Steven Euijong Whang

Yuji Roh, Kangwook Lee, Steven Euijong Whang et al.

Model fairness is an essential element for Trustworthy AI. While many techniques for model fairness have been proposed, most of them assume that the training and deployment data distributions are identical, which is often not true in practice. In particular, when the bias between labels and sensitive groups changes, the fairness of the trained model is directly influenced and can worsen. We make two contributions for solving this problem. First, we analytically show that existing in-processing fair algorithms have fundamental limits in accuracy and group fairness. We introduce the notion of correlation shifts, which can explicitly capture the change of the above bias. Second, we propose a novel pre-processing step that samples the input data to reduce correlation shifts and thus enables the in-processing approaches to overcome their limitations. We formulate an optimization problem for adjusting the data ratio among labels and sensitive groups to reflect the shifted correlation. A key benefit of our approach lies in decoupling the roles of pre- and in-processing approaches: correlation adjustment via pre-processing and unfairness mitigation on the processed data via in-processing. Experiments show that our framework effectively improves existing in-processing fair algorithms w.r.t. accuracy and fairness, both on synthetic and real datasets.

Hantian Zhang, Ki Hyun Tae, Jaeyoung Park et al.

As machine learning becomes prevalent, mitigating any unfairness present in the training data becomes critical. Among the various notions of fairness, this paper focuses on the well-known individual fairness, which states that similar individuals should be treated similarly. While individual fairness can be improved when training a model (in-processing), we contend that fixing the data before model training (pre-processing) is a more fundamental solution. In particular, we show that label flipping is an effective pre-processing technique for improving individual fairness. Our system iFlipper solves the optimization problem of minimally flipping labels given a limit to the individual fairness violations, where a violation occurs when two similar examples in the training data have different labels. We first prove that the problem is NP-hard. We then propose an approximate linear programming algorithm and provide theoretical guarantees on how close its result is to the optimal solution in terms of the number of label flips. We also propose techniques for making the linear programming solution more optimal without exceeding the violations limit. Experiments on real datasets show that iFlipper significantly outperforms other pre-processing baselines in terms of individual fairness and accuracy on unseen test sets. In addition, iFlipper can be combined with in-processing techniques for even better results.

Hyunseung Hwang, Steven Euijong Whang

We study the problem of explainability-first clustering where explainability becomes a first-class citizen for clustering. Previous clustering approaches use decision trees for explanation, but only after the clustering is completed. In contrast, our approach is to perform clustering and decision tree training holistically where the decision tree's performance and size also influence the clustering results. We assume the attributes for clustering and explaining are distinct, although this is not necessary. We observe that our problem is a monotonic optimization where the objective function is a difference of monotonic functions. We then propose an efficient branch-and-bound algorithm for finding the best parameters that lead to a balance of cluster distortion and decision tree explainability. Our experiments show that our method can improve the explainability of any clustering that fits in our framework.

Jio Oh, Steven Euijong Whang, James Evans et al.

Large Language Models (LLMs) offer a promising solution to complement traditional teaching and address global teacher shortages that affect hundreds of millions of children, but they fail to provide grade-appropriate responses for students at different educational levels. We introduce a framework for finetuning LLMs to generate age-appropriate educational content across six grade levels, from lower elementary to adult education. Our framework successfully adapts explanations to match students' comprehension capacities without sacrificing factual correctness. This approach integrates seven established readability metrics through a clustering method and builds a comprehensive dataset for grade-specific content generation. Evaluations across multiple datasets with 208 human participants demonstrate substantial improvements in grade-level alignment, achieving a 35.64 percentage point increase compared to prompt-based methods while maintaining response accuracy. AI-assisted learning tailored to different grade levels has the potential to advance educational engagement and equity.

