Don't blame Dataset Shift! Shortcut Learning due to Gradients and Cross Entropy

• URL: http://arxiv.org/abs/2308.12553v1
• Date: Thu, 24 Aug 2023 04:39:25 GMT
• Title: Don't blame Dataset Shift! Shortcut Learning due to Gradients and Cross Entropy
• Authors: Aahlad Puli, Lily Zhang, Yoav Wald, Rajesh Ranganath
• Abstract summary: We show that default-ERM's preference for maximizing the margin leads to models that depend more on the shortcut than the stable feature. We develop loss functions that encourage uniform-margin solutions, called margin control (MARG-CTRL)
• Score: 22.69591517487717
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Common explanations for shortcut learning assume that the shortcut improves prediction under the training distribution but not in the test distribution. Thus, models trained via the typical gradient-based optimization of cross-entropy, which we call default-ERM, utilize the shortcut. However, even when the stable feature determines the label in the training distribution and the shortcut does not provide any additional information, like in perception tasks, default-ERM still exhibits shortcut learning. Why are such solutions preferred when the loss for default-ERM can be driven to zero using the stable feature alone? By studying a linear perception task, we show that default-ERM's preference for maximizing the margin leads to models that depend more on the shortcut than the stable feature, even without overparameterization. This insight suggests that default-ERM's implicit inductive bias towards max-margin is unsuitable for perception tasks. Instead, we develop an inductive bias toward uniform margins and show that this bias guarantees dependence only on the perfect stable feature in the linear perception task. We develop loss functions that encourage uniform-margin solutions, called margin control (MARG-CTRL). MARG-CTRL mitigates shortcut learning on a variety of vision and language tasks, showing that better inductive biases can remove the need for expensive two-stage shortcut-mitigating methods in perception tasks.

Related papers

• Model Surgery: Modulating LLM's Behavior Via Simple Parameter Editing [63.20133320524577]
  Large Language Models (LLMs) have demonstrated great potential as generalist assistants. It is crucial that these models exhibit desirable behavioral traits, such as non-toxicity and resilience against jailbreak attempts. In this paper, we observe that directly editing a small subset of parameters can effectively modulate specific behaviors of LLMs.
  arXiv  Detail & Related papers  (2024-07-11T17:52:03Z)
• Not Eliminate but Aggregate: Post-Hoc Control over Mixture-of-Experts to Address Shortcut Shifts in Natural Language Understanding [5.4480125359160265]
  We propose a pessimistically aggregating the predictions of a mixture-of-experts, assuming each expert captures relatively different latent features. The experimental results demonstrate that our post-hoc control over the experts significantly enhances the model's robustness to the distribution shift in shortcuts.
  arXiv  Detail & Related papers  (2024-06-17T20:00:04Z)
• RESTORE: Towards Feature Shift for Vision-Language Prompt Learning [33.13407089704543]
  We show that prompt tuning along only one branch of CLIP is the reason why the misalignment occurs. Without proper regularization across the learnable parameters in different modalities, prompt learning violates the original pre-training constraints. We propose RESTORE, a multi-modal prompt learning method that exerts explicit constraints on cross-modal consistency.
  arXiv  Detail & Related papers  (2024-03-10T08:52:48Z)
• Marginal Debiased Network for Fair Visual Recognition [59.05212866862219]
  We propose a novel marginal debiased network (MDN) to learn debiased representations. Our MDN can achieve a remarkable performance on under-represented samples.
  arXiv  Detail & Related papers  (2024-01-04T08:57:09Z)
• Benign Shortcut for Debiasing: Fair Visual Recognition via Intervention with Shortcut Features [47.01860331227165]
  We propose emphShortcut Debiasing, to first transfer the target task's learning of bias attributes from bias features to shortcut features. We achieve significant improvements over the state-of-the-art debiasing methods in both accuracy and fairness.
  arXiv  Detail & Related papers  (2023-08-13T00:40:22Z)
• Chroma-VAE: Mitigating Shortcut Learning with Generative Classifiers [44.97660597940641]
  We show that generative models alone are not sufficient to prevent shortcut learning. In particular, we propose Chroma-VAE, a two-pronged approach where a VAE is initially trained to isolate the shortcut in a small latent subspace. In addition to demonstrating the efficacy of Chroma-VAE on benchmark and real-world shortcut learning tasks, our work highlights the potential for manipulating the latent space of generative classifiers to isolate or interpret specific correlations.
  arXiv  Detail & Related papers  (2022-11-28T11:27:50Z)
• Debiasing Meta-Gradient Reinforcement Learning by Learning the Outer Value Function [69.59204851882643]
  We identify a bias in the meta-gradient of current meta-gradient RL approaches. This bias comes from using the critic that is trained using the meta-learned discount factor for the advantage estimation in the outer objective. Because the meta-learned discount factor is typically lower than the one used in the outer objective, the resulting bias can cause the meta-gradient to favor myopic policies.
  arXiv  Detail & Related papers  (2022-11-19T00:59:20Z)
• ELM: Embedding and Logit Margins for Long-Tail Learning [70.19006872113862]
  Long-tail learning is the problem of learning under skewed label distributions. We present Embedding and Logit Margins (ELM), a unified approach to enforce margins in logit space. The ELM method is shown to perform well empirically, and results in tighter tail class embeddings.
  arXiv  Detail & Related papers  (2022-04-27T21:53:50Z)
• Margin Calibration for Long-Tailed Visual Recognition [14.991077564590128]
  We study the relationship between the margins and logits (classification scores) and empirically observe the biased margins and the biased logits are positively correlated. We propose MARC, a simple yet effective MARgin function to dynamically calibrate the biased margins for unbiased logits.
  arXiv  Detail & Related papers  (2021-12-14T08:25:29Z)
• Towards Interpreting and Mitigating Shortcut Learning Behavior of NLU models [53.36605766266518]
  We show that trained NLU models have strong preference for features located at the head of the long-tailed distribution. We propose a shortcut mitigation framework, to suppress the model from making overconfident predictions for samples with large shortcut degree.
  arXiv  Detail & Related papers  (2021-03-11T19:39:56Z)
• Direction Matters: On the Implicit Bias of Stochastic Gradient Descent with Moderate Learning Rate [105.62979485062756]
  This paper attempts to characterize the particular regularization effect of SGD in the moderate learning rate regime. We show that SGD converges along the large eigenvalue directions of the data matrix, while GD goes after the small eigenvalue directions.
  arXiv  Detail & Related papers  (2020-11-04T21:07:52Z)

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.