Revisit, Extend, and Enhance Hessian-Free Influence Functions
- URL: http://arxiv.org/abs/2405.17490v2
- Date: Sun, 20 Oct 2024 22:10:36 GMT
- Title: Revisit, Extend, and Enhance Hessian-Free Influence Functions
- Authors: Ziao Yang, Han Yue, Jian Chen, Hongfu Liu,
- Abstract summary: Influence functions serve as crucial tools for assessing sample influence in model interpretation, subset training set selection, and more.
In this paper, we revisit a specific, albeit effective approximation method known as Trac.
This method substitutes the inverse of the Hessian matrix with an identity matrix.
- Score: 26.105554752277648
- License:
- Abstract: Influence functions serve as crucial tools for assessing sample influence in model interpretation, subset training set selection, noisy label detection, and more. By employing the first-order Taylor extension, influence functions can estimate sample influence without the need for expensive model retraining. However, applying influence functions directly to deep models presents challenges, primarily due to the non-convex nature of the loss function and the large size of model parameters. This difficulty not only makes computing the inverse of the Hessian matrix costly but also renders it non-existent in some cases. Various approaches, including matrix decomposition, have been explored to expedite and approximate the inversion of the Hessian matrix, with the aim of making influence functions applicable to deep models. In this paper, we revisit a specific, albeit naive, yet effective approximation method known as TracIn. This method substitutes the inverse of the Hessian matrix with an identity matrix. We provide deeper insights into why this simple approximation method performs well. Furthermore, we extend its applications beyond measuring model utility to include considerations of fairness and robustness. Finally, we enhance TracIn through an ensemble strategy. To validate its effectiveness, we conduct experiments on synthetic data and extensive evaluations on noisy label detection, sample selection for large language model fine-tuning, and defense against adversarial attacks.
Related papers
- Influence Functions for Scalable Data Attribution in Diffusion Models [52.92223039302037]
Diffusion models have led to significant advancements in generative modelling.
Yet their widespread adoption poses challenges regarding data attribution and interpretability.
In this paper, we aim to help address such challenges by developing an textitinfluence functions framework.
arXiv Detail & Related papers (2024-10-17T17:59:02Z) - Most Influential Subset Selection: Challenges, Promises, and Beyond [9.479235005673683]
We study the Most Influential Subset Selection (MISS) problem, which aims to identify a subset of training samples with the greatest collective influence.
We conduct a comprehensive analysis of the prevailing approaches in MISS, elucidating their strengths and weaknesses.
We demonstrate that an adaptive version of theses which applies them iteratively, can effectively capture the interactions among samples.
arXiv Detail & Related papers (2024-09-25T20:00:23Z) - Outlier Gradient Analysis: Efficiently Identifying Detrimental Training Samples for Deep Learning Models [36.05242956018461]
In this paper, we establish a bridge between identifying detrimental training samples via influence functions and outlier gradient detection.
We first validate the hypothesis of our proposed outlier gradient analysis approach on synthetic datasets.
We then demonstrate its effectiveness in detecting mislabeled samples in vision models and selecting data samples for improving performance of natural language processing transformer models.
arXiv Detail & Related papers (2024-05-06T21:34:46Z) - Querying Easily Flip-flopped Samples for Deep Active Learning [63.62397322172216]
Active learning is a machine learning paradigm that aims to improve the performance of a model by strategically selecting and querying unlabeled data.
One effective selection strategy is to base it on the model's predictive uncertainty, which can be interpreted as a measure of how informative a sample is.
This paper proposes the it least disagree metric (LDM) as the smallest probability of disagreement of the predicted label.
arXiv Detail & Related papers (2024-01-18T08:12:23Z) - Self-Supervised Dataset Distillation for Transfer Learning [77.4714995131992]
We propose a novel problem of distilling an unlabeled dataset into a set of small synthetic samples for efficient self-supervised learning (SSL)
We first prove that a gradient of synthetic samples with respect to a SSL objective in naive bilevel optimization is textitbiased due to randomness originating from data augmentations or masking.
We empirically validate the effectiveness of our method on various applications involving transfer learning.
arXiv Detail & Related papers (2023-10-10T10:48:52Z) - Studying Large Language Model Generalization with Influence Functions [29.577692176892135]
Influence functions aim to answer a counterfactual: how would the model's parameters (and hence its outputs) change if a sequence were added to the training set?
We use the Eigenvalue-corrected Kronecker-Factored Approximate Curvature (EK-FAC) approximation to scale influence functions up to large language models (LLMs) with up to 52 billion parameters.
We investigate generalization patterns of LLMs, including the sparsity of the influence patterns, increasing abstraction with scale, math and programming abilities, cross-lingual generalization, and role-playing behavior.
arXiv Detail & Related papers (2023-08-07T04:47:42Z) - Efficient Multidimensional Functional Data Analysis Using Marginal
Product Basis Systems [2.4554686192257424]
We propose a framework for learning continuous representations from a sample of multidimensional functional data.
We show that the resulting estimation problem can be solved efficiently by the tensor decomposition.
We conclude with a real data application in neuroimaging.
arXiv Detail & Related papers (2021-07-30T16:02:15Z) - Tesseract: Tensorised Actors for Multi-Agent Reinforcement Learning [92.05556163518999]
MARL exacerbates matters by imposing various constraints on communication and observability.
For value-based methods, it poses challenges in accurately representing the optimal value function.
For policy gradient methods, it makes training the critic difficult and exacerbates the problem of the lagging critic.
We show that from a learning theory perspective, both problems can be addressed by accurately representing the associated action-value function.
arXiv Detail & Related papers (2021-05-31T23:08:05Z) - Hard-label Manifolds: Unexpected Advantages of Query Efficiency for
Finding On-manifold Adversarial Examples [67.23103682776049]
Recent zeroth order hard-label attacks on image classification models have shown comparable performance to their first-order, gradient-level alternatives.
It was recently shown in the gradient-level setting that regular adversarial examples leave the data manifold, while their on-manifold counterparts are in fact generalization errors.
We propose an information-theoretic argument based on a noisy manifold distance oracle, which leaks manifold information through the adversary's gradient estimate.
arXiv Detail & Related papers (2021-03-04T20:53:06Z) - Influence Functions in Deep Learning Are Fragile [52.31375893260445]
influence functions approximate the effect of samples in test-time predictions.
influence estimates are fairly accurate for shallow networks.
Hessian regularization is important to get highquality influence estimates.
arXiv Detail & Related papers (2020-06-25T18:25:59Z) - The Dilemma Between Data Transformations and Adversarial Robustness for
Time Series Application Systems [1.2056495277232115]
Adrial examples, or nearly indistinguishable inputs created by an attacker, significantly reduce machine learning accuracy.
This work explores how data transformations may impact an adversary's ability to create effective adversarial samples on a recurrent neural network.
A data transformation technique reduces the vulnerability to adversarial examples only if it approximates the dataset's intrinsic dimension.
arXiv Detail & Related papers (2020-06-18T22:43:37Z)
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.