Kahee Lim, Soyeon Kim, Steven Euijong Whang

As large language models (LLMs) are increasingly deployed in real-world applications, ensuring their fair responses across demographics has become crucial. Despite many efforts, an ongoing challenge is hidden bias: LLMs appear fair under standard evaluations, but can produce biased responses outside those evaluation settings. In this paper, we identify framing -- differences in how semantically equivalent prompts are expressed (e.g., "A is better than B" vs. "B is worse than A") -- as an underexplored contributor to this gap. We first introduce the concept of "framing disparity" to quantify the impact of framing on fairness evaluation. By augmenting fairness evaluation benchmarks with alternative framings, we find that (1) fairness scores vary significantly with framing and (2) existing debiasing methods improve overall (i.e., frame-averaged) fairness, but often fail to reduce framing-induced disparities. To address this, we propose a framing-aware debiasing method that encourages LLMs to be more consistent across framings. Experiments demonstrate that our approach reduces overall bias and improves robustness against framing disparities, enabling LLMs to produce fairer and more consistent responses.

Jio Oh, Paul Vicinanza, Thomas Butler et al.

More than 80% of the 1.6 billion English speakers do not use Standard American English (SAE) and experience higher failure rates and stereotyped responses when interacting with LLMs as a result. Yet multi-dialectal performance remains underexplored. We introduce $\textbf{MDial}$, the first large-scale framework for generating multi-dialectal conversational data encompassing the three pillars of written dialect -- lexical (vocabulary), orthographic (spelling), and morphosyntactic (grammar) features -- for nine English dialects. Partnering with native linguists, we design an annotated and scalable rule-based LLM transformation to ensure precision. Our approach challenges the assumption that models should mirror users' morphosyntactic features, showing that up to 90% of the grammatical features of a dialect should not be reproduced by models. Independent evaluations confirm data quality, with annotators preferring MDial outputs over prior methods in 98% of pairwise comparisons for dialect naturalness. Using this pipeline, we construct the dialect-parallel $\textbf{MDialBench}$mark with 50k+ dialogs, resulting in 97k+ QA pairs, and evaluate 17 LLMs on dialect identification and response generation tasks. Even frontier models achieve under 70% accuracy, fail to reach 50% for Canadian English, and systematically misclassify non-SAE dialects as American or British. As dialect identification underpins natural language understanding, these errors risk cascading failures into downstream tasks.

Soyeon Kim, Jindong Wang, Xing Xie et al.

Facts evolve over time, making it essential for Large Language Models (LLMs) to handle time-sensitive factual knowledge accurately and reliably. While factual Time-Sensitive Question-Answering (TSQA) tasks have been widely studied, existing benchmarks often rely on manual curation or a small, fixed set of predefined templates, which restricts scalable and comprehensive TSQA evaluation. To address these challenges, we propose TDBench, a new benchmark that systematically constructs TSQA pairs by harnessing temporal databases and database techniques such as temporal SQL and functional dependencies. We also introduce a fine-grained evaluation metric called time accuracy, which assesses the validity of time references in model explanations alongside traditional answer accuracy to enable a more reliable TSQA evaluation. Extensive experiments on contemporary LLMs show how \ours{} enables scalable and comprehensive TSQA evaluation while reducing the reliance on human labor, complementing existing Wikipedia/Wikidata-based TSQA evaluation approaches by enabling LLM evaluation on application-specific data and seamless multi-hop question generation. Code and data are publicly available at: https://github.com/ssoy0701/tdbench.git.

Jio Oh, Soyeon Kim, Junseok Seo et al.

Large language models (LLMs) have achieved unprecedented performances in various applications, yet evaluating them is still challenging. Existing benchmarks are either manually constructed or are automatic, but lack the ability to evaluate the thought process of LLMs with arbitrary complexity. We contend that utilizing existing relational databases based on the entity-relationship (ER) model is a promising approach for constructing benchmarks as they contain structured knowledge that can be used to question LLMs. Unlike knowledge graphs, which are also used to evaluate LLMs, relational databases have integrity constraints that can be used to better construct complex in-depth questions and verify answers: (1) functional dependencies can be used to pinpoint critical keywords that an LLM must know to properly answer a given question containing certain attribute values; and (2) foreign key constraints can be used to join relations and construct multi-hop questions, which can be arbitrarily long and used to debug intermediate answers. We thus propose ERBench, which uses these integrity constraints to convert any database into an LLM benchmark. ERBench supports continuous evaluation as databases change, multimodal questions, and various prompt engineering techniques. In our experiments, we construct LLM benchmarks using databases of multiple domains and make an extensive comparison of contemporary LLMs. We show how ERBench can properly evaluate any LLM by not only checking for answer correctness, but also effectively verifying the rationales by looking for the right keywords.

