Investigating better ways to reuse the released pre-trained language models (PLMs) can significantly reduce the computational cost and the potential environmental side-effects. This paper explores a novel PLM reuse paradigm, Knowledge Integration (KI). Without human annotations available, KI aims to merge the knowledge from different teacher-PLMs, each of which specializes in a different classification problem, into a versatile student model. To achieve this, we first derive the correlation between virtual golden supervision and teacher predictions. We then design a Model Uncertainty--aware Knowledge Integration (MUKI) framework to recover the golden supervision for the student. Specifically, MUKI adopts Monte-Carlo Dropout to estimate model uncertainty for the supervision integration. An instance-wise re-weighting mechanism based on the margin of uncertainty scores is further incorporated, to deal with the potential conflicting supervision from teachers. Experimental results demonstrate that MUKI achieves substantial improvements over baselines on benchmark datasets. Further analysis shows that MUKI can generalize well for merging teacher models with heterogeneous architectures, and even teachers major in cross-lingual datasets.

Few-shot dialogue state tracking (DST) is a realistic problem that trains the DST model with limited labeled data. Existing few-shot methods mainly transfer knowledge learned from external labeled dialogue data (e.g., from question answering, dialogue summarization, machine reading comprehension tasks, etc.) into DST, whereas collecting a large amount of external labeled data is laborious, and the external data may not effectively contribute to the DST-specific task. In this paper, we propose a few-shot DST framework called CSS, which Combines Self-training and Self-supervised learning methods. The unlabeled data of the DST task is incorporated into the self-training iterations, where the pseudo labels are predicted by a DST model trained on limited labeled data in advance. Besides, a contrastive self-supervised method is used to learn better representations, where the data is augmented by the dropout operation to train the model. Experimental results on the MultiWOZ dataset show that our proposed CSS achieves competitive performance in several few-shot scenarios.

Supervised learning algorithms generally assume the availability of enough memory to store data models during the training and test phases. However, this assumption is unrealistic when data comes in the form of infinite data streams, or when learning algorithms are deployed on devices with reduced amounts of memory. Such memory constraints impact the model behavior and assumptions. In this paper, we show that under memory constraints, increasing the size of a tree-based ensemble classifier can worsen its performance. In particular, we experimentally show the existence of an optimal ensemble size for a memory-bounded Mondrian forest on data streams and we design an algorithm to guide the forest toward that optimal number by using an estimation of overfitting. We tested different variations for this algorithm on a variety of real and simulated datasets, and we conclude that our method can achieve up to 95% of the performance of an optimally-sized Mondrian forest for stable datasets, and can even outperform it for datasets with concept drifts. All our methods are implemented in the OrpailleCC open-source library and are ready to be used on embedded systems and connected objects.

Improving the generalization of deep networks is an important open challenge, particularly in domains without plentiful data. The mixup algorithm improves generalization by linearly interpolating a pair of examples and their corresponding labels. These interpolated examples augment the original training set. Mixup has shown promising results in various classification tasks, but systematic analysis of mixup in regression remains underexplored. Using mixup directly on regression labels can result in arbitrarily incorrect labels. In this paper, we propose a simple yet powerful algorithm, C-Mixup, to improve generalization on regression tasks. In contrast with vanilla mixup, which picks training examples for mixing with uniform probability, C-Mixup adjusts the sampling probability based on the similarity of the labels. Our theoretical analysis confirms that C-Mixup with label similarity obtains a smaller mean square error in supervised regression and meta-regression than vanilla mixup and using feature similarity. Another benefit of C-Mixup is that it can improve out-of-distribution robustness, where the test distribution is different from the training distribution. By selectively interpolating examples with similar labels, it mitigates the effects of domain-associated information and yields domain-invariant representations. We evaluate C-Mixup on eleven datasets, ranging from tabular to video data. Compared to the best prior approach, C-Mixup achieves 6.56%, 4.76%, 5.82% improvements in in-distribution generalization, task generalization, and out-of-distribution robustness, respectively. Code is released at https://github.com/huaxiuyao/C-Mixup.

Multi-label learning (MLL) learns from the examples each associated with multiple labels simultaneously, where the high cost of annotating all relevant labels for each training example is challenging for real-world applications. To cope with the challenge, we investigate single-positive multi-label learning (SPMLL) where each example is annotated with only one relevant label, and show that one can successfully learn a theoretically grounded multi-label classifier for the problem. In this paper, a novel SPMLL method named SMILE, i.e., Single-positive MultI-label learning with Label Enhancement, is proposed. Specifically, an unbiased risk estimator is derived, which could be guaranteed to approximately converge to the optimal risk minimizer of fully supervised learning and shows that one positive label of each instance is sufficient to train the predictive model. Then, the corresponding empirical risk estimator is established via recovering the latent soft label as a label enhancement process, where the posterior density of the latent soft labels is approximate to the variational Beta density parameterized by an inference model. Experiments on benchmark datasets validate the effectiveness of the proposed method.

