Shubh Gupta

Nicholas Meisburger, Vihan Lakshman, Benito Geordie et al.

Efficient large-scale neural network training and inference on commodity CPU hardware is of immense practical significance in democratizing deep learning (DL) capabilities. Presently, the process of training massive models consisting of hundreds of millions to billions of parameters requires the extensive use of specialized hardware accelerators, such as GPUs, which are only accessible to a limited number of institutions with considerable financial resources. Moreover, there is often an alarming carbon footprint associated with training and deploying these models. In this paper, we take a step towards addressing these challenges by introducing BOLT, a sparse deep learning library for training large-scale search and recommendation models on standard CPU hardware. BOLT provides a flexible, high-level API for constructing models that will be familiar to users of existing popular DL frameworks. By automatically tuning specialized hyperparameters, BOLT also abstracts away the algorithmic details of sparse network training. We evaluate BOLT on a number of information retrieval tasks including product recommendations, text classification, graph neural networks, and personalization. We find that our proposed system achieves competitive performance with state-of-the-art techniques at a fraction of the cost and energy consumption and an order-of-magnitude faster inference time. BOLT has also been successfully deployed by multiple businesses to address critical problems, and we highlight one customer case study in the field of e-commerce.

Ishaan Rawal, Shubh Gupta, Yihan Hu et al.

Vision-Language-Action (VLA) models are advancing autonomous driving by replacing modular pipelines with unified end-to-end architectures. However, current VLAs face two expensive requirements: (1) massive dataset collection, and (2) dense reasoning annotations. In this work, we address both challenges with NORD (No Reasoning for Driving). Compared to existing VLAs, NORD achieves competitive performance while being fine-tuned on <60% of the data and no reasoning annotations, resulting in 3x fewer tokens. We identify that standard Group Relative Policy Optimization (GRPO) fails to yield significant improvements when applied to policies trained on such small, reasoning-free datasets. We show that this limitation stems from difficulty bias, which disproportionately penalizes reward signals from scenarios that produce high-variance rollouts within GRPO. NORD overcomes this by incorporating Dr. GRPO, a recent algorithm designed to mitigate difficulty bias in LLMs. As a result, NORD achieves competitive performance on Waymo and NAVSIM with a fraction of the training data and no reasoning overhead, enabling more efficient autonomous systems. Website: https://nord-vla-ai.github.io/

Anthony Corso, Kyu-Young Kim, Shubh Gupta et al.

An important step in the design of autonomous systems is to evaluate the probability that a failure will occur. In safety-critical domains, the failure probability is extremely small so that the evaluation of a policy through Monte Carlo sampling is inefficient. Adaptive importance sampling approaches have been developed for rare event estimation but do not scale well to sequential systems with long horizons. In this work, we develop two adaptive importance sampling algorithms that can efficiently estimate the probability of rare events for sequential decision making systems. The basis for these algorithms is the minimization of the Kullback-Leibler divergence between a state-dependent proposal distribution and a target distribution over trajectories, but the resulting algorithms resemble policy gradient and value-based reinforcement learning. We apply multiple importance sampling to reduce the variance of our estimate and to address the issue of multi-modality in the optimal proposal distribution. We demonstrate our approach on a control task with both continuous and discrete actions spaces and show accuracy improvements over several baselines.

Adam Dai, Shubh Gupta, Grace Gao · amazon-science

Autonomous vehicles such as the Mars rovers currently lead the vanguard of surface exploration on extraterrestrial planets and moons. In order to accelerate the pace of exploration and science objectives, it is critical to plan safe and efficient paths for these vehicles. However, current rover autonomy is limited by a lack of global maps which can be easily constructed and stored for onboard re-planning. Recently, Neural Radiance Fields (NeRFs) have been introduced as a detailed 3D scene representation which can be trained from sparse 2D images and efficiently stored. We propose to use NeRFs to construct maps for online use in autonomous navigation, and present a planning framework which leverages the NeRF map to integrate local and global information. Our approach interpolates local cost observations across global regions using kernel ridge regression over terrain features extracted from the NeRF map, allowing the rover to re-route itself around untraversable areas discovered during online operation. We validate our approach in high-fidelity simulation and demonstrate lower cost and higher percentage success rate path planning compared to various baselines.

