Author name cluster

Florian Jaeckle

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.

3 papers

2 author rows

ICLR Conference 2023 Conference Paper

FastFill: Efficient Compatible Model Update

Florian Jaeckle
Fartash Faghri
Ali Farhadi
Oncel Tuzel
Hadi Pouransari

In many retrieval systems the original high dimensional data (e.g., images) is mapped to a lower dimensional feature through a learned embedding model. The task of retrieving the most similar data from a gallery set to a given query data is performed through similarity comparison on features. When the embedding model is updated, it might produce features that are not comparable/compatible with features already in the gallery computed with the old model. Subsequently, all features in the gallery need to be re-computed using the new embedding model -- a computationally expensive process called backfilling. Recently, compatible representation learning methods have been proposed to avoid back-filling. Despite their relative success, there is an inherent trade-off between new model performance and its compatibility with the old model. In this work, we introduce FastFill: a compatible model update process using feature alignment and policy based partial backfilling to promptly elevate retrieval performance. We show that previous backfilling strategies suffer from decreased performance and demonstrate the importance of both the training objective and the ordering in online partial backfilling. We propose a new training method for feature alignment between old and new embedding models using uncertainty estimation. Compared to previous works, we obtain significantly improved backfilling results on a variety of datasets: mAP on ImageNet (+4.4%), Places-365 (+2.7%), and VGG-Face2 (+1.3%). Further, we demonstrate that when updating a biased model with FastFill, the minority subgroup accuracy gap promptly vanishes with a small fraction of partial backfilling.

Details

UAI Conference 2021 Conference Paper

Generating adversarial examples with graph neural networks

Florian Jaeckle
M. Pawan Kumar

Recent years have witnessed the deployment of adversarial attacks to evaluate the robustness of Neural Networks. Past work in this field has relied on traditional optimization algorithms that ignore the inherent structure of the problem and data, or generative methods that rely purely on learning and often fail to generate adversarial examples where they are hard to find. To alleviate these deficiencies, we propose a novel attack based on a graph neural network (GNN) that takes advantage of the strengths of both approaches; we call it AdvGNN. Our GNN architecture closely resembles the network we wish to attack. During inference, we perform forward-backward passes through the GNN layers to guide an iterative procedure towards adversarial examples. During training, its parameters are estimated via a loss function that encourages the efficient computation of adversarial examples over a time horizon. We show that our method beats state-of-the-art adversarial attacks, including PGD-attack, MI-FGSM, and Carlini and Wagner attack, reducing the time required to generate adversarial examples with small perturbation norms by over 65%. Moreover, AdvGNN achieves good generalization performance on unseen networks. Finally, we provide a new challenging dataset specifically designed to allow for a more illustrative comparison of adversarial attacks.

Details

AAAI Conference 2018 Conference Paper

On Recognising Nearly Single-Crossing Preferences

Florian Jaeckle
Dominik Peters
Edith Elkind

If voters’ preferences are one-dimensional, many hard problems in computational social choice become tractable. A preference proﬁle can be classiﬁed as one-dimensional if it has the singlecrossing property, which requires that the voters can be ordered from left to right so that their preferences are consistent with this order. In practice, preferences may exhibit some onedimensional structure, despite not being single-crossing in the formal sense. Hence, we ask whether one can identify preference proﬁles that are close to being single-crossing. We consider three distance measures, which are based on partitioning voters or candidates or performing a small number of swaps in each vote. We prove that it can be eﬃciently decided if voters can be split into two single-crossing groups. Also, for every ﬁxed k ⩾ 1 we can decide in polynomial time if a proﬁle can be made single-crossing by performing at most k candidate swaps per vote. In contrast, for each k ⩾ 3 it is NP-complete to decide whether candidates can be partitioned into k sets so that the restriction of the input proﬁle to each set is single-crossing.

PDF Details