Author name cluster

Sean Augenstein

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

4 papers

2 author rows

NeurIPS Conference 2025 Conference Paper

Gatekeeper: Improving Model Cascades Through Confidence Tuning

Stephan Rabanser
Nathalie Rauschmayr
Achin Kulshrestha
Petra Poklukar
Wittawat Jitkrittum
Sean Augenstein
Congchao Wang
Federico Tombari

Large-scale machine learning models deliver strong performance across a wide range of tasks but come with significant computational and resource constraints. To mitigate these challenges, local smaller models are often deployed alongside larger models, relying on routing and deferral mechanisms to offload complex tasks. However, existing approaches inadequately balance the capabilities of these models, often resulting in unnecessary deferrals or sub-optimal resource usage. In this work, we introduce a novel loss function called Gatekeeper for calibrating smaller models in cascade setups. Our approach fine-tunes the smaller model to confidently handle tasks it can perform correctly while deferring complex tasks to the larger model. Moreover, it incorporates a mechanism for managing the trade-off between model performance and deferral accuracy and is broadly applicable across various tasks and domains without any architectural changes. We evaluated our method on encoder-only, decoder-only, and encoder-decoder architectures. Experiments across image classification, language modeling, and vision-language tasks show that our approach substantially improves deferral performance.

PDF Details

ICLR Conference 2020 Conference Paper

Generative Models for Effective ML on Private, Decentralized Datasets

Sean Augenstein
H. Brendan McMahan
Daniel Ramage
Swaroop Ramaswamy
Peter Kairouz
Mingqing Chen
Rajiv Mathews
Blaise Agüera y Arcas

To improve real-world applications of machine learning, experienced modelers develop intuition about their datasets, their models, and how the two interact. Manual inspection of raw data—of representative samples, of outliers, of misclassifications—is an essential tool in a) identifying and fixing problems in the data, b) generating new modeling hypotheses, and c) assigning or refining human-provided labels. However, manual data inspection is risky for privacy-sensitive datasets, such as those representing the behavior of real-world individuals. Furthermore, manual data inspection is impossible in the increasingly important setting of federated learning, where raw examples are stored at the edge and the modeler may only access aggregated outputs such as metrics or model parameters. This paper demonstrates that generative models—trained using federated methods and with formal differential privacy guarantees—can be used effectively to debug data issues even when the data cannot be directly inspected. We explore these methods in applications to text with differentially private federated RNNs and to images using a novel algorithm for differentially private federated GANs.

Details

ICAPS Conference 2016 Conference Paper

Optimal Scheduling of a Constellation of Earth-Imaging Satellites, for Maximal Data Throughput and Efficient Human Management

Sean Augenstein
Alejandra Estanislao
Emmanuel Guere
Sean Blaes

A mixed-integer linear program (MILP) approach to scheduling a large constellation of Earth-imaging satellites is presented. The algorithm optimizes the assignment of imagery collects, image data downlinks, and "health & safety" contacts, generating schedules for all satellites and ground stations in a network. Hardware-driven constraints (e. g. , the limited agility of the satellites) and operations-driven constraints (e. g. , guaranteeing a minimum contact frequency for each satellite) are both addressed. Of critical importance to the use of this algorithm in real-world operations, it runs fast enough to allow for human operator interaction and repeated rescheduling. This is achieved by a partitioning of the problem into sequential steps for downlink scheduling and image scheduling, with a novel dynamic programming (DP) heuristic providing a stand-in for imaging activity in the MILP when scheduling the downlinks.

Details

ICRA Conference 2011 Conference Paper

Improved frame-to-frame pose tracking during vision-only SLAM/SFM with a tumbling target

Sean Augenstein
Stephen M. Rock

A hybrid algorithm for real-time frame-to-frame pose estimation during monocular vision-only SLAM/SFM is presented. The algorithm combines concepts from two existing approaches to pose tracking, Bayesian estimation methods and measurement inversion techniques, to achieve in real-time a feasible, smooth estimate of the relative pose between a robotic platform and a tumbling target. It is assumed that no a priori information about the target is available, and that only a monocular camera is available for measuring the relative motion of the target with respect to the robotic platform. The rationale for a hybrid approach is explained, and an algorithm is presented. A specific implementation using a modified Rao-Blackwellised particle filter is described and tested. Results from both numerical simulations and field experiments are included which demonstrate the performance and viability of the hybrid approach. The hybrid approach to pose estimation described here is applicable regardless of the method by which the map/reconstruction is estimated.

Details