Ashwin V. Kanhere, Shubh Gupta, Akshay Shetty et al.

Deep Neural Networks (DNNs) are a promising tool for Global Navigation Satellite System (GNSS) positioning in the presence of multipath and non-line-of-sight errors, owing to their ability to model complex errors using data. However, developing a DNN for GNSS positioning presents various challenges, such as 1) poor numerical conditioning caused by large variations in measurements and position values across the globe, 2) varying number and order within the set of measurements due to changing satellite visibility, and 3) overfitting to available data. In this work, we address the aforementioned challenges and propose an approach for GNSS positioning by applying DNN-based corrections to an initial position guess. Our DNN learns to output the position correction using the set of pseudorange residuals and satellite line-of-sight vectors as inputs. The limited variation in these input and output values improves the numerical conditioning for our DNN. We design our DNN architecture to combine information from the available GNSS measurements, which vary both in number and order, by leveraging recent advancements in set-based deep learning methods. Furthermore, we present a data augmentation strategy for reducing overfitting in the DNN by randomizing the initial position guesses. We first perform simulations and show an improvement in the initial positioning error when our DNN-based corrections are applied. After this, we demonstrate that our approach outperforms a WLS baseline on real-world data. Our implementation is available at github.com/Stanford-NavLab/deep_gnss.

Navyansh Mahla, Annie D'souza, Shubh Gupta et al.

The application of large-scale models in medical image segmentation demands substantial quantities of meticulously annotated data curated by experts along with high computational resources, both of which are challenges in resource-poor settings. In this study, we present the Medical Segment Anything Model with Galore MedSAGa where we adopt the Segment Anything Model (SAM) to achieve memory-efficient, few-shot medical image segmentation by applying Gradient Low-Rank Projection GaLore to the parameters of the image encoder of SAM. Meanwhile, the weights of the prompt encoder and mask decoder undergo full parameter fine-tuning using standard optimizers. We further assess MedSAGa's few-shot learning capabilities, reporting on its memory efficiency and segmentation performance across multiple standard medical image segmentation datasets. We compare it with several baseline models, including LoRA fine-tuned SAM (SAMed) and DAE-Former. Experiments across multiple datasets and these baseline models with different number of images for fine tuning demonstrated that the GPU memory consumption of MedSAGa is significantly less than that of the baseline models, achieving an average memory efficiency of 66% more than current state-of-the-art (SOTA) models for medical image segmentation. The combination of substantially lower memory requirements and comparable to SOTA results in few-shot learning for medical image segmentation positions MedSAGa as an optimal solution for deployment in resource-constrained settings.

Suraj Racha, Shubh Gupta, Humaira Firdowse et al.

Federated Learning (FL) is a pioneering approach in distributed machine learning, enabling collaborative model training across multiple clients while retaining data privacy. However, the inherent heterogeneity due to imbalanced resource representations across multiple clients poses significant challenges, often introducing bias towards the majority class. This issue is particularly prevalent in healthcare settings, where hospitals acting as clients share medical images. To address class imbalance and reduce bias, we propose a co-distillation driven framework in a federated healthcare setting. Unlike traditional federated setups with a designated server client, our framework promotes knowledge sharing among clients to collectively improve learning outcomes. Our experiments demonstrate that in a federated healthcare setting, co-distillation outperforms other federated methods in handling class imbalance. Additionally, we demonstrate that our framework has the least standard deviation with increasing imbalance while outperforming other baselines, signifying the robustness of our framework for FL in healthcare.

