arXiv subjects

Deepak Gupta

Publications and source records attributed to Deepak Gupta.

At least 37 records · Page 2Linked to original sources

arXiv2024

Overview of TREC 2024 Medical Video Question Answering (MedVidQA) Track

One of the key goals of artificial intelligence (AI) is the development of a multimodal system that facilitates communication with the visual world (image and video) using a natural language query. Earlier works on medical question answering primarily focused on textual and visual (image) modalities, which may be inefficient in answering questions requiring demonstration. In recent years, significant progress has been achieved due to the introduction of large-scale language-vision datasets and the development of efficient deep neural techniques that bridge the gap between language and visual understanding. Improvements have been made in numerous vision-and-language tasks, such as visual captioning visual question answering, and natural language video localization. Most of the existing work on language vision focused on creating datasets and developing solutions for open-domain applications. We believe medical videos may provide the best possible answers to many first aid, medical emergency, and medical education questions. With increasing interest in AI to support clinical decision-making and improve patient engagement, there is a need to explore such challenges and develop efficient algorithms for medical language-video understanding and generation. Toward this, we introduced new tasks to foster research toward designing systems that can understand medical videos to provide visual answers to natural language questions, and are equipped with multimodal capability to generate instruction steps from the medical video. These tasks have the potential to support the development of sophisticated downstream applications that can benefit the public and medical professionals.

cs.CV↗

arXiv2024

Overview of TREC 2024 Biomedical Generative Retrieval (BioGen) Track

With the advancement of large language models (LLMs), the biomedical domain has seen significant progress and improvement in multiple tasks such as biomedical question answering, lay language summarization of the biomedical literature, clinical note summarization, etc. However, hallucinations or confabulations remain one of the key challenges when using LLMs in the biomedical and other domains. Inaccuracies may be particularly harmful in high-risk situations, such as medical question answering, making clinical decisions, or appraising biomedical research. Studies on the evaluation of the LLMs abilities to ground generated statements in verifiable sources have shown that models perform significantly worse on lay-user-generated questions, and often fail to reference relevant sources. This can be problematic when those seeking information want evidence from studies to back up the claims from LLMs. Unsupported statements are a major barrier to using LLMs in any applications that may affect health. Methods for grounding generated statements in reliable sources along with practical evaluation approaches are needed to overcome this barrier. Towards this, in our pilot task organized at TREC 2024, we introduced the task of reference attribution as a means to mitigate the generation of false statements by LLMs answering biomedical questions.

cs.IR↗

arXiv2024

Dynamics of switching processes: general results and applications to intermittent active motion

Systems switching between different dynamical phases is an ubiquitous phenomenon. The general understanding of such a process is limited. To this end, we present a general expression that captures fluctuations of a system exhibiting a switching mechanism. Specifically, we obtain an exact expression of the Laplace-transformed characteristic function of the particle's position. Then, the characteristic function is used to compute the effective diffusion coefficient of a system performing intermittent dynamics. Further, we employ two examples: 1) Generalized run-and-tumble active particle, and 2) an active particle switching its dynamics between generalized active run-and-tumble motion and passive Brownian motion. In each case, explicit computations of the spatial cumulants are presented. Our findings reveal that the particle's position probability density function exhibit rich behaviours due to intermittent activity. Numerical simulations confirm our findings.

cond-mat.stat-mech↗

arXiv2024

Optimizing Cost through Dynamic Stochastic Resetting

The cost of stochastic resetting is considered within the context of a discrete random walk model. In addition to standard stochastic resetting, for which a reset occurs with a certain probability after \emph{each} step, we introduce a novel resetting protocol which we dubbed {\it dynamic resetting}. This protocol entails an additional dynamic constraint related to the direction of successive steps of the random walker. We study this novel protocol for a one-dimensional random walker on an infinite lattice. We analyze the impact of the constraint on the walker's mean-first passage time and the cost (fluctuations) of the resets as a function of distance of target from the resetting location. Further, cost optimized search strategies are discussed.

cond-mat.stat-mech↗

arXiv2024

Enhancing Deep Learning based RMT Data Inversion using Gaussian Random Field

Deep learning (DL) methods have emerged as a powerful tool for the inversion of geophysical data. When applied to field data, these models often struggle without additional fine-tuning of the network. This is because they are built on the assumption that the statistical patterns in the training and test datasets are the same. To address this, we propose a DL-based inversion scheme for Radio Magnetotelluric data where the subsurface resistivity models are generated using Gaussian Random Fields (GRF). The network's generalization ability was tested with an out-of-distribution (OOD) dataset comprising a homogeneous background and various rectangular-shaped anomalous bodies. After end-to-end training with the GRF dataset, the pre-trained network successfully identified anomalies in the OOD dataset. Synthetic experiments confirmed that the GRF dataset enhances generalization compared to a homogeneous background OOD dataset. The network accurately recovered structures in a checkerboard resistivity model, and demonstrated robustness to noise, outperforming traditional gradient-based methods. Finally, the developed scheme is tested using exemplary field data from a waste site near Roorkee, India. The proposed scheme enhances generalization in a data-driven supervised learning framework, suggesting a promising direction for OOD generalization in DL methods.