Ki Hyun Tae, Hantian Zhang, Jaeyoung Park et al.

Biased data can lead to unfair machine learning models, highlighting the importance of embedding fairness at the beginning of data analysis, particularly during dataset curation and labeling. In response, we propose Falcon, a scalable fair active learning framework. Falcon adopts a data-centric approach that improves machine learning model fairness via strategic sample selection. Given a user-specified group fairness measure, Falcon identifies samples from "target groups" (e.g., (attribute=female, label=positive)) that are the most informative for improving fairness. However, a challenge arises since these target groups are defined using ground truth labels that are not available during sample selection. To handle this, we propose a novel trial-and-error method, where we postpone using a sample if the predicted label is different from the expected one and falls outside the target group. We also observe the trade-off that selecting more informative samples results in higher likelihood of postponing due to undesired label prediction, and the optimal balance varies per dataset. We capture the trade-off between informativeness and postpone rate as policies and propose to automatically select the best policy using adversarial multi-armed bandit methods, given their computational efficiency and theoretical guarantees. Experiments show that Falcon significantly outperforms existing fair active learning approaches in terms of fairness and accuracy and is more efficient. In particular, only Falcon supports a proper trade-off between accuracy and fairness where its maximum fairness score is 1.8-4.5x higher than the second-best results.

Hyunseung Hwang, Andrew Bell, Joao Fonseca et al.

Local feature-based explanations are a key component of the XAI toolkit. These explanations compute feature importance values relative to an ``interpretable'' feature representation. In tabular data, feature values themselves are often considered interpretable. This paper examines the impact of data engineering choices on local feature-based explanations. We demonstrate that simple, common data engineering techniques, such as representing age with a histogram or encoding race in a specific way, can manipulate feature importance as determined by popular methods like SHAP. Notably, the sensitivity of explanations to feature representation can be exploited by adversaries to obscure issues like discrimination. While the intuition behind these results is straightforward, their systematic exploration has been lacking. Previous work has focused on adversarial attacks on feature-based explainers by biasing data or manipulating models. To the best of our knowledge, this is the first study demonstrating that explainers can be misled by standard, seemingly innocuous data engineering techniques.

Yuji Roh, Qingyun Liu, Huan Gui et al.

Fine-tuning is becoming widely used for leveraging the power of pre-trained foundation models in new downstream tasks. While there are many successes of fine-tuning on various tasks, recent studies have observed challenges in the generalization of fine-tuned models to unseen distributions (i.e., out-of-distribution; OOD). To improve OOD generalization, some previous studies identify the limitations of fine-tuning data and regulate fine-tuning to preserve the general representation learned from pre-training data. However, potential limitations in the pre-training data and models are often ignored. In this paper, we contend that overly relying on the pre-trained representation may hinder fine-tuning from learning essential representations for downstream tasks and thus hurt its OOD generalization. It can be especially catastrophic when new tasks are from different (sub)domains compared to pre-training data. To address the issues in both pre-training and fine-tuning data, we propose a novel generalizable fine-tuning method LEVI (Layer-wise Ensemble of different VIews), where the pre-trained model is adaptively ensembled layer-wise with a small task-specific model, while preserving its efficiencies. By combining two complementing models, LEVI effectively suppresses problematic features in both the fine-tuning data and pre-trained model and preserves useful features for new tasks. Broad experiments with large language and vision models show that LEVI greatly improves fine-tuning generalization via emphasizing different views from fine-tuning data and pre-trained features.

