Adversarial Learning for Cross-Project Semi-Supervised Defect Prediction

Abstract

Cross-project defect prediction (CPDP) aims to build a prediction model on existing source projects and predict the labels of target project. The data distribution difference between different projects makes CPDP very challenging. Besides, most existing CPDP methods usually require sufficient and labeled data. However, acquiring lots of labeled data for a new project is difficult while obtaining the unlabeled data is relatively easy. A desirable approach is building a prediction model on unlabeled data and labeled data. CPDP in this scenario is called cross-project semi-supervised defect prediction (CSDP). Recently, generative adversarial networks have achieved impressive results with these strong ability of learning data distribution and discriminative representation. For effectively learning the discriminative features of data from different projects, we propose a Discriminative Adversarial Feature Learning (DAFL) approach for CSDP. DAFL consists of feature transformer and project discriminator, which compete with each other. A feature transformer tries to generate feature representation, which learns the discriminant information and preserves intrinsic structure inferred from both labeled and unlabeled data. A project discriminator tries to discriminate source and target instances on the generated representation. Experiments on 16 projects show that DAFL performs significantly better than baselines. INDEX TERMS Cross-project defect prediction, adversarial learning, semi-supervised learning.

Disentangled Adversarial Autoencoder for Subject-Invariant Physiological Feature Extraction

“…Under such restrictions, a robust feature extractor can be constructed with transfer learning [10][11][12], which tries to discover shared data features that are invariant across subjects and tasks. In particular, promising results were demonstrated for transfer learning by censoring nuisance features via adversarial training [13][14][15][16][17][18][19]. These works use adversarial methods to learn universal features shared by an attribute group, where a discriminative unit distinguishes shared features with respect to the different attributes adversarially to the feature extractor.…”

Unsupervised Domain Adaptation Based on Discriminative Subspace Learning for Cross-Project Defect Prediction

“…CKSDL combines cost-sensitive technique and kernel technique to further improve the prediction model. Then, Sun et al [18] introduced adversarial learning into CPDP and embedded a triplet sampling strategy into an adversarial learning framework.…”

Universal Physiological Representation Learning with Soft-Disentangled Rateless Autoencoders

“…D-cAE and D-cRAE represent cAE with hard-split and soft-split bottleneck variables respectively linked to the nuisance network only. DA-cAE and DA-cRAE specify hard-split and soft-split representations connected to both adversary and nuisance networks respectively with decoder conditioned on s. Note that the A-cAE resembles to the traditional adversarial learning methods presented in [20][21][22][23] where only one adversarial unit is adopted.…”

“…Addressing biosignal datasets collected from a narrow amount of subjects, transfer learning methods [11][12][13][14] are applied to build strong feature learning machines to extract robust and invariant features across various tasks and/or unknown subjects. Particularly, adversarial transfer learning [15][16][17][18][19][20][21][22][23] demonstrated impressive results in constructing such discriminative feature extractors. Traditional adversarial transfer learning works aim to extract latent representations universally shared by a group of attributes using adversarial inference, where a discriminative network is trained adversarially towards the feature extractor in order to differentiate universal features from various attributes.…”