Abstract

Background Despite showing remarkable performance on tasks across modalities, deep neural networks remain opaque, with limited insight on how they internally organize information. A learned hierarchy is suggested in vision (Zhou et al., 2015) and language (Tenney et al., 2019) models in deep learning. Bregman (1990) describes the process of auditory scene analysis (ASA), where the auditory system abstracts information based on perceptual properties such as pitch and timbre, among others, in order to construct auditory streams that are then integrated to form higher order percepts like musical genre. This implies that the representation of audio goes through a series of hierarchical abstractions in the process. While hierarchical representations are indicated in human auditory processing (Kell et al., 2018), the same has not been thoroughly demonstrated in deep learning models for audio. Aims We study how convolutional neural networks (CNNs) represent the hierarchy of audio tasks, hypothesizing that representations at earlier layers of a CNN perform better at lower level tasks, and those at later layers perform better at higher level tasks. Methods On the basis of Bregman’s ASA model, we choose tasks that are arranged in a hierarchy from the domains of speech and music. We select three hierarchical musical tasks - note identification, instrument classification, and genre classification, representative of low-, mid-, and high-level tasks. We use class-balanced subsets of NSynth (Engel et al., 2017) with 1800 instances across 12 classes, Medley-solos-DB (Lostanlen et al., 2019) with 965 instances across 7 classes , and GTZAN (Tzanetakis et al., 2001) with 1000 instances across 10 classes, for each of the respective tasks. We inspect three models - VGGish (Hershey et al., 2017), CLAP (Elizalde et al., 2023), and MobileNetV3 (Schmid et al., 2023). We use k-Nearest Neighbour classifiers with class-balanced five-fold cross-validation to assess the model's accuracy on each task with intermediate representations extracted at six equally spaced convolutional blocks and the first fully-connected layer. We repeat this for three hierarchically related speech tasks - consonant classification, keyword recognition, and speaker count estimation, using similarly class-balanced subsets of PCVC (Malekzadeh et al., 2020) with 1794 instances across 23 classes, Speech Commands (Warden, 2018) with 1750 instances across 35 classes, LibriCount (Stöter et al., 2018) with 1100 instances across 11 classes respectively. Results Our results show that the CNN’s early layers learned low-level tasks and later layers learned high-level tasks even without explicit training. On CLAP, for note identification, the first layer’s accuracy (47%) is greater than later layers (<30%). For instrument classification, we see peak accuracy at the last layer (97%), but a remarkable change occurs at the fourth layer (71%) compared to the third layer (55%). Genre classification accuracy at the last layer (75%) is better than earlier layers (<66%). These trends hold across all models. We see similar trends for the speech tasks. Discussion The results support our hypothesis, providing strong evidence that CNNs implicitly learn a hierarchy in sound. This mirrors the human brain’s hierarchical encoding and deserves further investigation if other deep learning architectures and training methodologies also encode similar hierarchies. We also see how low-level tasks are implicitly learnt.

Abstract

Understanding the collective behavior of dynamical systems is essential for explaining various emergent phenomena in natural and engineered settings. A key step in this process is formulating an appropriate mathematical description of the individual systems and network of systems. In this context, a range of physical systems is considered here, including the classical pendula, superconducting Josephson junctions, power grids, and various others. Despite the diversity of the systems in terms of physical structure and their application domains, they exhibit strikingly similar dynamical features, namely, phase dynamics governed by inertia and damping, and in their response to external forcing. This observation creates interest and motivates a search for a unified theoretical framework capable of capturing the fundamentals of their dynamical behaviors exhibited across the systems. This review critically examines the up-to-date research activities on the dynamics of the second-order phase oscillator, henceforth claimed here as a universality class by its own merits as a simple nonlinear dynamical model representing a broad class of physical systems. It offers a common mathematical framework to develop a comprehensive understanding, from a general perspective, that bridges, the theoretical and experimental observations of pendulum motion, Josephson junctions, and power grids and their collective behaviors. While each of these systems has been discussed in disparate physical contexts, their underlying mathematical structures reveal strong commonalities. In particular, we highlight the importance of analyzing these systems through the lens of nonlinear phase dynamics to uncover their shared mechanisms and system-specific variety of behaviors as well. This survey mainly focuses on some specific interrelated themes: (i) collective phenomena and emergent synchronization; (ii) the role of heterogeneity in terms of system parameters and effects of noise on the emergent dynamics; (iii) multi-stability and complex transient regimes; (iv) the integration of machine learning for model discovery, control, and prediction; and (v) the broader applicability of phase oscillator models across diverse domains beyond the canonical systems considered here. By systematically comparing the dynamical behaviors of the varied physical systems within a cohesive mathematical framework of second-order phase oscillators, this review seeks for the universal and distinctive features of nonlinear dynamics of the three systems, their collective behaviors such as emergent synchrony, partial synchrony, or chimera states, and specifically explains real-life phenomena, and crowd synchrony that may lead to a collapse of a footbridge and the failure of a power grid. Besides our main emphasis on these systems, brief notes have been added on other systems where this second-order phase model explains their dynamical properties. A broad synthesis on the topic will not only deepen our theoretical understanding but also suggest any design and control of complex dynamical systems in both natural and engineered settings.

