Stabilizing Linear Passive-Aggressive Online Learning with Weighted Reservoir Sampling

• URL: http://arxiv.org/abs/2410.23601v1
• Date: Thu, 31 Oct 2024 03:35:48 GMT
• Title: Stabilizing Linear Passive-Aggressive Online Learning with Weighted Reservoir Sampling
• Authors: Skyler Wu, Fred Lu, Edward Raff, James Holt,
• Abstract summary: Online learning methods are still highly effective for high-dimensional streaming data, out-of-core processing, and other throughput-sensitive applications. Many such algorithms rely on fast adaptation to individual errors as a key to their convergence. While such algorithms enjoy low theoretical regret, in real-world deployment they can be sensitive to individual outliers that cause the algorithm to over-correct.
• Score: 46.01254613933967
• License:
• Abstract: Online learning methods, like the seminal Passive-Aggressive (PA) classifier, are still highly effective for high-dimensional streaming data, out-of-core processing, and other throughput-sensitive applications. Many such algorithms rely on fast adaptation to individual errors as a key to their convergence. While such algorithms enjoy low theoretical regret, in real-world deployment they can be sensitive to individual outliers that cause the algorithm to over-correct. When such outliers occur at the end of the data stream, this can cause the final solution to have unexpectedly low accuracy. We design a weighted reservoir sampling (WRS) approach to obtain a stable ensemble model from the sequence of solutions without requiring additional passes over the data, hold-out sets, or a growing amount of memory. Our key insight is that good solutions tend to be error-free for more iterations than bad solutions, and thus, the number of passive rounds provides an estimate of a solution's relative quality. Our reservoir thus contains $K$ previous intermediate weight vectors with high survival times. We demonstrate our WRS approach on the Passive-Aggressive Classifier (PAC) and First-Order Sparse Online Learning (FSOL), where our method consistently and significantly outperforms the unmodified approach. We show that the risk of the ensemble classifier is bounded with respect to the regret of the underlying online learning method.

Related papers

• Adaptive Robust Learning using Latent Bernoulli Variables [50.223140145910904]
  We present an adaptive approach for learning from corrupted training sets. We identify corrupted non-corrupted samples with latent Bernoulli variables. The resulting problem is solved via variational inference.
  arXiv  Detail & Related papers  (2023-12-01T13:50:15Z)
• Active Learning in the Predict-then-Optimize Framework: A Margin-Based Approach [5.371816551086118]
  We develop a learning method that sequentially decides whether to request the "labels" of feature samples from an unlabeled data stream. Our active learning method is the first to be directly informed by the decision error induced by the predicted parameters.
  arXiv  Detail & Related papers  (2023-05-11T05:44:36Z)
• MAPS: A Noise-Robust Progressive Learning Approach for Source-Free Domain Adaptive Keypoint Detection [76.97324120775475]
  Cross-domain keypoint detection methods always require accessing the source data during adaptation. This paper considers source-free domain adaptive keypoint detection, where only the well-trained source model is provided to the target domain.
  arXiv  Detail & Related papers  (2023-02-09T12:06:08Z)
• Can we achieve robustness from data alone? [0.7366405857677227]
  Adversarial training and its variants have come to be the prevailing methods to achieve adversarially robust classification using neural networks. We devise a meta-learning method for robust classification, that optimize the dataset prior to its deployment in a principled way. Experiments on MNIST and CIFAR-10 demonstrate that the datasets we produce enjoy very high robustness against PGD attacks.
  arXiv  Detail & Related papers  (2022-07-24T12:14:48Z)
• Log Barriers for Safe Black-box Optimization with Application to Safe Reinforcement Learning [72.97229770329214]
  We introduce a general approach for seeking high dimensional non-linear optimization problems in which maintaining safety during learning is crucial. Our approach called LBSGD is based on applying a logarithmic barrier approximation with a carefully chosen step size. We demonstrate the effectiveness of our approach on minimizing violation in policy tasks in safe reinforcement learning.
  arXiv  Detail & Related papers  (2022-07-21T11:14:47Z)
• Bilevel Online Deep Learning in Non-stationary Environment [4.565872584112864]
  Bilevel Online Deep Learning (BODL) framework combines bilevel optimization strategy and online ensemble classifier. When the concept drift is detected, our BODL algorithm can adaptively update the model parameters via bilevel optimization and then circumvent the large drift and encourage positive transfer.
  arXiv  Detail & Related papers  (2022-01-25T11:05:51Z)
• Simple Stochastic and Online Gradient DescentAlgorithms for Pairwise Learning [65.54757265434465]
  Pairwise learning refers to learning tasks where the loss function depends on a pair instances. Online descent (OGD) is a popular approach to handle streaming data in pairwise learning. In this paper, we propose simple and online descent to methods for pairwise learning.
  arXiv  Detail & Related papers  (2021-11-23T18:10:48Z)
• Attentional-Biased Stochastic Gradient Descent [74.49926199036481]
  We present a provable method (named ABSGD) for addressing the data imbalance or label noise problem in deep learning. Our method is a simple modification to momentum SGD where we assign an individual importance weight to each sample in the mini-batch. ABSGD is flexible enough to combine with other robust losses without any additional cost.
  arXiv  Detail & Related papers  (2020-12-13T03:41:52Z)
• Low-Rank Robust Online Distance/Similarity Learning based on the Rescaled Hinge Loss [0.34376560669160383]
  Existing online methods usually assume training triplets or pairwise constraints are exist in advance. We formulate the online Distance-Similarity learning problem with the robust Rescaled hinge loss function. The proposed model is rather general and can be applied to any PA-based online Distance-Similarity algorithm.
  arXiv  Detail & Related papers  (2020-10-07T08:38:34Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.