Minsu Kim, Seong-Hyeon Hwang, Steven Euijong Whang

Continuous machine learning pipelines are common in industrial settings where models are periodically trained on data streams. Unfortunately, concept drifts may occur in data streams where the joint distribution of the data X and label y, P(X, y), changes over time and possibly degrade model accuracy. Existing concept drift adaptation approaches mostly focus on updating the model to the new data possibly using ensemble techniques of previous models and tend to discard the drifted historical data. However, we contend that explicitly utilizing the drifted data together leads to much better model accuracy and propose Quilt, a data-centric framework for identifying and selecting data segments that maximize model accuracy. To address the potential downside of efficiency, Quilt extends existing data subset selection techniques, which can be used to reduce the training data without compromising model accuracy. These techniques cannot be used as is because they only assume virtual drifts where the posterior probabilities P(y|X) are assumed not to change. In contrast, a key challenge in our setup is to also discard undesirable data segments with concept drifts. Quilt thus discards drifted data segments and selects data segment subsets holistically for accurate and efficient model training. The two operations use gradient-based scores, which have little computation overhead. In our experiments, we show that Quilt outperforms state-of-the-art drift adaptation and data selection baselines on synthetic and real datasets.

Seong-Hyeon Hwang, Minsu Kim, Steven Euijong Whang

We study model confidence calibration in class-incremental learning, where models learn from sequential tasks with different class sets. While existing works primarily focus on accuracy, maintaining calibrated confidence has been largely overlooked. Unfortunately, most post-hoc calibration techniques are not designed to work with the limited memories of old-task data typical in class-incremental learning, as retaining a sufficient validation set would be impractical. Thus, we propose T-CIL, a novel temperature scaling approach for class-incremental learning without a validation set for old tasks, that leverages adversarially perturbed exemplars from memory. Directly using exemplars is inadequate for temperature optimization, since they are already used for training. The key idea of T-CIL is to perturb exemplars more strongly for old tasks than for the new task by adjusting the perturbation direction based on feature distance, with the single magnitude determined using the new-task validation set. This strategy makes the perturbation magnitude computed from the new task also applicable to old tasks, leveraging the tendency that the accuracy of old tasks is lower than that of the new task. We empirically show that T-CIL significantly outperforms various baselines in terms of calibration on real datasets and can be integrated with existing class-incremental learning techniques with minimal impact on accuracy.

Hyunseung Hwang, Seungeun Lee, Lucas Rosenblatt et al.

Post-hoc explanations are widely used to justify, contest, and audit automated decisions in high-stakes domains. SHAP, in particular, is often treated as a reliable account of which features drove an individual prediction. Yet SHAP explanations can vary substantially across repeated runs even when the input, task, and trained model are held fixed. We term this phenomenon explanation multiplicity: multiple internally valid but substantively different explanations for the same decision. We present a methodology to characterize multiplicity in feature-attribution explanations and to disentangle sources due to model training/selection from stochasticity intrinsic to the explanation pipeline. We further show that apparent stability depends on the metric: magnitude-based distances can remain near zero while rank-based measures reveal substantial churn in the identity and ordering of top features. To contextualize observed disagreement, we derive randomized baseline values under plausible null models. Across datasets, model classes, and confidence regimes, we find explanation multiplicity is pervasive and persists even for high-confidence predictions, highlighting the need for metrics and baselines that match the intended use of explanations.

Seong-Hyeon Hwang, Soyoung Choi, Steven Euijong Whang

Multimodal models often over-rely on dominant modalities, failing to achieve optimal performance. While prior work focuses on modifying training objectives or optimization procedures, data-centric solutions remain underexplored. We propose MIDAS, a novel data augmentation strategy that generates misaligned samples with semantically inconsistent cross-modal information, labeled using unimodal confidence scores to compel learning from contradictory signals. However, this confidence-based labeling can still favor the more confident modality. To address this within our misaligned samples, we introduce weak-modality weighting, which dynamically increases the loss weight of the least confident modality, thereby helping the model fully utilize weaker modality. Furthermore, when misaligned features exhibit greater similarity to the aligned features, these misaligned samples pose a greater challenge, thereby enabling the model to better distinguish between classes. To leverage this, we propose hard-sample weighting, which prioritizes such semantically ambiguous misaligned samples. Experiments on multiple multimodal classification benchmarks demonstrate that MIDAS significantly outperforms related baselines in addressing modality imbalance.

