ActiveClean: Interactive Data Cleaning For Statistical Modeling

By Sanjay Krishnan, Jiannan Wang, Eugene Wu, M. Franklin, Ken Goldberg. 2016

282 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Management Science and Operations Research, Statistics and Probability, ML Systems & Hardware

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Management Science and Operations Research

In Statistics and Probability

In ML Systems & Hardware

Citations in the corpus

Benchmark and Survey of Automated Machine Learning Frameworks (2019, background): “Yet, most current approaches still aim to assist a human data scientist instead of fully automated data cleaning, (e.g., Krishnan et al., 2015; Khayyaty et al., 2015; Krishnan et al., 2016; Eduardo & Sutton, 2016; Rekatsinas et al., 2017).”
Survey on Automated Machine Learning (2019, background): “00; Raman and Hellerstein, 2001). Yet, most current approaches still aim to assist a human data scientist instead of fully automate data cleaning, e.g., (Krishnan et al., 2015; Khayyaty et al., 2015; Krishnan et al., 2016; Eduardo and Sutton, 2016; Rekatsinas et al., 2017). Krishnan and Wu (2019) pr”
A Survey on Data Collection for Machine Learning: A Big Data - AI Integration Perspective (2018, neutral): “ActiveClean [159] is a model training framework”
BoostClean: Automated Error Detection and Repair for Machine Learning (2017, methodology): “ver a dirty relation [11]. While the work on relational queries is extensive, analytical queries (aggregates, advanced statistical analytics, learning etc.) is less studied. Projects like ActiveClean [32] have studied algorithms for prioritizing user-deﬁned cleaning using the downstream ML model, Ac”
PrIU: A Provenance-Based Approach for Incrementally Updating Regression Models (2020, methodology): “when large volumes of data are processed. One approach is to start each training phase by setting the initial model parameters to the ones generated by the previous training phase over the dirty data [34]. Our contribution is orthogonal to this approach, and updates the machine learning model parame”
Towards Automated Data Cleaning Workflows (2019, methodology): “Finally, ActiveClean cleans the training data for a machine learning application and requires the user to specify how to clean and how to featurize the dataset [14].”
“Everyone wants to do the model work, not the data work”: Data Cascades in High-Stakes AI (2021, background): “and clean data include ActiveClean and BoostClean [70, 72].”
Advances, challenges and opportunities in creating data for trustworthy AI (2022, methodology): “Methods such as ActiveClean improve cleaning efficiency by identifying the subset of the dirty data that are critical for the AI model and hence are important to clean 56 .”
Data Management Challenges in Production Machine Learning (2017, methodology): “A recent study [12] also looked at applying similar techniques for a speciﬁc class of machine learning algorithms.”
TFX: A TensorFlow-Based Production-Scale Machine Learning Platform (2017, background): “As a machine learning platform scales to larger data and runs continuously, there is a strong need for a reusable component that enables rigorous checks for data quality and promotes best practices for data management in the context of machine learning platforms [11].”
Cleaning Crowdsourced Labels Using Oracles For Statistical Classification (2018, background): “This is because that ActiveClean is focused on a different cleaning scenario, where there is a large dirty dataset (both features and labels are dirty), and an oracle cleans the data in batches.”
Machine Learning Testing: Survey, Landscapes and Horizons (2019, unknown): “[225], [226] proposed a model training framework, ActiveClean, that allows for iterative data cleaning while preserving provable convergence properties.”
PALM: Machine Learning Explanations For Iterative Debugging (2017, background): “Explaining a prediction exactly in terms of features is highly useful for an end-user, but developers also need to be able to trace modeling errors to training data [12].”