 pages, 17 figures, 7 tables, Accepted for Publication in ApJ

Our main goal is to examine the role of communities in epidemic spread in a random graph model. More precisely, we consider a random graph model which consists of overlapping complete graphs, representing households, workplaces, school classes, and which also has a simple geometric structure. We study the model's sensitivity to infection parameters and other tunable parameters of the model, which might be helpful in finding efficient social distancing strategies. We also quantitatively compare different vaccination strategies to see which order is the best to defend the most vulnerable groups or the population in general, and how important it is to gather and use information on the position of infected individuals in the network., Comment: 16 pages, 7 figures

We provide an $\Omega(n\log n) $ lower bound and an $O(n^2)$ upper bound for the smallest size of rooted binary trees (a.k.a. phylogenetic tree shapes), which are universal for rooted binary trees with $n$ leaves, i.e., contain all of them as induced binary subtrees. We explicitly compute the smallest universal trees for $n\leq 11$. We also provide an $\Omega(n^2) $ lower bound and an $O(n^4)$ upper bound for the smallest size of tanglegrams, which are universal for size $n$ tanglegrams, i.e., which contain all of them as induced subtanglegrams. Some of our results generalize to rooted $d$-ary trees and to $d$-ary tanglegrams., Comment: 18 pages, 14 figures

Self-supervised learning (SSL) speech representations learned from large amounts of diverse, mixed-quality speech data without transcriptions are gaining ground in many speech technology applications. Prior work has shown that SSL is an effective intermediate representation in two-stage text-to-speech (TTS) for both read and spontaneous speech. However, it is still not clear which SSL and which layer from each SSL model is most suited for spontaneous TTS. We address this shortcoming by extending the scope of comparison for SSL in spontaneous TTS to 6 different SSLs and 3 layers within each SSL. Furthermore, SSL has also shown potential in predicting the mean opinion scores (MOS) of synthesized speech, but this has only been done in read-speech MOS prediction. We extend an SSL-based MOS prediction framework previously developed for scoring read speech synthesis and evaluate its performance on synthesized spontaneous speech. All experiments are conducted twice on two different spontaneous corpora in order to find generalizable trends. Overall, we present comprehensive experimental results on the use of SSL in spontaneous TTS and MOS prediction to further quantify and understand how SSL can be used in spontaneous TTS. Audios samples: https://www.speech.kth.se/tts-demos/sp_ssl_tts, Comment: 7 pages, 2 figures. 12th ISCA Speech Synthesis Workshop (SSW) 2023

A tanglegram $\cal T$ consists of two rooted binary trees with the same number of leaves, and a perfect matching between the two leaf sets. In a layout, the tanglegrams is drawn with the leaves on two parallel lines, the trees on either side of the strip created by these lines are drawn as plane trees, and the perfect matching is drawn in straight line segments inside the strip. The tanglegram crossing number ${\rm cr}({\cal T})$ of $\cal T$ is the smallest number of crossings of pairs of matching edges, over all possible layouts of $\cal T$. The size of the tanglegram is the number of matching edges, say $n$. An earlier paper showed that the maximum of the tanglegram crossing number of size $n$ tanglegrams is $<\frac{1}{2}\binom{n}{2}$; but is at least $\frac{1}{2}\binom{n}{2}-\frac{n^{3/2}-n}{2}$ for infinitely many $n$. Now we make better bounds: the maximum crossing number of a size $n$ tanglegram is at most $ \frac{1}{2}\binom{n}{2}-\frac{n}{4}$, but for infinitely many $n$, at least $\frac{1}{2}\binom{n}{2}-\frac{n\log_2 n}{4}$. The problem shows analogy with the Unbalancing Lights Problem of Gale and Berlekamp.

The nearby spiral galaxy NGC 3147 hosted three Type Ia supernovae (SNe Ia) in the past decades, which have been subjects of intense follow-up observations. Simultaneous analysis of their data provides a unique opportunity for testing the different light curve fitting methods and distance estimations. The detailed optical follow-up of SN 2021hpr allows us to revise the previous distance estimations to NGC 3147, and compare the widely used light curve fitting algorithms to each other. After the combination of the available and newly published data of SN 2021hpr, its physical properties can be also estimated with higher accuracy. We present and analyse new BVgriz and Swift photometry of SN 2021hpr to constrain its general physical properties. Together with its siblings, SNe 1997bq and 2008fv, we cross-compare the individual distance estimates of these three SNe given by the SALT code, and also check their consistency with the results from the MLCS2k2 method. The early spectral series of SN 2021hpr are also fit with the radiative spectral code TARDIS in order to verify the explosion properties and constrain the chemical distribution of the outer ejecta. After combining the distance estimates for the three SNe, the mean distance to their host galaxy, NGC 3127, is 42.5 $\pm$ 1.0 Mpc, which matches with the distance inferred by the most up-to-date LC fitters, SALT3 and BayeSN. We confirm that SN~2021hpr is a Branch-normal Type Ia SN that ejected $\sim 1.12 \pm 0.28$ M$_\odot$ from its progenitor white dwarf, and synthesized $\sim 0.44 \pm 0.14$ M$_\odot$ of radioactive $^{56}$Ni., Comment: 16 pages, 17 figures, 11 tables; accepted for publication in A&A