Minsu Kim, Seong-Hyeon Hwang, Steven Euijong Whang

In the context of continual learning, acquiring new knowledge while maintaining previous knowledge presents a significant challenge. Existing methods often use experience replay techniques that store a small portion of previous task data for training. In experience replay approaches, data augmentation has emerged as a promising strategy to further improve the model performance by mixing limited previous task data with sufficient current task data. However, we theoretically and empirically analyze that training with mixed samples from random sample pairs may harm the knowledge of previous tasks and cause greater catastrophic forgetting. We then propose GradMix, a robust data augmentation method specifically designed for mitigating catastrophic forgetting in class-incremental learning. GradMix performs gradient-based selective mixup using a class-based criterion that mixes only samples from helpful class pairs and not from detrimental class pairs for reducing catastrophic forgetting. Our experiments on various real datasets show that GradMix outperforms data augmentation baselines in accuracy by minimizing the forgetting of previous knowledge.

Jio Oh, Geon Heo, Seungjun Oh et al.

Large Language Models (LLMs) often struggle with data-analytics requests related to information retrieval and data manipulation that frequently arise in real-world scenarios under multiple conditions. In this paper, we introduce Thinking with Tables, where we inject tabular structures into LLMs for data-analytics requests. Through comprehensive evaluations across various request types, we show that providing tabular structures yields a 40.29 percent average performance gain along with better robustness and token efficiency. Through attention-value analysis, we uncover that tables help LLMs better attend to relevant information, explaining these improvements. Beyond tables and text, we evaluate whether (1) blending structuredness within text, such as providing templates or fixing the order of attributes, and (2) other representative structures, such as knowledge graphs and JSON, are helpful. We observe that utilizing tables offers the best balance between efficiency and effectiveness. These advantages remain consistent under increased task complexity and even when all input data cannot be structured. Finally, as data analytics typically relies on structured factual inputs, our text-to-table conversion demonstrates the method's applicability to text-compatible data sources.

Geon Heo, Junseok Seo, Steven Euijong Whang

Personalized privacy becomes critical in deep learning for Trustworthy AI. While Differentially Private Stochastic Gradient Descent (DP-SGD) is widely used in deep learning methods supporting privacy, it provides the same level of privacy to all individuals, which may lead to overprotection and low utility. In practice, different users may require different privacy levels, and the model can be improved by using more information about the users with lower privacy requirements. There are also recent works on differential privacy of individuals when using DP-SGD, but they are mostly about individual privacy accounting and do not focus on satisfying different privacy levels. We thus extend DP-SGD to support a recent privacy notion called ($Φ$,$Δ$)-Personalized Differential Privacy (($Φ$,$Δ$)-PDP), which extends an existing PDP concept called $Φ$-PDP. Our algorithm uses a multi-round personalized sampling mechanism and embeds it within the DP-SGD iterations. Experiments on real datasets show that our algorithm outperforms DP-SGD and simple combinations of DP-SGD with existing PDP mechanisms in terms of model performance and efficiency due to its embedded sampling mechanism.

Geon Heo, Steven Euijong Whang

Information leakage is becoming a critical problem as various information becomes publicly available by mistake, and machine learning models train on that data to provide services. As a result, one's private information could easily be memorized by such trained models. Unfortunately, deleting information is out of the question as the data is already exposed to the Web or third-party platforms. Moreover, we cannot necessarily control the labeling process and the model trainings by other parties either. In this setting, we study the problem of targeted disinformation generation where the goal is to dilute the data and thus make a model safer and more robust against inference attacks on a specific target (e.g., a person's profile) by only inserting new data. Our method finds the closest points to the target in the input space that will be labeled as a different class. Since we cannot control the labeling process, we instead conservatively estimate the labels probabilistically by combining decision boundaries of multiple classifiers using data programming techniques. Our experiments show that a probabilistic decision boundary can be a good proxy for labelers, and that our approach is effective in defending against inference attacks and can scale to large data.

