Information-Theoretic Measures of Dataset Difficulty

• URL: http://arxiv.org/abs/2110.08420v1
• Date: Sat, 16 Oct 2021 00:21:42 GMT
• Title: Information-Theoretic Measures of Dataset Difficulty
• Authors: Kawin Ethayarajh, Yejin Choi, Swabha Swayamdipta
• Abstract summary: Estimating difficulty of a dataset typically involves comparing state-of-the-art models to humans. We propose an information-theoretic perspective, framing dataset difficulty as the absence of usable information.
• Score: 54.538766940287864
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Estimating the difficulty of a dataset typically involves comparing state-of-the-art models to humans; the bigger the performance gap, the harder the dataset is said to be. Not only is this framework informal, but it also provides little understanding of how difficult each instance is, or what attributes make it difficult for a given model. To address these problems, we propose an information-theoretic perspective, framing dataset difficulty as the absence of $\textit{usable information}$. Measuring usable information is as easy as measuring performance, but has certain theoretical advantages. While the latter only allows us to compare different models w.r.t the same dataset, the former also allows us to compare different datasets w.r.t the same model. We then introduce $\textit{pointwise}$ $\mathcal{V}-$$\textit{information}$ (PVI) for measuring the difficulty of individual instances, where instances with higher PVI are easier for model $\mathcal{V}$. By manipulating the input before measuring usable information, we can understand $\textit{why}$ a dataset is easy or difficult for a given model, which we use to discover annotation artefacts in widely-used benchmarks.