Abstract

The likelihood of intense heatwaves in South Asia is increasing due to climate change, highlighting the need to understand their evolving spatiotemporal patterns. Using a complex network-based approach, we analyze synchronous extreme heat events across South and West Asia over three 30-year periods: two historical phases (1960–1989, 1990–2019) and a near-future projection (2020–2049) under the SSP2-4.5 scenario. Our findings reveal a shift in heatwave synchronization from western and central Asia before 1990 towards Pakistan, northwest India, and the southwestern Tibetan Plateau by the mid-21st century. This shift is primarily driven by increased surface sensible heat flux, which enhances atmospheric diabatic heating and strengthens the early-summer circumglobal teleconnection. Additionally, atmospheric conditions over the North Atlantic-Greenland sector modulate South Asian heatwave synchronization. Our study provides novel insights into the evolving land-atmosphere interactions driving extreme heat events, with implications for heatwave predictability and risk assessment in a warming world.

Abstract

We investigate the effect of partial order parameter adaptation in the form of general functions on the synchronization behavior of coupled Kuramoto oscillators on top of random hypergraph models. The interactions between the oscillators are considered pairwise and triangular. Using the Ott–Antonsen ansatz, we obtain a set of self-consistent equations of the order parameter that describe the synchronization patterns. A broad diversity of synchronization transitions is observed as a result of the interplay between the partial adaptation approach, generalized adaptation functions, and coupling strengths. The system specifically shows a double-jump transition under a power law form of the adaptation function. A polynomial form of the adaptation function leads to the emergence of an intermediate synchronization state for specific combinations of one negative and one positive coefficient. Moreover, the synchronization transition may become continuous or explosive when the pairwise coupling strength varies. This type of synchronization behavior is further found by using a Gaussian adaptation function.

Abstract

Complex biological networks, encompassing metabolic pathways, gene regulatory systems, and protein–protein interaction networks, often exhibit scale-free structures characterized by heavy-tailed degree distributions. However, empirical studies reveal significant deviations from ideal power-law behavior, underscoring the need for more flexible and accurate probabilistic models. In this work, we propose the Compounded Burr (CBurr) distribution, a novel four-parameter family derived by compounding the Burr distribution with a discrete mixing process. This model is specifically designed to capture both the body and tail behavior of real-world network degree distributions with applications to biological networks. We rigorously derive its statistical properties, including moments, hazard and risk functions, and tail behavior, and develop an efficient maximum likelihood estimation framework. The CBurr model demonstrates broad applicability to networks with complex connectivity patterns, particularly in biological, social, and technological domains. Extensive experiments on large-scale biological network datasets show that CBurr consistently outperforms classical power-law, lognormal, and other heavy-tailed models across the full degree spectrum. By providing a statistically grounded and interpretable framework, the CBurr model enhances our ability to characterize the structural heterogeneity of biological networks.

Abstract

This study aims to develop a generalized concept that will enable double explosive transitions in the forward and backward directions or a combination thereof. We found two essential factors for generating such phase transitions: the use of higher-order (triadic) interactions and the partial adaptation of a global order parameter acting on the triadic coupling. A compromise between the two factors may result in a double explosive transition. To reinforce numerical observations, we employed the Ott-Antonsen ansatz. We observed that for a wide class of hypergraphs, combining two elements can result in a double explosive transition.