Steven Euijong Whang, Yuji Roh, Hwanjun Song et al.

Data-centric AI is at the center of a fundamental shift in software engineering where machine learning becomes the new software, powered by big data and computing infrastructure. Here software engineering needs to be re-thought where data becomes a first-class citizen on par with code. One striking observation is that a significant portion of the machine learning process is spent on data preparation. Without good data, even the best machine learning algorithms cannot perform well. As a result, data-centric AI practices are now becoming mainstream. Unfortunately, many datasets in the real world are small, dirty, biased, and even poisoned. In this survey, we study the research landscape for data collection and data quality primarily for deep learning applications. Data collection is important because there is lesser need for feature engineering for recent deep learning approaches, but instead more need for large amounts of data. For data quality, we study data validation, cleaning, and integration techniques. Even if the data cannot be fully cleaned, we can still cope with imperfect data during model training using robust model training techniques. In addition, while bias and fairness have been less studied in traditional data management research, these issues become essential topics in modern machine learning applications. We thus study fairness measures and unfairness mitigation techniques that can be applied before, during, or after model training. We believe that the data management community is well poised to solve these problems.

Yuji Roh, Kangwook Lee, Steven Euijong Whang et al.

Fairness and robustness are critical elements of Trustworthy AI that need to be addressed together. Fairness is about learning an unbiased model while robustness is about learning from corrupted data, and it is known that addressing only one of them may have an adverse affect on the other. In this work, we propose a sample selection-based algorithm for fair and robust training. To this end, we formulate a combinatorial optimization problem for the unbiased selection of samples in the presence of data corruption. Observing that solving this optimization problem is strongly NP-hard, we propose a greedy algorithm that is efficient and effective in practice. Experiments show that our algorithm obtains fairness and robustness that are better than or comparable to the state-of-the-art technique, both on synthetic and benchmark real datasets. Moreover, unlike other fair and robust training baselines, our algorithm can be used by only modifying the sampling step in batch selection without changing the training algorithm or leveraging additional clean data.

Seong-Hyeon Hwang, Steven Euijong Whang

Data augmentation is becoming essential for improving regression performance in critical applications including manufacturing, climate prediction, and finance. Existing techniques for data augmentation largely focus on classification tasks and do not readily apply to regression tasks. In particular, the recent Mixup techniques for classification have succeeded in improving the model performance, which is reasonable due to the characteristics of the classification task, but has limitations in regression. We show that mixing examples that have large data distances using linear interpolations may have increasingly-negative effects on model performance. Our key idea is thus to limit the distances between examples that are mixed. We propose RegMix, a data augmentation framework for regression that learns for each example how many nearest neighbors it should be mixed with for the best model performance using a validation set. Our experiments conducted both on synthetic and real datasets show that RegMix outperforms state-of-the-art data augmentation baselines applicable to regression.

Steven Euijong Whang, Ki Hyun Tae, Yuji Roh et al.

Responsible AI is becoming critical as AI is widely used in our everyday lives. Many companies that deploy AI publicly state that when training a model, we not only need to improve its accuracy, but also need to guarantee that the model does not discriminate against users (fairness), is resilient to noisy or poisoned data (robustness), is explainable, and more. In addition, these objectives are not only relevant to model training, but to all steps of end-to-end machine learning, which include data collection, data cleaning and validation, model training, model evaluation, and model management and serving. Finally, responsible AI is conceptually challenging, and supporting all the objectives must be as easy as possible. We thus propose three key research directions towards this vision - depth, breadth, and usability - to measure progress and introduce our ongoing research. First, responsible AI must be deeply supported where multiple objectives like fairness and robust must be handled together. To this end, we propose FR-Train, a holistic framework for fair and robust model training in the presence of data bias and poisoning. Second, responsible AI must be broadly supported, preferably in all steps of machine learning. Currently we focus on the data pre-processing steps and propose Slice Tuner, a selective data acquisition framework for training fair and accurate models, and MLClean, a data cleaning framework that also improves fairness and robustness. Finally, responsible AI must be usable where the techniques must be easy to deploy and actionable. We propose FairBatch, a batch selection approach for fairness that is effective and simple to use, and Slice Finder, a model evaluation tool that automatically finds problematic slices. We believe we scratched the surface of responsible AI for end-to-end machine learning and suggest research challenges moving forward.

