Search

Bird's eye view (BEV)-based 3D perception plays a crucial role in autonomous driving applications. The rise of large language models has spurred interest in BEV-based captioning to understand object behavior in the surrounding environment. However, existing approaches treat perception and captioning as separate tasks, focusing on the performance of only one of the tasks and overlooking the potential benefits of multimodal alignment. To bridge this gap between modalities, we introduce MTA, a novel multimodal task alignment framework that boosts both BEV perception and captioning. MTA consists of two key components: (1) BEV-Language Alignment (BLA), a contextual learning mechanism that aligns the BEV scene representations with ground-truth language representations, and (2) Detection-Captioning Alignment (DCA), a cross-modal prompting mechanism that aligns detection and captioning outputs. MTA integrates into state-of-the-art baselines during training, adding no extra computational complexity at runtime. Extensive experiments on the nuScenes and TOD3Cap datasets show that MTA significantly outperforms state-of-the-art baselines, achieving a 4.9% improvement in perception and a 9.2% improvement in captioning. These results underscore the effectiveness of unified alignment in reconciling BEV-based perception and captioning., Comment: 10 pages

Scaling model sizes to scale performance has worked remarkably well for the current large language models paradigm. The research and empirical findings of various scaling studies led to novel scaling results and laws that guides subsequent research. High training costs for contemporary scales of data and models result in a lack of thorough understanding of how to tune and arrive at such training setups. One direction to ameliorate the cost of pretraining large models is to warmstart the large-scale training from smaller models that are cheaper to tune. In this work, we attempt to understand if the behavior of optimal hyperparameters can be retained under warmstarting for scaling. We explore simple operations that allow the application of theoretically motivated methods of zero-shot transfer of optimal hyperparameters using {\mu}Transfer. We investigate the aspects that contribute to the speedup in convergence and the preservation of stable training dynamics under warmstarting with {\mu}Transfer. We find that shrinking smaller model weights, zero-padding, and perturbing the resulting larger model with scaled initialization from {\mu}P enables effective warmstarting of $\mut{}$.

Gate-based quantum computers are an innovative tool for experimentally studying the core principles of quantum mechanics. This work presents the first observation of quantum anomalous heat flow between two qubits and investigates the role of mid-circuit measurements in this context. Using mid-circuit measurements, we designed quantum circuits that violate the semi-classical heat flow bound, witnessing negativities in the underlying Kirkwood-Dirac quasiprobability distribution, which indicates the presence of quantum correlations between the subsystems. Mid-circuit measurements, crucial for probing qubits during the experiment, enabled these observations but also introduced disturbances, such as energy leakage, leading to deviations from theoretical predictions. We modeled these noise effects, providing insight into the limitations of current mid-circuit measurement techniques., Comment: 9 pages, 9 figures

We investigate the path integral of the four-dimensional Einstein-Hilbert gravity for the deSitter-like Universe with fluctuations and study the transition amplitude from one boundary configuration to another for the gravitational system described by lapse, scale-factor and metric fluctuation field. For variational consistency, the bulk theory needs suitable augmentation of boundary action. A specific boundary choice on the scale-factor is achieved by an infinite family of covariant boundary actions, each restricting the boundary choices for the fluctuation field. The boundary choices for the two are intimately tied to each other via general covariance, and can't be independently chosen. In the background field formalism, the full gauge-fixed path-integral on ignoring back-reaction, decouples to a path-integral over scale-factor and fluctuation field, coupling to each other only via lapse. After exactly doing the path-integral over the scale factor, the fluctuation path-integral is computed to one-loop via mode-decomposition leading to a quantum corrected lapse action. Its complex saddles shows background saddles getting corrected due to contribution from fluctuations. ``No-boundary'' saddles are seen to be no longer the saddles of the full theory, implying a Universe starting with a non-vanishing size at the corrected saddles. One-loop contribution is seen to diverge logarithmically at the background ``no-boundary'' saddles. Counter term added to cancel divergence lead to finite lapse action and transition amplitude., Comment: v1: 1+42 pages

Characterization of the participant-zone (PZ) in the $^{129,124}$Xe + $^{112,124}$Sn reaction at the energy range 65-150 MeV/nucleon reveals copious cluster production. A detailed study of the chemical composition as a function of the impact parameter shows that heavier clusters ($^6$He, $^{6-8}$Li, $^{7-10}$Be...) are most likely produced for central collisions. A hierarchy of the cluster production with the neutron-richness of the total system is observed, suggesting a full mixing of the projectile and target in the PZ. An estimate of the maximum density in central collisions has been deduced from the kinetic energy of the emitted fragments, reaching almost 2 time the normal density (2$\rho_0$) at 150 MeV/nucleon., Comment: 5 pages, 3 figures, nn2024 conference

