Suvodeep Majumder

Suvodeep Majumder, Joymallya Chakraborty, Tim Menzies

Labeling a module defective or non-defective is an expensive task. Hence, there are often limits on how much-labeled data is available for training. Semi-supervised classifiers use far fewer labels for training models. However, there are numerous semi-supervised methods, including self-labeling, co-training, maximal-margin, and graph-based methods, to name a few. Only a handful of these methods have been tested in SE for (e.g.) predicting defects and even there, those methods have been tested on just a handful of projects. This paper applies a wide range of 55 semi-supervised learners to over 714 projects. We find that semi-supervised "co-training methods" work significantly better than other approaches. Specifically, after labeling, just 2.5% of data, then make predictions that are competitive to those using 100% of the data. That said, co-training needs to be used cautiously since the specific choice of co-training methods needs to be carefully selected based on a user's specific goals. Also, we warn that a commonly-used co-training method ("multi-view"-- where different learners get different sets of columns) does not improve predictions (while adding too much to the run time costs 11 hours vs. 1.8 hours). It is an open question, worthy of future work, to test if these reductions can be seen in other areas of software analytics. To assist with exploring other areas, all the codes used are available at https://github.com/ai-se/Semi-Supervised.

Suvodeep Majumder, Stanislas Lauly, Maria Nadejde et al. · amazon-science

This paper addresses the task of contextual translation using multi-segment models. Specifically we show that increasing model capacity further pushes the limits of this approach and that deeper models are more suited to capture context dependencies. Furthermore, improvements observed with larger models can be transferred to smaller models using knowledge distillation. Our experiments show that this approach achieves competitive performance across several languages and benchmarks, without additional language-specific tuning and task specific architectures.

Suvodeep Majumder, Tianpei Xia, Rahul Krishna et al.

Despite decades of research, SE lacks widely accepted models (that offer precise quantitative stable predictions) about what factors most influence software quality. This paper provides a promising result showing such stable models can be generated using a new transfer learning framework called "STABILIZER". Given a tree of recursively clustered projects (using project meta-data), STABILIZER promotes a model upwards if it performs best in the lower clusters (stopping when the promoted model performs worse than the models seen at a lower level). The number of models found by STABILIZER is minimal: one for defect prediction (756 projects) and less than a dozen for project health (1628 projects). Hence, via STABILIZER, it is possible to find a few projects which can be used for transfer learning and make conclusions that hold across hundreds of projects at a time. Further, the models produced in this manner offer predictions that perform as well or better than the prior state-of-the-art. To the best of our knowledge, STABILIZER is order of magnitude faster than the prior state-of-the-art transfer learners which seek to find conclusion stability, and these case studies are the largest demonstration of the generalizability of quantitative predictions of project quality yet reported in the SE literature. In order to support open science, all our scripts and data are online at https://github.com/Anonymous633671/STABILIZER.

Suvodeep Majumder, Joymallya Chakraborty, Amritanshu Agrawal et al.

Prior literature has suggested that in many projects 80\% or more of the contributions are made by a small called group of around 20% of the development team. Most prior studies deprecate a reliance on such a small inner group of "heroes", arguing that it causes bottlenecks in development and communication. Despite this, such projects are very common in open source projects. So what exactly is the impact of "heroes" in code quality? Herbsleb argues that if code is strongly connected yet their developers are not, then that code will be buggy. To test the Hersleb hypothesis, we develop and apply two metrics of (a) "social-ness'"and (b) "hero-ness" that measure (a) how much one developer comments on the issues of another; and (b) how much one developer changes another developer's code (and "heroes" are those that change the most code, all around the system). In a result endorsing the Hersleb hypothesis, in over 1000 open source projects, we find that "social-ness" is a statistically stronger indicate for code quality (number of bugs) than "hero-ness". Hence we say that debates over the merits of "hero-ness" is subtly misguided. Our results suggest that the real benefits of these so-called "heroes" is not so much the code they generate but the pattern of communication required when the interaction between a large community of programmers passes through a small group of centralized developers. To say that another way, to build better code, build better communication flows between core developers and the rest. In order to allow other researchers to confirm/improve/refute our results, all our scripts and data are available, on-line at https://github.com/Anonymous633671/A-Comparison-on-Communication-and-Code-Dependency-Effects-on-Software-Code-Quality.

Joymallya Chakraborty, Suvodeep Majumder, Huy Tu

Ethical bias in machine learning models has become a matter of concern in the software engineering community. Most of the prior software engineering works concentrated on finding ethical bias in models rather than fixing it. After finding bias, the next step is mitigation. Prior researchers mainly tried to use supervised approaches to achieve fairness. However, in the real world, getting data with trustworthy ground truth is challenging and also ground truth can contain human bias. Semi-supervised learning is a machine learning technique where, incrementally, labeled data is used to generate pseudo-labels for the rest of the data (and then all that data is used for model training). In this work, we apply four popular semi-supervised techniques as pseudo-labelers to create fair classification models. Our framework, Fair-SSL, takes a very small amount (10%) of labeled data as input and generates pseudo-labels for the unlabeled data. We then synthetically generate new data points to balance the training data based on class and protected attribute as proposed by Chakraborty et al. in FSE 2021. Finally, the classification model is trained on the balanced pseudo-labeled data and validated on test data. After experimenting on ten datasets and three learners, we find that Fair-SSL achieves similar performance as three state-of-the-art bias mitigation algorithms. That said, the clear advantage of Fair-SSL is that it requires only 10% of the labeled training data. To the best of our knowledge, this is the first SE work where semi-supervised techniques are used to fight against ethical bias in SE ML models.