Yuji Roh, Kangwook Lee, Steven Euijong Whang et al.

Training a fair machine learning model is essential to prevent demographic disparity. Existing techniques for improving model fairness require broad changes in either data preprocessing or model training, rendering themselves difficult-to-adopt for potentially already complex machine learning systems. We address this problem via the lens of bilevel optimization. While keeping the standard training algorithm as an inner optimizer, we incorporate an outer optimizer so as to equip the inner problem with an additional functionality: Adaptively selecting minibatch sizes for the purpose of improving model fairness. Our batch selection algorithm, which we call FairBatch, implements this optimization and supports prominent fairness measures: equal opportunity, equalized odds, and demographic parity. FairBatch comes with a significant implementation benefit -- it does not require any modification to data preprocessing or model training. For instance, a single-line change of PyTorch code for replacing batch selection part of model training suffices to employ FairBatch. Our experiments conducted both on synthetic and benchmark real data demonstrate that FairBatch can provide such functionalities while achieving comparable (or even greater) performances against the state of the arts. Furthermore, FairBatch can readily improve fairness of any pre-trained model simply via fine-tuning. It is also compatible with existing batch selection techniques intended for different purposes, such as faster convergence, thus gracefully achieving multiple purposes.

Geon Heo, Yuji Roh, Seonghyeon Hwang et al.

As machine learning for images becomes democratized in the Software 2.0 era, one of the serious bottlenecks is securing enough labeled data for training. This problem is especially critical in a manufacturing setting where smart factories rely on machine learning for product quality control by analyzing industrial images. Such images are typically large and may only need to be partially analyzed where only a small portion is problematic (e.g., identifying defects on a surface). Since manual labeling these images is expensive, weak supervision is an attractive alternative where the idea is to generate weak labels that are not perfect, but can be produced at scale. Data programming is a recent paradigm in this category where it uses human knowledge in the form of labeling functions and combines them into a generative model. Data programming has been successful in applications based on text or structured data and can also be applied to images usually if one can find a way to convert them into structured data. In this work, we expand the horizon of data programming by directly applying it to images without this conversion, which is a common scenario for industrial applications. We propose Inspector Gadget, an image labeling system that combines crowdsourcing, data augmentation, and data programming to produce weak labels at scale for image classification. We perform experiments on real industrial image datasets and show that Inspector Gadget obtains better performance than other weak-labeling techniques: Snuba, GOGGLES, and self-learning baselines using convolutional neural networks (CNNs) without pre-training.

Ki Hyun Tae, Steven Euijong Whang

As machine learning becomes democratized in the era of Software 2.0, a serious bottleneck is acquiring enough data to ensure accurate and fair models. Recent techniques including crowdsourcing provide cost-effective ways to gather such data. However, simply acquiring data as much as possible is not necessarily an effective strategy for optimizing accuracy and fairness. For example, if an online app store has enough training data for certain slices of data (say American customers), but not for others, obtaining more American customer data will only bias the model training. Instead, we contend that one needs to selectively acquire data and propose Slice Tuner, which acquires possibly-different amounts of data per slice such that the model accuracy and fairness on all slices are optimized. This problem is different than labeling existing data (as in active learning or weak supervision) because the goal is obtaining the right amounts of new data. At its core, Slice Tuner maintains learning curves of slices that estimate the model accuracies given more data and uses convex optimization to find the best data acquisition strategy. The key challenges of estimating learning curves are that they may be inaccurate if there is not enough data, and there may be dependencies among slices where acquiring data for one slice influences the learning curves of others. We solve these issues by iteratively and efficiently updating the learning curves as more data is acquired. We evaluate Slice Tuner on real datasets using crowdsourcing for data acquisition and show that Slice Tuner significantly outperforms baselines in terms of model accuracy and fairness, even when the learning curves cannot be reliably estimated.