With read-aloud speech synthesis achieving high naturalness scores, there is a growing research interest in synthesising spontaneous speech. However, human spontaneous face-to-face conversation has both spoken and non-verbal aspects (here, co-speech gestures). Only recently has research begun to explore the benefits of jointly synthesising these two modalities in a single system. The previous state of the art used non-probabilistic methods, which fail to capture the variability of human speech and motion, and risk producing oversmoothing artefacts and sub-optimal synthesis quality. We present the first diffusion-based probabilistic model, called Diff-TTSG, that jointly learns to synthesise speech and gestures together. Our method can be trained on small datasets from scratch. Furthermore, we describe a set of careful uni- and multi-modal subjective tests for evaluating integrated speech and gesture synthesis systems, and use them to validate our proposed approach. Please see https://shivammehta25.github.io/Diff-TTSG/ for video examples, data, and code., Comment: 7 pages, 2 figures, presented at the ISCA Speech Synthesis Workshop (SSW) 2023

Turn-taking is a fundamental aspect of human communication where speakers convey their intention to either hold, or yield, their turn through prosodic cues. Using the recently proposed Voice Activity Projection model, we propose an automatic evaluation approach to measure these aspects for conversational speech synthesis. We investigate the ability of three commercial, and two open-source, Text-To-Speech (TTS) systems ability to generate turn-taking cues over simulated turns. By varying the stimuli, or controlling the prosody, we analyze the models performances. We show that while commercial TTS largely provide appropriate cues, they often produce ambiguous signals, and that further improvements are possible. TTS, trained on read or spontaneous speech, produce strong turn-hold but weak turn-yield cues. We argue that this approach, that focus on functional aspects of interaction, provides a useful addition to other important speech metrics, such as intelligibility and naturalness., Comment: Accepted at INTERSPEECH 2023, 5 pages, 2 figures, 4 tables

Can we find a self-similar set on the line with positive Lebesgue measure and empty interior? Currently, we do not have the answer for this question for deterministic self-similar sets. In this paper we answer this question negatively for random self-similar sets which are defined with the construction introduced in the paper Jordan, Pollicott and Simon (Commun. Math. Phys., 2007). For the same type of random self-similar sets we prove the Palis-Takens conjecture which asserts that at least typically the algebraic difference of dynamically defined Cantor sets is either large in the sense that it contains an interval or small in the sense that it is a set of zero Lebesgue measure.

The National Football League (NFL) Scouting Combine serves as a tool to evaluate the skills of prospective players and assess their readiness to play in the NFL. The development of machine learning brings new opportunities in assessing the utility of the Scouting Combine. Using machine and statistical learning, it may be possible to predict future success of prospective athletes, as well as predict which Scouting Combine tests are the most important. Results from statistical learning research have been contradicting whether the Scouting combine is a useful metric for player success. In this study, we investigate if machine learning can be used to determine matriculation and future success in the NFL. Using Scouting Combine data, we evaluate six different algorithms' ability to predict whether a potential draft pick will play a single NFL snap (matriculation). If a player is drafted, we predict how many snaps they go on to play (success). We are able to predict matriculation with 83% accuracy; however, we are unable to predict later success. Our best performing algorithm returns large error and low explained variance (RMSE=1,210 snaps; ${R}^2$=0.17). These findings indicate that while the Scouting Combine can predict NFL matriculation, it may not be a reliable predictor of long-term player success., Comment: 16 pages, 2 figures, 1 table

Recent work has explored using self-supervised learning (SSL) speech representations such as wav2vec2.0 as the representation medium in standard two-stage TTS, in place of conventionally used mel-spectrograms. It is however unclear which speech SSL is the better fit for TTS, and whether or not the performance differs between read and spontaneous TTS, the later of which is arguably more challenging. This study aims at addressing these questions by testing several speech SSLs, including different layers of the same SSL, in two-stage TTS on both read and spontaneous corpora, while maintaining constant TTS model architecture and training settings. Results from listening tests show that the 9th layer of 12-layer wav2vec2.0 (ASR finetuned) outperforms other tested SSLs and mel-spectrogram, in both read and spontaneous TTS. Our work sheds light on both how speech SSL can readily improve current TTS systems, and how SSLs compare in the challenging generative task of TTS. Audio examples can be found at https://www.speech.kth.se/tts-demos/ssr_tts, Comment: 5 pages, 2 figures. ICASSP Workshop SASB (Self-Supervision in Audio, Speech and Beyond)2023

Fingerprints are key tools in climate change detection and attribution (D&A) that are used to determine whether changes in observations are different from internal climate variability (detection), and whether observed changes can be assigned to specific external drivers (attribution). We propose a direct D&A approach based on supervised learning to extract fingerprints that lead to robust predictions under relevant interventions on exogenous variables, i.e., climate drivers other than the target. We employ anchor regression, a distributionally-robust statistical learning method inspired by causal inference that extrapolates well to perturbed data under the interventions considered. The residuals from the prediction achieve either uncorrelatedness or mean independence with the exogenous variables, thus guaranteeing robustness. We define D&A as a unified hypothesis testing framework that relies on the same statistical model but uses different targets and test statistics. In the experiments, we first show that the CO2 forcing can be robustly predicted from temperature spatial patterns under strong interventions on the solar forcing. Second, we illustrate attribution to the greenhouse gases and aerosols while protecting against interventions on the aerosols and CO2 forcing, respectively. Our study shows that incorporating robustness constraints against relevant interventions may significantly benefit detection and attribution of climate change.