cs.LG↗

arXiv2024

Thermodynamic cost of finite-time stochastic resetting

Recent experiments have implemented resetting by means of an external trap, whereby a system relaxes to the minimum of the trap and is reset in a finite time. In this work, we set up and analyse the thermodynamics of such a protocol. We present a general framework, even valid for non-Poissonian resetting, that captures the thermodynamic work required to maintain a resetting process up to a given observation time, and exactly calculate the moment generating function of this work. Our framework is valid for a wide range of systems, the only assumption being relaxation to equilibrium in the resetting trap. Examples and extensions are considered. In the case of Brownian motion, we investigate optimal resetting schemes that minimise work and its fluctuations, the mean work for arbitrary switching protocols and comparisons to previously studied resetting schemes. Numerical simulations are performed to validate our findings.

cond-mat.stat-mech↗

arXiv2024

Bioimpedance a Diagnostic Tool for Tobacco Induced Oral Lesions: a Mixed Model cross-sectional study

Introduction: Electrical impedance spectroscopy (EIS) has recently developed as a novel diagnostic device for screening and evaluating cervical dysplasia, prostate cancer, breast cancer and basal cell carcinoma. The current study aimed to validate and evaluate bioimpedance as a diagnostic tool for tobacco-induced oral lesions. Methodology: The study comprised 50 OSCC and OPMD tissue specimens for in-vitro study and 320 subjects for in vivo study. Bioimpedance device prepared and calibrated. EIS measurements were done for the habit and control groups and were compared. Results: The impedance value in the control group was significantly higher compared to the OPMD and OSCC groups. Diagnosis based on BIS measurements has a sensitivity of 95.9% and a specificity of 86.7%. Conclusion: Bioimpedance device can help in decision-making for differentiating OPMD and OSCC cases and their management, especially in primary healthcare settings. Keywords: Impedance, Cancer, Diagnosis, Device, Community

q-bio.QM↗

arXiv2024

Thermodynamic work of partial resetting

Partial resetting, whereby a state variable $x(t)$ is reset at random times to a value $a x (t)$, $0\leq a \leq 1$, generalizes conventional resetting by introducing the resetting strength $a$ as a parameter. Partial resetting generates a broad family of non-equilibrium steady states (NESS) that interpolates between the conventional NESS at strong resetting ($a=0$) and a Gaussian distribution at weak resetting ($a \to 1$). Here, such resetting processes are studied from a thermodynamic perspective, and the mean cost associated with maintaining such NESS are derived. The resetting phase of the dynamics is implemented by a resetting potential $Φ(x)$ that mediates the resets in finite time. By working in an ensemble of trajectories with a fixed number of resets, we study both the steady-state properties of the propagator and its moments. The thermodynamic work needed to sustain the resulting NESS is then investigated. We find that different resetting traps can give rise to rates of work with widely different dependencies on the resetting strength $a$. Surprisingly, in the case of resets mediated by a harmonic trap with otherwise free diffusive motion, the asymptotic rate of work is insensitive to the value of $a$. For general anharmonic traps, the asymptotic rate of work can be either increasing or decreasing as a function of the strength $a$, depending on the degree of anharmonicity. Counter to intuition, the rate of work can therefore in some cases increase as the resetting becomes weaker $(a\to 1)$ although the work vanishes at $a=1$. Work in the presence of a background potential is also considered. Numerical simulations confirm our findings.

cond-mat.stat-mech↗

arXiv2024

Interpretable User Satisfaction Estimation for Conversational Systems with Large Language Models

Accurate and interpretable user satisfaction estimation (USE) is critical for understanding, evaluating, and continuously improving conversational systems. Users express their satisfaction or dissatisfaction with diverse conversational patterns in both general-purpose (ChatGPT and Bing Copilot) and task-oriented (customer service chatbot) conversational systems. Existing approaches based on featurized ML models or text embeddings fall short in extracting generalizable patterns and are hard to interpret. In this work, we show that LLMs can extract interpretable signals of user satisfaction from their natural language utterances more effectively than embedding-based approaches. Moreover, an LLM can be tailored for USE via an iterative prompting framework using supervision from labeled examples. The resulting method, Supervised Prompting for User satisfaction Rubrics (SPUR), not only has higher accuracy but is more interpretable as it scores user satisfaction via learned rubrics with a detailed breakdown.

cs.IR↗

arXiv2024

Surgical Feature-Space Decomposition of LLMs: Why, When and How?