Yuji Roh, Kangwook Lee, Steven Euijong Whang et al.

Trustworthy AI is a critical issue in machine learning where, in addition to training a model that is accurate, one must consider both fair and robust training in the presence of data bias and poisoning. However, the existing model fairness techniques mistakenly view poisoned data as an additional bias to be fixed, resulting in severe performance degradation. To address this problem, we propose FR-Train, which holistically performs fair and robust model training. We provide a mutual information-based interpretation of an existing adversarial training-based fairness-only method, and apply this idea to architect an additional discriminator that can identify poisoned data using a clean validation set and reduce its influence. In our experiments, FR-Train shows almost no decrease in fairness and accuracy in the presence of data poisoning by both mitigating the bias and defending against poisoning. We also demonstrate how to construct clean validation sets using crowdsourcing, and release new benchmark datasets.

Ki Hyun Tae, Yuji Roh, Young Hun Oh et al.

The wide use of machine learning is fundamentally changing the software development paradigm (a.k.a. Software 2.0) where data becomes a first-class citizen, on par with code. As machine learning is used in sensitive applications, it becomes imperative that the trained model is accurate, fair, and robust to attacks. While many techniques have been proposed to improve the model training process (in-processing approach) or the trained model itself (post-processing), we argue that the most effective method is to clean the root cause of error: the data the model is trained on (pre-processing). Historically, there are at least three research communities that have been separately studying this problem: data management, machine learning (model fairness), and security. Although a significant amount of research has been done by each community, ultimately the same datasets must be preprocessed, and there is little understanding how the techniques relate to each other and can possibly be integrated. We contend that it is time to extend the notion of data cleaning for modern machine learning needs. We identify dependencies among the data preprocessing techniques and propose MLClean, a unified data cleaning framework that integrates the techniques and helps train accurate and fair models. This work is part of a broader trend of Big data -- Artificial Intelligence (AI) integration.

Yuji Roh, Geon Heo, Steven Euijong Whang

Data collection is a major bottleneck in machine learning and an active research topic in multiple communities. There are largely two reasons data collection has recently become a critical issue. First, as machine learning is becoming more widely-used, we are seeing new applications that do not necessarily have enough labeled data. Second, unlike traditional machine learning, deep learning techniques automatically generate features, which saves feature engineering costs, but in return may require larger amounts of labeled data. Interestingly, recent research in data collection comes not only from the machine learning, natural language, and computer vision communities, but also from the data management community due to the importance of handling large amounts of data. In this survey, we perform a comprehensive study of data collection from a data management point of view. Data collection largely consists of data acquisition, data labeling, and improvement of existing data or models. We provide a research landscape of these operations, provide guidelines on which technique to use when, and identify interesting research challenges. The integration of machine learning and data management for data collection is part of a larger trend of Big data and Artificial Intelligence (AI) integration and opens many opportunities for new research.

Yeounoh Chung, Tim Kraska, Neoklis Polyzotis et al.

As machine learning systems become democratized, it becomes increasingly important to help users easily debug their models. However, current data tools are still primitive when it comes to helping users trace model performance problems all the way to the data. We focus on the particular problem of slicing data to identify subsets of the validation data where the model performs poorly. This is an important problem in model validation because the overall model performance can fail to reflect that of the smaller subsets, and slicing allows users to analyze the model performance on a more granular-level. Unlike general techniques (e.g., clustering) that can find arbitrary slices, our goal is to find interpretable slices (which are easier to take action compared to arbitrary subsets) that are problematic and large. We propose Slice Finder, which is an interactive framework for identifying such slices using statistical techniques. Applications include diagnosing model fairness and fraud detection, where identifying slices that are interpretable to humans is crucial. This research is part of a larger trend of Big data and Artificial Intelligence (AI) integration and opens many opportunities for new research.