Shubh Gupta, Grace X. Gao

For reliable operation on urban roads, navigation using the Global Navigation Satellite System (GNSS) requires both accurately estimating the positioning detail from GNSS pseudorange measurements and determining when the estimated position is safe to use, or available. However, multiple GNSS measurements in urban environments contain biases, or faults, due to signal reflection and blockage from nearby buildings which are difficult to mitigate for estimating the position and availability. This paper proposes a novel particle filter-based framework that employs a Gaussian Mixture Model (GMM) likelihood of GNSS measurements to robustly estimate the position of a navigating vehicle under multiple measurement faults. Using the probability distribution tracked by the filter and the designed GMM likelihood, we measure the accuracy and the risk associated with localization and determine the availability of the navigation system at each time instant. Through experiments conducted on challenging simulated and real urban driving scenarios, we show that our method achieves small horizontal positioning errors compared to existing filter-based state estimation techniques when multiple GNSS measurements contain faults. Furthermore, we verify using several simulations that our method determines system availability with smaller probability of false alarms and integrity risk than the existing particle filter-based integrity monitoring approach.

Shubh Gupta, Grace X. Gao

Reliably assessing the error in an estimated vehicle position is integral for ensuring the vehicle's safety in urban environments. Many existing approaches use GNSS measurements to characterize protection levels (PLs) as probabilistic upper bounds on the position error. However, GNSS signals might be reflected or blocked in urban environments, and thus additional sensor modalities need to be considered to determine PLs. In this paper, we propose a novel approach for computing PLs by matching camera image measurements to a LiDAR-based 3D map of the environment. We specify a Gaussian mixture model probability distribution of position error using deep neural network-based data-driven models and statistical outlier weighting techniques. From the probability distribution, we compute the PLs by evaluating the position error bound using numerical line-search methods. Through experimental validation with real-world data, we demonstrate that the PLs computed from our method are reliable bounds on the position error in urban environments.

Adyasha Mohanty, Shubh Gupta, Grace Xingxin Gao

Adopting a joint approach towards state estimation and integrity monitoring results in unbiased integrity monitoring unlike traditional approaches. So far, a joint approach was used in Particle RAIM [l] for GNSS measurements only. In our work, we extend Particle RAIM to a GNSS-camera fused system for joint state estimation and integrity monitoring. To account for vision faults, we derive a probability distribution over position from camera images using map-matching. We formulate a Kullback-Leibler Divergence metric to assess the consistency of GNSS and camera measurements and mitigate faults during sensor fusion. The derived integrity risk upper bounds the probability of Hazardously Misleading Information (HMI). Experimental validation on a real-world dataset shows that our algorithm produces less than 11 m position error and the integrity risk over bounds the probability of HMI with 0.11 failure rate for an 8 m Alert Limit in an urban scenario.

Nikita Jaipuria, Shubh Gupta, Praveen Narayanan et al.

Generative Adversarial Networks (GANs) are now widely used for photo-realistic image synthesis. In applications where a simulated image needs to be translated into a realistic image (sim-to-real), GANs trained on unpaired data from the two domains are susceptible to failure in semantic content retention as the image is translated from one domain to the other. This failure mode is more pronounced in cases where the real data lacks content diversity, resulting in a content \emph{mismatch} between the two domains - a situation often encountered in real-world deployment. In this paper, we investigate the role of the discriminator's receptive field in GANs for unsupervised image-to-image translation with mismatched data, and study its effect on semantic content retention. Experiments with the discriminator architecture of a state-of-the-art coupled Variational Auto-Encoder (VAE) - GAN model on diverse, mismatched datasets show that the discriminator receptive field is directly correlated with semantic content discrepancy of the generated image.

Prakhar Gupta, Shubh Gupta, Ajaykrishnan Jayagopal et al.

We introduce models for saliency prediction for mobile user interfaces. A mobile interface may include elements like buttons, text, etc. in addition to natural images which enable performing a variety of tasks. Saliency in natural images is a well studied area. However, given the difference in what constitutes a mobile interface, and the usage context of these devices, we postulate that saliency prediction for mobile interface images requires a fresh approach. Mobile interface design involves operating on elements, the building blocks of the interface. We first collected eye-gaze data from mobile devices for free viewing task. Using this data, we develop a novel autoencoder based multi-scale deep learning model that provides saliency prediction at the mobile interface element level. Compared to saliency prediction approaches developed for natural images, we show that our approach performs significantly better on a range of established metrics.