Semiconductor research has shifted towards exploring two-dimensional (2D) materials as candidates for next-generation electronic devices due to the limitations of existing silicon technology. Transition Metal Dichalcogenides (TMDCs) stand out for their exceptional optoelectronic properties and potential for advanced device integration. This thesis focuses on the synthesis of 2D TMDCs using Chemical Vapor Deposition (CVD) for their potential applications in transistors, memory, and neuromorphic computing. By optimizing the NaCl-assisted CVD method and examining their optical properties through Raman and photoluminescence spectroscopy, challenges such as premature growth, defects, and non-uniformity in MoS2 samples are addressed. The thesis highlights device fabrication techniques and electrical performance of salt-assisted CVD-grown MoS2 field-effect transistors, which exhibit hysteresis-free behavior and high field-effect mobility. A novel etching-free transfer technique is introduced, improving transistor performance and enabling applications in flexible optoelectronics. The thesis also explores monolayer MoS2 mem-transistors, demonstrating multifunctional room temperature transistor and high-temperature multi-level memory behaviour. These devices leverage interfacial physics and ion dynamics to achieve non-volatile memory with multi-level storage capabilities. Additionally, high density memory devices using monolayer WS2 are developed, which demonstrate 6-bit memory operation with neuromorphic biomimetic plasticity. The study also includes 2D TMDCs and their hetero-bilayers as potential 2D dilute magnetic semiconductors via doping, strain engineering using density functional theory and micromagnetic simulations, revealing potential applications in spintronics. This thesis makes significant contributions to advancing 2D materials for next-generation electronics and spintronic devices., Comment: PhD Thesis

The importance of tuning hyperparameters in Machine Learning (ML) and Deep Learning (DL) is established through empirical research and applications, evident from the increase in new hyperparameter optimization (HPO) algorithms and benchmarks steadily added by the community. However, current benchmarking practices using averaged performance across many datasets may obscure key differences between HPO methods, especially for pairwise comparisons. In this work, we apply Linear Mixed-Effect Models-based (LMEMs) significance testing for post-hoc analysis of HPO benchmarking runs. LMEMs allow flexible and expressive modeling on the entire experiment data, including information such as benchmark meta-features, offering deeper insights than current analysis practices. We demonstrate this through a case study on the PriorBand paper's experiment data to find insights not reported in the original work.

The design of energy-efficient, high-performance, and reliable Convolutional Neural Network (CNN) accelerators involves significant challenges due to complex power and thermal management issues. This paper introduces SAfEPaTh, a novel system-level approach for accurately estimating power and temperature in tile-based CNN accelerators. By addressing both steady-state and transient-state scenarios, SAfEPaTh effectively captures the dynamic effects of pipeline bubbles in interlayer pipelines, utilizing real CNN workloads for comprehensive evaluation. Unlike traditional methods, it eliminates the need for circuit-level simulations or on-chip measurements. Our methodology leverages TANIA, a cutting-edge hybrid digital-analog tile-based accelerator featuring analog-in-memory computing cores alongside digital cores. Through rigorous simulation results using the ResNet18 model, we demonstrate SAfEPaTh's capability to accurately estimate power and temperature within 500 seconds, encompassing CNN model accelerator mapping exploration and detailed power and thermal estimations. This efficiency and accuracy make SAfEPaTh an invaluable tool for designers, enabling them to optimize performance while adhering to stringent power and thermal constraints. Furthermore, SAfEPaTh's adaptability extends its utility across various CNN models and accelerator architectures, underscoring its broad applicability in the field. This study contributes significantly to the advancement of energy-efficient and reliable CNN accelerator designs, addressing critical challenges in dynamic power and thermal management.

