Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userGarbage In, Garbage Out Revisited: What Do Machine Learning Application Papers Report About Human-Labeled Training Data?
147
0
0.0
(
0
)
Added by
R.Stuart Geiger
Publication date
2021
fields
Informatics Engineering
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
Supervised machine learning, in which models are automatically derived from labeled training data, is only as good as the quality of that data. This study builds on prior work that investigated to what extent best practices around labeling training data were followed in applied ML publications within a single domain (social media platforms). In this paper, we expand by studying publications that apply supervised ML in a far broader spectrum of disciplines, focusing on human-labeled data. We report to what extent a random sample of ML application papers across disciplines give specific details about whether best practices were followed, while acknowledging that a greater range of application fields necessarily produces greater diversity of labeling and annotation methods. Because much of machine learning research and education only focuses on what is done once a ground truth or gold standard of training data is available, it is especially relevant to discuss issues around the equally-important aspect of whether such data is reliable in the first place. This determination becomes increasingly complex when applied to a variety of specialized fields, as labeling can range from a task requiring little-to-no background knowledge to one that must be performed by someone with career expertise.
rate research
Read More
Many machine learning projects for new application areas involve teams of humans who label data for a particular purpose, from hiring crowdworkers to the papers authors labeling the data themselves. Such a task is quite similar to (or a form of) structured content analysis, which is a longstanding methodology in the social sciences and humanities, with many established best practices. In this paper, we investigate to what extent a sample of machine learning application papers in social computing --- specifically papers from ArXiv and traditional publications performing an ML classification task on Twitter data --- give specific details about whether such best practices were followed. Our team conducted multiple rounds of structured content analysis of each paper, making determinations such as: Does the paper report who the labelers were, what their qualifications were, whether they independently labeled the same items, whether inter-rater reliability metrics were disclosed, what level of training and/or instructions were given to labelers, whether compensation for crowdworkers is disclosed, and if the training data is publicly available. We find a wide divergence in whether such practices were followed and documented. Much of machine learning research and education focuses on what is done once a gold standard of training data is available, but we discuss issues around the equally-important aspect of whether such data is reliable in the first place.
We propose a bandit algorithm that explores by randomizing its history of rewards. Specifically, it pulls the arm with the highest mean reward in a non-parametric bootstrap sample of its history with pseudo rewards. We design the pseudo rewards such that the bootstrap mean is optimistic with a sufficiently high probability. We call our algorithm Giro, which stands for garbage in, reward out. We analyze Giro in a Bernoulli bandit and derive a $O(K Delta^{-1} log n)$ bound on its $n$-round regret, where $Delta$ is the difference in the expected rewards of the optimal and the best suboptimal arms, and $K$ is the number of arms. The main advantage of our exploration design is that it easily generalizes to structured problems. To show this, we propose contextual Giro with an arbitrary reward generalization model. We evaluate Giro and its contextual variant on multiple synthetic and real-world problems, and observe that it performs well.
As the numbers of submissions to conferences grow quickly, the task of assessing the quality of academic papers automatically, convincingly, and with high accuracy attracts increasing attention. We argue that studying interpretable dimensions of these submissions could lead to scalable solutions. We extract a collection of writing features, and construct a suite of prediction tasks to assess the usefulness of these features in predicting citation counts and the publication of AI-related papers. Depending on the venues, the writing features can predict the conference vs. workshop appearance with F1 scores up to 60-90, sometimes even outperforming the content-based tf-idf features and RoBERTa. We show that the features describe writing style more than content. To further understand the results, we estimate the causal impact of the most indicative features. Our analysis on writing features provides a perspective to assessing and refining the writing of academic articles at scale.
We show that asynchronous $t$ faults Byzantine system is equivalent to asynchronous $t$-resilient system, where unbeknownst to all, the private inputs of at most $t$ processors were altered and installed by a malicious oracle. The immediate ramification is that dealing with asynchronous Byzantine systems does not call for new topological methods, as was recently employed by various researchers: Asynchronous Byzantine is a standard asynchronous system with an input caveat. It also shows that two recent independent investigations of vector $epsilon$-agreement in the Byzantine model, and then in the fail-stop model, one was superfluous - in these problems the change of $t$ inputs allowed in the Byzantine has no effect compared to the fail-stop case. This result was motivated by the aim of casting any asynchronous system as a synchronous system where all processors are correct and it is the communication substrate in the form of message-adversary that misbehaves. Thus, in addition, we get such a characterization for the asynchronous Byzantine system.
Increasingly, software is making autonomous decisions in case of criminal sentencing, approving credit cards, hiring employees, and so on. Some of these decisions show bias and adversely affect certain social groups (e.g. those defined by sex, race, age, marital status). Many prior works on bias mitigation take the following form: change the data or learners in multiple ways, then see if any of that improves fairness. Perhaps a better approach is to postulate root causes of bias and then applying some resolution strategy. This paper postulates that the root causes of bias are the prior decisions that affect- (a) what data was selected and (b) the labels assigned to those examples. Our Fair-SMOTE algorithm removes biased labels; and rebalances internal distributions such that based on sensitive attribute, examples are equal in both positive and negative classes. On testing, it was seen that this method was just as effective at reducing bias as prior approaches. Further, models generated via Fair-SMOTE achieve higher performance (measured in terms of recall and F1) than other state-of-the-art fairness improvement algorithms. To the best of our knowledge, measured in terms of number of analyzed learners and datasets, this study is one of the largest studies on bias mitigation yet presented in the literature.
suggested questions
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