Suvodeep Majumder, Joymallya Chakraborty, Gina R. Bai et al.

Testing machine learning software for ethical bias has become a pressing current concern. In response, recent research has proposed a plethora of new fairness metrics, for example, the dozens of fairness metrics in the IBM AIF360 toolkit. This raises the question: How can any fairness tool satisfy such a diverse range of goals? While we cannot completely simplify the task of fairness testing, we can certainly reduce the problem. This paper shows that many of those fairness metrics effectively measure the same thing. Based on experiments using seven real-world datasets, we find that (a) 26 classification metrics can be clustered into seven groups, and (b) four dataset metrics can be clustered into three groups. Further, each reduced set may actually predict different things. Hence, it is no longer necessary (or even possible) to satisfy all fairness metrics. In summary, to simplify the fairness testing problem, we recommend the following steps: (1)~determine what type of fairness is desirable (and we offer a handful of such types); then (2) lookup those types in our clusters; then (3) just test for one item per cluster.

Joymallya Chakraborty, Suvodeep Majumder, Tim Menzies

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.

N. C. Shrikanth, Suvodeep Majumder, Tim Menzies

Many researchers assume that, for software analytics, "more data is better." We write to show that, at least for learning defect predictors, this may not be true. To demonstrate this, we analyzed hundreds of popular GitHub projects. These projects ran for 84 months and contained 3,728 commits (median values). Across these projects, most of the defects occur very early in their life cycle. Hence, defect predictors learned from the first 150 commits and four months perform just as well as anything else. This means that, at least for the projects studied here, after the first few months, we need not continually update our defect prediction models. We hope these results inspire other researchers to adopt a "simplicity-first" approach to their work. Some domains require a complex and data-hungry analysis. But before assuming complexity, it is prudent to check the raw data looking for "short cuts" that can simplify the analysis.

Suvodeep Majumder, Pranav Mody, Tim Menzies

Numerous methods can build predictive models from software data. However, what methods and conclusions should we endorse as we move from analytics in-the-small (dealing with a handful of projects) to analytics in-the-large (dealing with hundreds of projects)? To answer this question, we recheck prior small-scale results (about process versus product metrics for defect prediction and the granularity of metrics) using 722,471 commits from 700 Github projects. We find that some analytics in-the-small conclusions still hold when scaling up to analytics in-the-large. For example, like prior work, we see that process metrics are better predictors for defects than product metrics (best process/product-based learners respectively achieve recalls of 98\%/44\% and AUCs of 95\%/54\%, median values). That said, we warn that it is unwise to trust metric importance results from analytics in-the-small studies since those change dramatically when moving to analytics in-the-large. Also, when reasoning in-the-large about hundreds of projects, it is better to use predictions from multiple models (since single model predictions can become confused and exhibit a high variance).

Joymallya Chakraborty, Suvodeep Majumder, Zhe Yu et al.

Machine learning software is increasingly being used to make decisions that affect people's lives. But sometimes, the core part of this software (the learned model), behaves in a biased manner that gives undue advantages to a specific group of people (where those groups are determined by sex, race, etc.). This "algorithmic discrimination" in the AI software systems has become a matter of serious concern in the machine learning and software engineering community. There have been works done to find "algorithmic bias" or "ethical bias" in the software system. Once the bias is detected in the AI software system, the mitigation of bias is extremely important. In this work, we a)explain how ground-truth bias in training data affects machine learning model fairness and how to find that bias in AI software,b)propose a methodFairwaywhich combines pre-processing and in-processing approach to remove ethical bias from training data and trained model. Our results show that we can find bias and mitigate bias in a learned model, without much damaging the predictive performance of that model. We propose that (1) test-ing for bias and (2) bias mitigation should be a routine part of the machine learning software development life cycle. Fairway offers much support for these two purposes.

Suvodeep Majumder, Nikhila Balaji, Katie Brey et al.

Deep learning methods are useful for high-dimensional data and are becoming widely used in many areas of software engineering. Deep learners utilizes extensive computational power and can take a long time to train-- making it difficult to widely validate and repeat and improve their results. Further, they are not the best solution in all domains. For example, recent results show that for finding related Stack Overflow posts, a tuned SVM performs similarly to a deep learner, but is significantly faster to train. This paper extends that recent result by clustering the dataset, then tuning very learners within each cluster. This approach is over 500 times faster than deep learning (and over 900 times faster if we use all the cores on a standard laptop computer). Significantly, this faster approach generates classifiers nearly as good (within 2\% F1 Score) as the much slower deep learning method. Hence we recommend this faster methods since it is much easier to reproduce and utilizes far fewer CPU resources. More generally, we recommend that before researchers release research results, that they compare their supposedly sophisticated methods against simpler alternatives (e.g applying simpler learners to build local models).