The gravitational path-integral of Gauss-Bonnet gravity is investigated and the transition from one spacelike boundary configuration to another is analyzed. Of particular interest is the case of Neumann and Robin boundary conditions which is known to lead to stable Universe in Einstein-Hilbert gravity in four spacetime dimensions. After setting up the variational problem and computing the necessary boundary terms, the transition amplitude can be exactly computed in the mini-superspace approximation. The $\hbar\to0$ limit brings out the dominant pieces in the path-integral which is traced to an initial configuration corresponding to Hartle-Hawking no-boundary Universe. A deeper study involving Picard-Lefschetz methods not only allow us to find the integration contour along which the path-integral becomes convergent but also aids in understanding the crossover from Euclidean to Lorentzian signature. Saddle analysis further highlights the boundary configurations giving dominant contribution to the path-integral which is seen to be those corresponding to Hartle-Hawking no-boundary proposal and agrees with the exact computation. To ensure completeness, a comparison with the results from Wheeler-DeWitt equation is done., Comment: v1: 1 + 31 pages, 4 figures. For a special edition of Gen. Rel. Grav., "QG@RRI 2023''

Aligning future system design with the ever-increasing compute needs of large language models (LLMs) is undoubtedly an important problem in today's world. Here, we propose a general performance modeling methodology and workload analysis of distributed LLM training and inference through an analytical framework that accurately considers compute, memory sub-system, network, and various parallelization strategies (model parallel, data parallel, pipeline parallel, and sequence parallel). We validate our performance predictions with published data from literature and relevant industry vendors (e.g., NVIDIA). For distributed training, we investigate the memory footprint of LLMs for different activation re-computation methods, dissect the key factors behind the massive performance gain from A100 to B200 ($\sim$ 35x speed-up closely following NVIDIA's scaling trend), and further run a design space exploration at different technology nodes (12 nm to 1 nm) to study the impact of logic, memory, and network scaling on the performance. For inference, we analyze the compute versus memory boundedness of different operations at a matrix-multiply level for different GPU systems and further explore the impact of DRAM memory technology scaling on inference latency. Utilizing our modeling framework, we reveal the evolution of performance bottlenecks for both LLM training and inference with technology scaling, thus, providing insights to design future systems for LLM training and inference., Comment: 12 pages, 9 figures

Graph games lie at the algorithmic core of many automated design problems in computer science. These are games usually played between two players on a given graph, where the players keep moving a token along the edges according to pre-determined rules, and the winner is decided based on the infinite path traversed by the token from a given initial position. In bidding games, the players initially get some monetary budgets which they need to use to bid for the privilege of moving the token at each step. Each round of bidding affects the players' available budgets, which is the only form of update that the budgets experience. We introduce bidding games with charging where the players can additionally improve their budgets during the game by collecting vertex-dependent charges. Unlike traditional bidding games (where all charges are zero), bidding games with charging allow non-trivial recurrent behaviors. We show that the central property of traditional bidding games generalizes to bidding games with charging: For each vertex there exists a threshold ratio, which is the necessary and sufficient fraction of the total budget for winning the game from that vertex. While the thresholds of traditional bidding games correspond to unique fixed points of linear systems of equations, in games with charging, these fixed points are no longer unique. This significantly complicates the proof of existence and the algorithmic computation of thresholds for infinite-duration objectives. We also provide the lower complexity bounds for computing thresholds for Rabin and Streett objectives, which are the first known lower bounds in any form of bidding games (with or without charging), and we solve the following repair problem for safety and reachability games that have unsatisfiable objectives: Can we distribute a given amount of charge to the players in a way such that the objective can be satisfied?

Extended reality (XR) is at the center of attraction in the research community due to the emergence of augmented, mixed, and virtual reality applications. The performance of such applications needs to be uptight to maintain the requirements of latency, energy consumption, and freshness of data. Therefore, a comprehensive performance analysis model is required to assess the effectiveness of an XR application but is challenging to design due to the dependence of the performance metrics on several difficult-to-model parameters, such as computing resources and hardware utilization of XR and edge devices, which are controlled by both their operating systems and the application itself. Moreover, the heterogeneity in devices and wireless access networks brings additional challenges in modeling. In this paper, we propose a novel modeling framework for performance analysis of XR applications considering edge-assisted wireless networks and validate the model with experimental data collected from testbeds designed specifically for XR applications. In addition, we present the challenges associated with performance analysis modeling and present methods to overcome them in detail. Finally, the performance evaluation shows that the proposed analytical model can analyze XR applications' performance with high accuracy compared to the state-of-the-art analytical models., Comment: 12 pages, 4 figures; To appear in Proceedings of IEEE International Conference on Distributed Computing Systems (ICDCS), 2024

In Spectrum cartography (SC), the generation of exposure maps for radio frequency electromagnetic fields (RF-EMF) spans dimensions of frequency, space, and time, which relies on a sparse collection of sensor data, posing a challenging ill-posed inverse problem. Cartography methods based on models integrate designed priors, such as sparsity and low-rank structures, to refine the solution of this inverse problem. In our previous work, EMF exposure map reconstruction was achieved by Generative Adversarial Networks (GANs) where physical laws or structural constraints were employed as a prior, but they require a large amount of labeled data or simulated full maps for training to produce efficient results. In this paper, we present a method to reconstruct EMF exposure maps using only the generator network in GANs which does not require explicit training, thus overcoming the limitations of GANs, such as using reference full exposure maps. This approach uses a prior from sensor data as Local Image Prior (LIP) captured by deep convolutional generative networks independent of learning the network parameters from images in an urban environment. Experimental results show that, even when only sparse sensor data are available, our method can produce accurate estimates.