Low-rank approximations, of the weight and feature space can enhance the performance of deep learning models, whether in terms of improving generalization or reducing the latency of inference. However, there is no clear consensus yet on \emph{how}, \emph{when} and \emph{why} these approximations are helpful for large language models (LLMs). In this work, we empirically study the efficacy of weight and feature space decomposition in transformer-based LLMs. We demonstrate that surgical decomposition not only provides critical insights into the trade-off between compression and language modelling performance, but also sometimes enhances commonsense reasoning performance of LLMs. Our empirical analysis identifies specific network segments that intrinsically exhibit a low-rank structure. Furthermore, we extend our investigation to the implications of low-rank approximations on model bias. Overall, our findings offer a novel perspective on optimizing LLMs, presenting the low-rank approximation not only as a tool for performance enhancements, but also as a means to potentially rectify biases within these models. Our code is available at \href{https://github.com/nyunAI/SFSD-LLM}{GitHub}.

cs.CL↗

arXiv2024

Faster and Lighter LLMs: A Survey on Current Challenges and Way Forward

Despite the impressive performance of LLMs, their widespread adoption faces challenges due to substantial computational and memory requirements during inference. Recent advancements in model compression and system-level optimization methods aim to enhance LLM inference. This survey offers an overview of these methods, emphasizing recent developments. Through experiments on LLaMA(/2)-7B, we evaluate various compression techniques, providing practical insights for efficient LLM deployment in a unified setting. The empirical analysis on LLaMA(/2)-7B highlights the effectiveness of these methods. Drawing from survey insights, we identify current limitations and discuss potential future directions to improve LLM inference efficiency. We release the codebase to reproduce the results presented in this paper at https://github.com/nyunAI/Faster-LLM-Survey

cs.LG↗

arXiv2024

Stochastic resetting with refractory periods: pathway formulation and exact results

We look into the problem of stochastic resetting with refractory periods. The model dynamics comprises diffusive and motionless phases. The diffusive phase ends at random time instants, at which the system is reset to a given position -- where the system remains at rest for a random time interval, termed the refractory period. A pathway formulation is introduced to derive exact analytical results for the relevant observables in a broad framework, with the resetting time and the refractory period following arbitrary distributions. For the paradigmatic case of Poissonian distributions of the resetting and refractory times, in general with different characteristic rates, closed-form expressions are obtained that successfully describe the relaxation to the steady state. Finally, we focus on the single-target search problem, in which the survival probability and the mean first passage time to the target can be exactly computed. Therein, we also discuss optimal strategies, which show a non-trivial dependence on the refractory period.

cond-mat.stat-mech↗

arXiv2024

VERSE: Virtual-Gradient Aware Streaming Lifelong Learning with Anytime Inference

Lifelong learning or continual learning is the problem of training an AI agent continuously while also preventing it from forgetting its previously acquired knowledge. Streaming lifelong learning is a challenging setting of lifelong learning with the goal of continuous learning in a dynamic non-stationary environment without forgetting. We introduce a novel approach to lifelong learning, which is streaming (observes each training example only once), requires a single pass over the data, can learn in a class-incremental manner, and can be evaluated on-the-fly (anytime inference). To accomplish these, we propose a novel \emph{virtual gradients} based approach for continual representation learning which adapts to each new example while also generalizing well on past data to prevent catastrophic forgetting. Our approach also leverages an exponential-moving-average-based semantic memory to further enhance performance. Experiments on diverse datasets with temporally correlated observations demonstrate our method's efficacy and superior performance over existing methods.

cs.LG↗

arXiv2024

Beyond Uniform Scaling: Exploring Depth Heterogeneity in Neural Architectures

Conventional scaling of neural networks typically involves designing a base network and growing different dimensions like width, depth, etc. of the same by some predefined scaling factors. We introduce an automated scaling approach leveraging second-order loss landscape information. Our method is flexible towards skip connections a mainstay in modern vision transformers. Our training-aware method jointly scales and trains transformers without additional training iterations. Motivated by the hypothesis that not all neurons need uniform depth complexity, our approach embraces depth heterogeneity. Extensive evaluations on DeiT-S with ImageNet100 show a 2.5% accuracy gain and 10% parameter efficiency improvement over conventional scaling. Scaled networks demonstrate superior performance upon training small scale datasets from scratch. We introduce the first intact scaling mechanism for vision transformers, a step towards efficient model scaling.

