ABSTRACT
The reliable prediction of the temporal behavior of complex systems is key in numerous scientific fields. This strong interest is however hindered by modeling issues: Often, the governing equations describing the physics of the system under consideration are not accessible or, when known, their solution might require a computational time incompatible with the prediction time constraints. Not surprisingly, approximating complex systems in a generic functional format and informing it ex-nihilo from available observations has become common practice in the age of machine learning, as illustrated by the numerous successful examples based on deep neural networks. However, generalizability of the models, margins of guarantee and the impact of data are often overlooked or examined mainly by relying on prior knowledge of the physics. We tackle these issues from a different viewpoint, by adopting a curriculum learning strategy. In curriculum learning, the dataset is structured such that the training process starts from simple samples toward more complex ones in order to favor convergence and generalization. The concept has been developed and successfully applied in robotics and control of systems. Here, we apply this concept for the learning of complex dynamical systems in a systematic way. First, leveraging insights from the ergodic theory, we assess the amount of data sufficient for a-priori guaranteeing a faithful model of the physical system and thoroughly investigate the impact of the training set and its structure on the quality of long-term predictions. Based on that, we consider entropy as a metric of complexity of the dataset; we show how an informed design of the training set based on the analysis of the entropy significantly improves the resulting models in terms of generalizability and provide insights on the amount and the choice of data required for an effective data-driven modeling.
ABSTRACT
This study focuses on prosodic evolution in the French news announcer style, based on acoustic and perceptual analysis of French audiovisual archives. A 10-hour corpus covering six decades of broadcast news is investigated automatically. Two prosodic features, which may give an impression of emphatic style, are explored: word-initial stress and penultimate vowel lengthening, especially before a pause. Objective measurements suggest that the following features have decreased since the 40s: mean pitch, pitch rise associated with initial stress, vowel duration characterizing an emphatic initial stress, and prepausal penultimate lengthening. The onsets of stressed initial syllables have become longer while speech rate (measured at the phonemic level) has not changed. This puzzling outcome raises interesting questions for research on French prosody, suggesting that the durational correlates of word-initial stress have changed over time, in the French news announcer style. Three perceptual experiments were conducted using prosody transplantation (copy of fundamental frequency and duration parameters on a synthetic voice), delexicalization and imitation. Rather than manipulating the parameters of,say, word-initial stress, we selected a subset of the corpus to represent the different decades under investigation. Results show that, among other factors, fundamental frequency and duration correlates of prosody contribute to distinguishing early recordings from more recent ones.The higher the pitch and the greater the pitch movements associated with word-initial stress, the more the speech samples are perceived as dating back to the 40s or 50s.