Abstract

Successful aerial manipulation largely depends on how effectively a controller can tackle the coupling dynamic forces between the aerial vehicle and the manipulator. However, this control problem has remained largely unsolved as the existing control approaches either require precise knowledge of the aerial vehicle/manipulator inertial couplings, or neglect the statedependent uncertainties especially arising during the interaction phase. This work proposes an adaptive control solution to overcome this long standing control challenge without any a priori knowledge of the coupling dynamic terms. Additionally, in contrast to the existing adaptive control solutions, the proposed control framework is modular, that is, it allows independent tuning of the adaptive gains for the vehicle position sub-dynamics, the vehicle attitude sub-dynamics, and the manipulator subdynamics. Stability of the closed loop under the proposed scheme is derived analytically, and real-time experiments validate the effectiveness of the proposed scheme over the state-of-the-art approaches.

Abstract

An embodiment herein provides a re-configurable unmanned aerial vehicle that re-configures its shape based on the shape, size, weight of a payload, and efficiently performs payload delivery in real-time. The re-configurable unmanned aerial vehicle includes one or more rotor units placed at corners and is connected by one or more scissor units. The re-configurable unmanned aerial vehicle approaches the payload in a first location, and analyses the position and dimension of the payload with a camera, that enables the one or more scissor units to adjust its length by at least one elongation or compression following size and shape of the payload and fit the payload within the re- configurable unmanned aerial vehicle. The re-configurable unmanned aerial vehicle takes off carrying the payload from the first location and lands at a second location.

Abstract

Grasping using an aerial robot can have many applications ranging from infrastructure inspection and maintenance to precise agriculture. However, aerial grasping is a challenging problem since the robot has to maintain an accurate position and orientation relative to the grasping object, while negotiating various forms of uncertainties (e.g., contact force from the object). To address such challenges, in this paper, we integrate a novel passive gripper design and advanced adaptive control methods to enable robust aerial grasping. The gripper is enabled by a pre-stressed band with two stable states (a flat shape and a curled shape). In this case, it can automatically initiate the grasping process upon contact with an object. The gripper also features a cable-driven system by a single DC motor to open the gripper without using cumbersome pneumatics. Since the gripper is passively triggered and initially has a straight shape, it can function without precisely aligning the gripper with the object (within an 80 mm tolerance). Our adaptive control scheme eliminates the need for any a priori knowledge (nominal or upper bounds) of uncertainties. The closed-loop stability of the system is analyzed via Lyapunov-based method. Combining the gripper and the adaptive control, we conduct comparative realtime experimental results to demonstrate the effectiveness of the proposed integrated system for grasping. Our integrated approach can pave the way to enhance aerial grasping for different applications.

Abstract

Contributing catchment land use patterns under rapid urbanization, agricultural practices and climatic variability in tropical regions significantly influence the sediment load of inland waterbodies such as reservoirs. This study presents a modeling framework to estimate qualitative turbidity dynamics in two tropical reservoirs (Bhadra and Tungabhadra) within same river basin but different landscape characteristics, using Sentinel-2 satellite data from 2016 to 2021 using Google Earth Engine (GEE). The framework demonstrates a robust capability of Sentinel-2 data, in detecting, mapping spatiotemporal trends and classifying turbidity levels from low to high along with impact of potential catchment characteristics. The Normalized Difference Turbidity Index (NDTI) values have a strong correlation with in-situ data (Earth Observing and Mapping EOMAP and field-based observations) having values > 10 NTU (nephelometric turbidity unit) with R2 = 0.81 and standard Error Estimate (SEE) of 5.3 respectively. In addition, the correlation between red band reflectance values with <10 NTU, has demonstrated a strong relation with an R2 and SEE of 0.74 and 1.7 respectively. Thus, the combination of NDTI and single red band reflectance values were used further for classifying the different turbidity levels (low, moderate and high). It was also observed that diverse land use patterns, particularly high proportions of urban and agriculture area along with rainfall, impact the annual and seasonal fluctuation of turbidity level in the reservoirs. The study demonstrates that remote sensing (RS) approach can effectively estimate turbidity levels in tropical reservoirs, highlighting the impact of catchment characteristics and supporting improved monitoring and sustainable water resource management.