cs.LG↗

arXiv2024

Self-healing Nodes with Adaptive Data-Sharding

Data sharding, a technique for partitioning and distributing data among multiple servers or nodes, offers enhancements in the scalability, performance, and fault tolerance of extensive distributed systems. Nonetheless, this strategy introduces novel challenges, including load balancing among shards, management of node failures and data loss, and adaptation to evolving data and workload patterns. This paper proposes an innovative approach to tackle these challenges by empowering self-healing nodes with adaptive data sharding. Leveraging concepts such as self-replication, fractal regeneration, sentient data sharding, and symbiotic node clusters, our approach establishes a dynamic and resilient data sharding scheme capable of addressing diverse scenarios and meeting varied requirements. Implementation and evaluation of our approach involve a prototype system simulating a large-scale distributed database across various data sharding scenarios. Comparative analyses against existing data sharding techniques highlight the superior scalability, performance, fault tolerance, and adaptability of our approach. Additionally, the paper delves into potential applications and limitations, providing insights into the future research directions that can further advance this innovative approach.

cs.DC↗

arXiv2023

Rethinking Compression: Reduced Order Modelling of Latent Features in Large Language Models

Due to the substantial scale of Large Language Models (LLMs), the direct application of conventional compression methodologies proves impractical. The computational demands associated with even minimal gradient updates present challenges, particularly on consumer-grade hardware. This paper introduces an innovative approach for the parametric and practical compression of LLMs based on reduced order modelling, which entails low-rank decomposition within the feature space and re-parameterization in the weight space. Notably, this compression technique operates in a layer-wise manner, obviating the need for a GPU device and enabling the compression of billion-scale models within stringent constraints of both memory and time. Our method represents a significant advancement in model compression by leveraging matrix decomposition, demonstrating superior efficacy compared to the prevailing state-of-the-art structured pruning method.

cs.LG↗

arXiv2023

Efficient Expansion and Gradient Based Task Inference for Replay Free Incremental Learning

This paper proposes a simple but highly efficient expansion-based model for continual learning. The recent feature transformation, masking and factorization-based methods are efficient, but they grow the model only over the global or shared parameter. Therefore, these approaches do not fully utilize the previously learned information because the same task-specific parameter forgets the earlier knowledge. Thus, these approaches show limited transfer learning ability. Moreover, most of these models have constant parameter growth for all tasks, irrespective of the task complexity. Our work proposes a simple filter and channel expansion based method that grows the model over the previous task parameters and not just over the global parameter. Therefore, it fully utilizes all the previously learned information without forgetting, which results in better knowledge transfer. The growth rate in our proposed model is a function of task complexity; therefore for a simple task, the model has a smaller parameter growth while for complex tasks, the model requires more parameters to adapt to the current task. Recent expansion based models show promising results for task incremental learning (TIL). However, for class incremental learning (CIL), prediction of task id is a crucial challenge; hence, their results degrade rapidly as the number of tasks increase. In this work, we propose a robust task prediction method that leverages entropy weighted data augmentations and the models gradient using pseudo labels. We evaluate our model on various datasets and architectures in the TIL, CIL and generative continual learning settings. The proposed approach shows state-of-the-art results in all these settings. Our extensive ablation studies show the efficacy of the proposed components.

cs.LG↗

arXiv2023

One-for-All: Generalized LoRA for Parameter-Efficient Fine-tuning

We present Generalized LoRA (GLoRA), an advanced approach for universal parameter-efficient fine-tuning tasks. Enhancing Low-Rank Adaptation (LoRA), GLoRA employs a generalized prompt module to optimize pre-trained model weights and adjust intermediate activations, providing more flexibility and capability across diverse tasks and datasets. Moreover, GLoRA facilitates efficient parameter adaptation by employing a scalable, modular, layer-wise structure search that learns individual adapter of each layer. Originating from a unified mathematical formulation, GLoRA exhibits strong transfer learning, few-shot learning and domain generalization abilities, as it adapts to new tasks through not only weights but also additional dimensions like activations. Comprehensive experiments demonstrate that GLoRA outperforms all previous methods in natural, specialized, and structured vision benchmarks, achieving superior accuracy with fewer parameters and computations. The proposed method on LLaMA-1 and LLaMA-2 also show considerable enhancements compared to the original LoRA in the language domain. Furthermore, our structural re-parameterization design ensures that GLoRA incurs no extra inference cost, rendering it a practical solution for resource-limited applications. Code and models are available at: https://github.com/Arnav0400/ViT-Slim/tree/master/GLoRA.

cs.LG↗