Mixup is a regularization technique that artificially produces new samples using convex combinations of original training points. This simple technique has shown strong empirical performance, and has been heavily used as part of semi-supervised learning techniques such as mixmatch~\citep{berthelot2019mixmatch} and interpolation consistent training (ICT)~\citep{verma2019interpolation}. In this paper, we look at Mixup through a \emph{representation learning} lens in a semi-supervised learning setup. In particular, we study the role of Mixup in promoting linearity in the learned network representations. Towards this, we study two questions: (1) how does the Mixup loss that enforces linearity in the \emph{last} network layer propagate the linearity to the \emph{earlier} layers?; and (2) how does the enforcement of stronger Mixup loss on more than two data points affect the convergence of training? We empirically investigate these properties of Mixup on vision datasets such as CIFAR-10, CIFAR-100 and SVHN. Our results show that supervised Mixup training does not make \emph{all} the network layers linear; in fact the \emph{intermediate layers} become more non-linear during Mixup training compared to a network that is trained \emph{without} Mixup. However, when Mixup is used as an unsupervised loss, we observe that all the network layers become more linear resulting in faster training convergence.

Abstract: Object detection for autonomous vehicles has received increasing attention in recent years, where labeled data are often expensive while unlabeled data can be collected readily, calling for research on semi-supervised learning for this area. Existing semi-supervised object detection (SSOD) methods usually assume that the labeled and unlabeled data come from the same data distribution. In autonomous driving, however, data are usually collected from different scenarios, such as different weather conditions or different times in a day. Motivated by this, we study a novel but challenging domain inconsistent SSOD problem. It involves two kinds of distribution shifts among different domains, including (1) data distribution discrepancy, and (2) class distribution shifts, making existing SSOD methods suffer from inaccurate pseudo-labels and hurting model performance.

Physiological measurements involves observing variables that attribute to the normative functioning of human systems and subsystems directly or indirectly. The measurements can be used to detect affective states of a person with aims such as improving human-computer interactions. There are several methods of collecting physiological data, but wearable sensors are a common, non-invasive tool for accurate readings. However, valuable information is hard to extract from the raw physiological data, especially for affective state detection. Machine Learning techniques are used to detect the affective state of a person through labeled physiological data. A clear problem with using labeled data is creating accurate labels. An expert is needed to analyze a form of recording of participants and mark sections with different states such as stress and calm. While expensive, this method delivers a complete dataset with labeled data that can be used in any number of supervised algorithms. An interesting question arises from the expensive labeling: how can we reduce the cost while maintaining high accuracy? Semi-Supervised learning (SSL) is a potential solution to this problem. These algorithms allow for machine learning models to be trained with only a small subset of labeled data (unlike unsupervised which use no labels). They provide a way of avoiding expensive labeling. This paper compares a fully supervised algorithm to a SSL on the public WESAD (Wearable Stress and Affect Detection) Dataset for stress detection. This paper shows that Semi-Supervised algorithms are a viable method for inexpensive affective state detection systems with accurate results.

Many clinical studies require the follow-up of patients over time. This is challenging: apart from frequently observed drop-out, there are often also organizational and financial challenges, which can lead to reduced data collection and, in turn, can complicate subsequent analyses. In contrast, there is often plenty of baseline data available of patients with similar characteristics and background information, e.g., from patients that fall outside the study time window. In this article, we investigate whether we can benefit from the inclusion of such unlabeled data instances to predict accurate survival times. In other words, we introduce a third level of supervision in the context of survival analysis, apart from fully observed and censored instances, we also include unlabeled instances. We propose three approaches to deal with this novel setting and provide an empirical comparison over fifteen real-life clinical and gene expression survival datasets. Our results demonstrate that all approaches are able to increase the predictive performance over independent test data. We also show that integrating the partial supervision provided by censored data in a semi-supervised wrapper approach generally provides the best results, often achieving high improvements, compared to not using unlabeled data.