Related papers

• Inverse Entropic Optimal Transport Solves Semi-supervised Learning via Data Likelihood Maximization [65.8915778873691]
  conditional distributions is a central problem in machine learning. We propose a new learning paradigm that integrates both paired and unpaired data. Our approach also connects intriguingly with inverse entropic optimal transport (OT)
  arXiv  Detail & Related papers  (2024-10-03T16:12:59Z)
• Making Multi-Axis Gaussian Graphical Models Scalable to Millions of Samples and Features [0.30723404270319693]
  We introduce a method that has $O(n2)$ runtime and $O(n)$ space complexity, without assuming independence. We demonstrate that our approach can be used on unprecedentedly large datasets, such as a real-world 1,000,000-cell scRNA-seq dataset.
  arXiv  Detail & Related papers  (2024-07-29T11:15:25Z)
• $\textit{GeoHard}$: Towards Measuring Class-wise Hardness through Modelling Class Semantics [90.9047957137981]
  This work formally initiates the concept of $textitclass-wise hardness$. Experiments across eight natural language understanding (NLU) datasets demonstrate a consistent hardness distribution across learning paradigms, models, and human judgment. $textitGeoHard$ surpasses instance-level metrics by over 59 percent on $textitPearson$'s correlation on measuring class-wise hardness.
  arXiv  Detail & Related papers  (2024-07-17T11:53:39Z)
• Scaling Laws for Data Filtering -- Data Curation cannot be Compute Agnostic [99.3682210827572]
  Vision-language models (VLMs) are trained for thousands of GPU hours on carefully curated web datasets. Data curation strategies are typically developed agnostic of the available compute for training. We introduce neural scaling laws that account for the non-homogeneous nature of web data.
  arXiv  Detail & Related papers  (2024-04-10T17:27:54Z)
• Variance Alignment Score: A Simple But Tough-to-Beat Data Selection Method for Multimodal Contrastive Learning [17.40655778450583]
  We propose a principled metric named Variance Alignment Score (VAS), which has the form $langle Sigma_texttest, Sigma_irangle$. We show that applying VAS and CLIP scores together can outperform baselines by a margin of $1.3%$ on 38 evaluation sets for noisy dataset DataComp and $2.5%$ on VTAB for high-quality dataset CC12M.
  arXiv  Detail & Related papers  (2024-02-03T06:29:04Z)
• The Unreasonable Effectiveness of Easy Training Data for Hard Tasks [84.30018805150607]
  We present the surprising conclusion that current pretrained language models often generalize relatively well from easy to hard data. We demonstrate this kind of easy-to-hard generalization using simple finetuning methods like in-context learning, linear heads, and QLoRA. We conclude that easy-to-hard generalization in LMs is surprisingly strong for the tasks studied.
  arXiv  Detail & Related papers  (2024-01-12T18:36:29Z)
• Compressive Recovery of Sparse Precision Matrices [5.557600489035657]
  We consider the problem of learning a graph modeling the statistical relations of the $d$ variables from a dataset with $n$ samples $X in mathbbRn times d$. We show that it is possible to estimate it from a sketch of size $m=Omegaleft((d+2k)log(d)right)$ where $k$ is the maximal number of edges of the underlying graph. We investigate the possibility of achieving practical recovery with an iterative algorithm based on the graphical lasso, viewed as a specific denoiser.
  arXiv  Detail & Related papers  (2023-11-08T13:29:08Z)
• Data Factors for Better Compositional Generalization [60.698130703909804]
  We conduct an empirical analysis by training Transformer models on a variety of training sets with different data factors. We show that increased dataset complexity can lead to better generalization behavior on multiple different generalization challenges. We explore how training examples of different difficulty levels influence generalization differently.
  arXiv  Detail & Related papers  (2023-11-08T01:27:34Z)
• DatasetEquity: Are All Samples Created Equal? In The Quest For Equity Within Datasets [4.833815605196965]
  This paper presents a novel method for addressing data imbalance in machine learning. It computes sample likelihoods based on image appearance using deep perceptual embeddings and clustering. It then uses these likelihoods to weigh samples differently during training with a proposed $bfGeneralized Focal Loss$ function.
  arXiv  Detail & Related papers  (2023-08-19T02:11:49Z)
• Simplicity Bias Leads to Amplified Performance Disparities [8.60453031364566]
  We show that SGD-trained models have a bias towards simplicity, leading them to prioritize learning a majority class. A model may prioritize any class or group of the dataset that it finds simple-at the expense of what it finds complex.
  arXiv  Detail & Related papers  (2022-12-13T15:24:41Z)
• PyHard: a novel tool for generating hardness embeddings to support data-centric analysis [0.38233569758620045]
  PyHard produces a hardness embedding of a dataset relating the predictive performance of multiple ML models. The user can interact with this embedding in multiple ways to obtain useful insights about data and algorithmic performance. We show in a COVID prognosis dataset how this analysis supported the identification of pockets of hard observations that challenge ML models.
  arXiv  Detail & Related papers  (2021-09-29T14:08:26Z)
• Competency Problems: On Finding and Removing Artifacts in Language Data [50.09608320112584]
  We argue that for complex language understanding tasks, all simple feature correlations are spurious. We theoretically analyze the difficulty of creating data for competency problems when human bias is taken into account.
  arXiv  Detail & Related papers  (2021-04-17T21:34:10Z)
• DAIL: Dataset-Aware and Invariant Learning for Face Recognition [67.4903809903022]
  To achieve good performance in face recognition, a large scale training dataset is usually required. It is problematic and troublesome to naively combine different datasets due to two major issues. Naively treating the same person as different classes in different datasets during training will affect back-propagation. manually cleaning labels may take formidable human efforts, especially when there are millions of images and thousands of identities.
  arXiv  Detail & Related papers  (2021-01-14T01:59:52Z)
• List-Decodable Mean Estimation in Nearly-PCA Time [50.79691056481693]
  We study the fundamental task of list-decodable mean estimation in high dimensions. Our algorithm runs in time $widetildeO(ndk)$ for all $k = O(sqrtd) cup Omega(d)$, where $n$ is the size of the dataset. A variant of our algorithm has runtime $widetildeO(ndk)$ for all $k$, at the expense of an $O(sqrtlog k)$ factor in the recovery guarantee
  arXiv  Detail & Related papers  (2020-11-19T17:21:37Z)
• Evaluating Models' Local Decision Boundaries via Contrast Sets [119.38387782979474]
  We propose a new annotation paradigm for NLP that helps to close systematic gaps in the test data. We demonstrate the efficacy of contrast sets by creating them for 10 diverse NLP datasets. Although our contrast sets are not explicitly adversarial, model performance is significantly lower on them than on the original test sets.
  arXiv  Detail & Related papers  (2020-04-06T14:47:18Z)

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.