With the increasing computational costs associated with deep learning, automated hyperparameter optimization methods, strongly relying on black-box Bayesian optimization (BO), face limitations. Freeze-thaw BO offers a promising grey-box alternative, strategically allocating scarce resources incrementally to different configurations. However, the frequent surrogate model updates inherent to this approach pose challenges for existing methods, requiring retraining or fine-tuning their neural network surrogates online, introducing overhead, instability, and hyper-hyperparameters. In this work, we propose FT-PFN, a novel surrogate for Freeze-thaw style BO. FT-PFN is a prior-data fitted network (PFN) that leverages the transformers' in-context learning ability to efficiently and reliably do Bayesian learning curve extrapolation in a single forward pass. Our empirical analysis across three benchmark suites shows that the predictions made by FT-PFN are more accurate and 10-100 times faster than those of the deep Gaussian process and deep ensemble surrogates used in previous work. Furthermore, we show that, when combined with our novel acquisition mechanism (MFPI-random), the resulting in-context freeze-thaw BO method (ifBO), yields new state-of-the-art performance in the same three families of deep learning HPO benchmarks considered in prior work., Comment: Published at the 41st International Conference on Machine Learning (ICML), Vienna, Austria

A connected graph, on four or more vertices, is matching covered if every edge is present in some perfect matching. An ear decomposition theorem (similar to the one for $2$-connected graphs) exists for bipartite matching covered graphs due to Hetyei. From the results and proofs of Lov\'asz and Plummer, that rely on Hetyei's theorem, one may deduce that any minimal bipartite matching covered graph has at least $2(m-n+2)$ vertices of degree two (where minimal means that deleting any edge results in a graph that is not matching covered); such a graph is said to be extremal if it attains the stated lower bound. In this paper, we provide a complete characterization of the class of extremal minimal bipartite matching covered graphs. In particular, we prove that every such graph $G$ is obtained from two copies of a tree devoid of degree two vertices, say $T$ and $T'$, by adding edges -- each of which joins a leaf of $T$ with the corresponding leaf of $T'$. Apart from the aforementioned bound, there are four other bounds that appear in, or may be deduced from, the work of Lov\'asz and Plummer. Each of these bounds leads to a notion of extremality. In this paper, we obtain a complete characterization of all of these extremal classes and also establish relationships between them. Two of our characterizations are in the same spirit as the one stated above. For the remaining two extremal classes, we reduce each of them to one of the already characterized extremal classes using standard matching theoretic operations., Comment: Submitted to Innovations in Graph Theory

Reinforcement learning (RL) based autonomous driving has emerged as a promising alternative to data-driven imitation learning approaches. However, crafting effective reward functions for RL poses challenges due to the complexity of defining and quantifying good driving behaviors across diverse scenarios. Recently, large pretrained models have gained significant attention as zero-shot reward models for tasks specified with desired linguistic goals. However, the desired linguistic goals for autonomous driving such as "drive safely" are ambiguous and incomprehensible by pretrained models. On the other hand, undesired linguistic goals like "collision" are more concrete and tractable. In this work, we introduce LORD, a novel large models based opposite reward design through undesired linguistic goals to enable the efficient use of large pretrained models as zero-shot reward models. Through extensive experiments, our proposed framework shows its efficiency in leveraging the power of large pretrained models for achieving safe and enhanced autonomous driving. Moreover, the proposed approach shows improved generalization capabilities as it outperforms counterpart methods across diverse and challenging driving scenarios.