Most previous neural text-to-speech (TTS) methods are mainly based on supervised learning methods, which means they depend on a large training dataset and hard to achieve comparable performance under low-resource conditions. To address this issue, we propose a semi-supervised learning method for neural TTS in which labeled target data is limited, which can also resolve the problem of exposure bias in the previous auto-regressive models. Specifically, we pre-train the reference model based on Fastspeech2 with much source data, fine-tuned on a limited target dataset. Meanwhile, pseudo labels generated by the original reference model are used to guide the fine-tuned model's training further, achieve a regularization effect, and reduce the overfitting of the fine-tuned model during training on the limited target data. Experimental results show that our proposed semi-supervised learning scheme with limited target data significantly improves the voice quality for test data to achieve naturalness and robustness in speech synthesis.

We analyze the growth of dataset sizes used in machine learning for natural language processing and computer vision, and extrapolate these using two methods; using the historical growth rate and estimating the compute-optimal dataset size for future predicted compute budgets. We investigate the growth in data usage by estimating the total stock of unlabeled data available on the internet over the coming decades. Our analysis indicates that the stock of high-quality language data will be exhausted soon; likely before 2026. By contrast, the stock of low-quality language data and image data will be exhausted only much later; between 2030 and 2050 (for low-quality language) and between 2030 and 2060 (for images). Our work suggests that the current trend of ever-growing ML models that rely on enormous datasets might slow down if data efficiency is not drastically improved or new sources of data become available.

Quantum machine learning promises to efficiently solve important problems. There are two persistent challenges in classical machine learning: the lack of labeled data, and the limit of computational power. We propose a novel framework that resolves both issues: quantum semi-supervised learning. Moreover, we provide a protocol in systematically designing quantum machine learning algorithms with quantum supremacy, which can be extended beyond quantum semi-supervised learning. In the meantime, we show that naive quantum matrix product estimation algorithm outperforms the best known classical matrix multiplication algorithm. We showcase two concrete quantum semi-supervised learning algorithms: a quantum self-training algorithm named the propagating nearest-neighbor classifier, and the quantum semi-supervised K-means clustering algorithm. By doing time complexity analysis, we conclude that they indeed possess quantum supremacy.

Self-training methods have been explored in recent years and have exhibited great performance in improving semi-supervised learning. This work presents a Simple instance-Adaptive self-Training method (SAT) for semi-supervised text classification. SAT first generates two augmented views for each unlabeled data and then trains a meta-learner to automatically identify the relative strength of augmentations based on the similarity between the original view and the augmented views. The weakly-augmented view is fed to the model to produce a pseudo-label and the strongly-augmented view is used to train the model to predict the same pseudo-label. We conducted extensive experiments and analyses on three text classification datasets and found that with varying sizes of labeled training data, SAT consistently shows competitive performance compared to existing semi-supervised learning methods. Our code can be found at \url{https://github.com/declare-lab/SAT.git}.

Class imbalance in a dataset is one of the major challenges that can significantly impact the performance of machine learning models resulting in biased predictions. Numerous techniques have been proposed to address class imbalanced problems, including, but not limited to, Oversampling, Undersampling, and cost-sensitive approaches. Due to its ability to generate synthetic data, oversampling techniques such as the Synthetic Minority Oversampling Technique (SMOTE) is among the most widely used methodology by researchers. However, one of SMOTE's potential disadvantages is that newly created minor samples may overlap with major samples. As an effect, the probability of ML models' biased performance towards major classes increases. Recently, generative adversarial network (GAN) has garnered much attention due to its ability to create almost real samples. However, GAN is hard to train even though it has much potential. This study proposes two novel techniques: GAN-based Oversampling (GBO) and Support Vector Machine-SMOTE-GAN (SSG) to overcome the limitations of the existing oversampling approaches. The preliminary computational result shows that SSG and GBO performed better on the expanded imbalanced eight benchmark datasets than the original SMOTE. The study also revealed that the minor sample generated by SSG demonstrates Gaussian distributions, which is often difficult to achieve using original SMOTE.

Although most of the applications of Machine Learning in our world are based on supervised machine learning algorithms and that's why this is where most of the investment goes into this direction. However, the majority of the available data is actually unlabeled: we have the input feature X, but we do not have the labels y. From here comes the importance of unsupervised learning. There are a lot of applications for unsupervised learning for example if you want to create a system that will take a few pictures of each item on a manufacturing production line and detect which items are defective. You can fairly easily create a system that will take pictures automatically, and this might give you thousands of pictures every day.