Predicting Out-of-Domain Generalization with Neighborhood Invariance
- URL: http://arxiv.org/abs/2207.02093v3
- Date: Mon, 17 Jul 2023 15:16:56 GMT
- Title: Predicting Out-of-Domain Generalization with Neighborhood Invariance
- Authors: Nathan Ng and Neha Hulkund and Kyunghyun Cho and Marzyeh Ghassemi
- Abstract summary: We propose a measure of a classifier's output invariance in a local transformation neighborhood.
Our measure is simple to calculate, does not depend on the test point's true label, and can be applied even in out-of-domain (OOD) settings.
In experiments on benchmarks in image classification, sentiment analysis, and natural language inference, we demonstrate a strong and robust correlation between our measure and actual OOD generalization.
- Score: 59.05399533508682
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: Developing and deploying machine learning models safely depends on the
ability to characterize and compare their abilities to generalize to new
environments. Although recent work has proposed a variety of methods that can
directly predict or theoretically bound the generalization capacity of a model,
they rely on strong assumptions such as matching train/test distributions and
access to model gradients. In order to characterize generalization when these
assumptions are not satisfied, we propose neighborhood invariance, a measure of
a classifier's output invariance in a local transformation neighborhood.
Specifically, we sample a set of transformations and given an input test point,
calculate the invariance as the largest fraction of transformed points
classified into the same class. Crucially, our measure is simple to calculate,
does not depend on the test point's true label, makes no assumptions about the
data distribution or model, and can be applied even in out-of-domain (OOD)
settings where existing methods cannot, requiring only selecting a set of
appropriate data transformations. In experiments on robustness benchmarks in
image classification, sentiment analysis, and natural language inference, we
demonstrate a strong and robust correlation between our neighborhood invariance
measure and actual OOD generalization on over 4,600 models evaluated on over
100 unique train/test domain pairs.
Related papers
- Conformal Inference for Invariant Risk Minimization [12.049545417799125]
The application of machine learning models can be significantly impeded by the occurrence of distributional shifts.
One way to tackle this problem is to use invariant learning, such as invariant risk minimization (IRM), to acquire an invariant representation.
This paper develops methods for obtaining distribution-free prediction regions to describe uncertainty estimates for invariant representations.
arXiv Detail & Related papers (2023-05-22T03:48:38Z) - Variational Classification [51.2541371924591]
We derive a variational objective to train the model, analogous to the evidence lower bound (ELBO) used to train variational auto-encoders.
Treating inputs to the softmax layer as samples of a latent variable, our abstracted perspective reveals a potential inconsistency.
We induce a chosen latent distribution, instead of the implicit assumption found in a standard softmax layer.
arXiv Detail & Related papers (2023-05-17T17:47:19Z) - Malign Overfitting: Interpolation Can Provably Preclude Invariance [30.776243638012314]
We show that "benign overfitting" in which models generalize well despite interpolating might not favorably extend to settings in which robustness or fairness are desirable.
We propose and analyze an algorithm that successfully learns a non-interpolating classifier that is provably invariant.
arXiv Detail & Related papers (2022-11-28T19:17:31Z) - Learning Invariant Representations under General Interventions on the
Response [2.725698729450241]
We focus on linear structural causal models (SCMs) and introduce invariant matching property (IMP)
We analyze the generalization errors of our method under both the discrete and continuous environment settings.
arXiv Detail & Related papers (2022-08-22T03:09:17Z) - On the Strong Correlation Between Model Invariance and Generalization [54.812786542023325]
Generalization captures a model's ability to classify unseen data.
Invariance measures consistency of model predictions on transformations of the data.
From a dataset-centric view, we find a certain model's accuracy and invariance linearly correlated on different test sets.
arXiv Detail & Related papers (2022-07-14T17:08:25Z) - Predicting with Confidence on Unseen Distributions [90.68414180153897]
We connect domain adaptation and predictive uncertainty literature to predict model accuracy on challenging unseen distributions.
We find that the difference of confidences (DoC) of a classifier's predictions successfully estimates the classifier's performance change over a variety of shifts.
We specifically investigate the distinction between synthetic and natural distribution shifts and observe that despite its simplicity DoC consistently outperforms other quantifications of distributional difference.
arXiv Detail & Related papers (2021-07-07T15:50:18Z) - Domain Conditional Predictors for Domain Adaptation [3.951376400628575]
We consider a conditional modeling approach in which predictions, in addition to being dependent on the input data, use information relative to the underlying data-generating distribution.
We argue that such an approach is more generally applicable than current domain adaptation methods.
arXiv Detail & Related papers (2021-06-25T22:15:54Z) - Performance-Agnostic Fusion of Probabilistic Classifier Outputs [2.4206828137867107]
We propose a method for combining probabilistic outputs of classifiers to make a single consensus class prediction.
Our proposed method works well in situations where accuracy is the performance metric.
It does not output calibrated probabilities, so it is not suitable in situations where such probabilities are required for further processing.
arXiv Detail & Related papers (2020-09-01T16:53:29Z) - Estimating Generalization under Distribution Shifts via Domain-Invariant
Representations [75.74928159249225]
We use a set of domain-invariant predictors as a proxy for the unknown, true target labels.
The error of the resulting risk estimate depends on the target risk of the proxy model.
arXiv Detail & Related papers (2020-07-06T17:21:24Z) - Good Classifiers are Abundant in the Interpolating Regime [64.72044662855612]
We develop a methodology to compute precisely the full distribution of test errors among interpolating classifiers.
We find that test errors tend to concentrate around a small typical value $varepsilon*$, which deviates substantially from the test error of worst-case interpolating model.
Our results show that the usual style of analysis in statistical learning theory may not be fine-grained enough to capture the good generalization performance observed in practice.
arXiv Detail & Related papers (2020-06-22T21:12:31Z)
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.