Connected and autonomous vehicles (CAVs) rely heavily upon time-sensitive information update services to ensure the safety of people and assets, and satisfactory entertainment applications. Therefore, the freshness of information is a crucial performance metric for CAV services. However, information from roadside sensors and nearby vehicles can get delayed in transmission due to the high mobility of vehicles. Our research shows that a CAV's relative distance and speed play an essential role in determining the Age-of-Information (AoI). With an increase in AoI, incremental service aggregation issues are observed with out-of-sequence information updates, which hampers the performance of low-latency applications in CAVs. In this paper, we propose a novel AoI-based service aggregation method for CAVs, which can process the information updates according to their update cycles. First, the AoI for sensors and vehicles is modeled, and a predictive AoI system is designed. Then, to reduce the overall service aggregation time and computational load, intervals are used for periodic AoI prediction, and information sources are clustered based on the AoI value. Finally, the system aggregates services for CAV applications using the predicted AoI. We evaluate the system performance based on data sequencing success rate (DSSR) and overall system latency. Lastly, we compare the performance of our proposed system with three other state-of-the-art methods. The evaluation and comparison results show that our proposed predictive AoI-based service aggregation system maintains satisfactory latency and DSSR for CAV applications and outperforms other existing methods., Comment: 6 pages, 8 figures, to appear in the Proceedings of IEEE International Conference on Communications (IEEE ICC, 9-13 June 2024, Denver, CO, USA)

While deep learning has celebrated many successes, its results often hinge on the meticulous selection of hyperparameters (HPs). However, the time-consuming nature of deep learning training makes HP optimization (HPO) a costly endeavor, slowing down the development of efficient HPO tools. While zero-cost benchmarks, which provide performance and runtime without actual training, offer a solution for non-parallel setups, they fall short in parallel setups as each worker must communicate its queried runtime to return its evaluation in the exact order. This work addresses this challenge by introducing a user-friendly Python package that facilitates efficient parallel HPO with zero-cost benchmarks. Our approach calculates the exact return order based on the information stored in file system, eliminating the need for long waiting times and enabling much faster HPO evaluations. We first verify the correctness of our approach through extensive testing and the experiments with 6 popular HPO libraries show its applicability to diverse libraries and its ability to achieve over 1000x speedup compared to a traditional approach. Our package can be installed via pip install mfhpo-simulator., Comment: Accepted to AutoML Conference 2024 ABCD Track

The deposition of micron particles finds importance in meteorology and several engineering applications such as deposition of dust in gas lines, carbon deposition in engine exhaust, designing effective air-cleaning systems and estimating deposition of inhaled drug or atmospheric pollutants to determine its consequences on human health. Although the existing literature on deposition in straight tubes is quite mature, an experimental study on deposition in micro capillaries with a wide ranges of Re that models particle dynamics in lungs, is missing. The deposition of atmospheric pollutants and nebulized drugs in the lung depends on various biological factors such as flow properties, lung morphology, breathing patterns, particle properties, deposition mechanism, etc. To complicate matters, each breath manifests flows spanning a wide range of Reynolds numbers in various regions of the lung. In this study, the deposition of nebulized aerosol was experimentally investigated in phantom bronchioles of diameters relevant to the 7th to the 23rd branching generations and over the entire range of Re manifest during one breathing cycle. The aerosol fluid was loaded with boron doped carbon quantum dots as a fluorophore. An aerosol was generated of this mixture fluid using an ultrasonic nebulizer, producing droplets of 6.5$\mu$m as the mean diameter. The amount of aerosol deposited on the bronchiole walls was measured using a spectrofluorometer. Finally, a universal bronchiole scale deposition model is proposed which can form the building block for lung-scale aerosol deposition prediction., Comment: MS Thesis submitted at IIT Madras, India

In this article, we consider a local projection stabilisation for a Hybrid High-Order (HHO) approximation of the Oseen problem. We prove an existence-uniqueness result under a stronger SUPG-like norm. We improve the stability and provide error estimation in stronger norm for convection dominated Oseen problem. We also derive an optimal order error estimate under the SUPG-like norm for equal-order polynomial discretisation of velocity and pressure spaces. Numerical experiments are performed to validate the theoretical results.

Autonomous driving is a complex and challenging task that aims at safe motion planning through scene understanding and reasoning. While vision-only autonomous driving methods have recently achieved notable performance, through enhanced scene understanding, several key issues, including lack of reasoning, low generalization performance and long-tail scenarios, still need to be addressed. In this paper, we present VLP, a novel Vision-Language-Planning framework that exploits language models to bridge the gap between linguistic understanding and autonomous driving. VLP enhances autonomous driving systems by strengthening both the source memory foundation and the self-driving car's contextual understanding. VLP achieves state-of-the-art end-to-end planning performance on the challenging NuScenes dataset by achieving 35.9\% and 60.5\% reduction in terms of average L2 error and collision rates, respectively, compared to the previous best method. Moreover, VLP shows improved performance in challenging long-tail scenarios and strong generalization capabilities when faced with new urban environments., Comment: CVPR2024