Machine learning technologies have been used in a wide range of practical systems. In practical situations, it is natural to expect the input-output pairs of a machine learning model to satisfy some requirements. However, it is difficult to obtain a model that satisfies requirements by just learning from examples. A simple solution is to add a module that checks whether the input-output pairs meet the requirements and then modifies the model's outputs. Such a module, which we call a {\em concurrent verifier} (CV), can give a certification, although how the generalizability of the machine learning model changes using a CV is unclear. This paper gives a generalization analysis of learning with a CV. We analyze how the learnability of a machine learning model changes with a CV and show a condition where we can obtain a guaranteed hypothesis using a verifier only in the inference time. We also show that typical error bounds based on Rademacher complexity will be no larger than that of the original model when using a CV in multi-class classification and structured prediction settings.

Federated Learning allows the training of machine learning models by using the computation and private data resources of many distributed clients. Most existing results on Federated Learning (FL) assume the clients have ground-truth labels. However, in many practical scenarios, clients may be unable to label task-specific data due to a lack of expertise or resource. We propose SemiFL to address the problem of combining communication-efficient FL such as FedAvg with Semi-Supervised Learning (SSL). In SemiFL, clients have completely unlabeled data and can train multiple local epochs to reduce communication costs, while the server has a small amount of labeled data. We provide a theoretical understanding of the success of data augmentation-based SSL methods to illustrate the bottleneck of a vanilla combination of communication-efficient FL with SSL. To address this issue, we propose alternate training to `fine-tune global model with labeled data' and `generate pseudo-labels with the global model.' We conduct extensive experiments and demonstrate that our approach significantly improves the performance of a labeled server with unlabeled clients training with multiple local epochs. Moreover, our method outperforms many existing SSFL baselines and performs competitively with the state-of-the-art FL and SSL results.

The centrality and diversity of the labeled data are very influential to the performance of semi-supervised learning (SSL), but most SSL models select the labeled data randomly. This study first construct a leading forest that forms a partially ordered topological space in an unsupervised way, and select a group of most representative samples to label with one shot (differs from active learning essentially) using property of homeomorphism. Then a kernelized large margin metric is efficiently learned for the selected data to classify the remaining unlabeled sample. Optimal leading forest (OLF) has been observed to have the advantage of revealing the difference evolution along a path within a subtree. Therefore, we formulate an optimization problem based on OLF to select the samples. Also with OLF, the multiple local metrics learning is facilitated to address multi-modal and mix-modal problem in SSL, especially when the number of class is large. Attribute to this novel design, stableness and accuracy of the performance is significantly improved when compared with the state-of-the-art graph SSL methods. The extensive experimental studies have shown that the proposed method achieved encouraging accuracy and efficiency. Code has been made available at https://github.com/alanxuji/DeLaLA.

State-of-the-art (SOTA) semi-supervised learning (SSL) methods have been highly successful in leveraging a mix of labeled and unlabeled data by combining techniques of consistency regularization and pseudo-labeling. During pseudo-labeling, the model's predictions on unlabeled data are used for training and thus, model calibration is important in mitigating confirmation bias. Yet, many SOTA methods are optimized for model performance, with little focus directed to improve model calibration. In this work, we empirically demonstrate that model calibration is strongly correlated with model performance and propose to improve calibration via approximate Bayesian techniques. We introduce a family of new SSL models that optimizes for calibration and demonstrate their effectiveness across standard vision benchmarks of CIFAR-10, CIFAR-100 and ImageNet, giving up to 15.9% improvement in test accuracy. Furthermore, we also demonstrate their effectiveness in additional realistic and challenging problems, such as class-imbalanced datasets and in photonics science.

Unsupervised representational learning is now on the very rim to take over supervised representation learning. It is supposed to be a perfect solver for real-world scenarios full of unlabeled data. Among them, self-supervised learning has drawn the most attention for its good data efficiency and generalizability. Self-supervised learning typically involves a proxy task to learn discriminative representations from self-derived labels. Among all manners of these proxy tasks [13, 18, 29, 33, 36], instance discrimination [31, 49], known as contrastive representation learning, has emerged as the most effective paradigm. Its subsequent methods [7, 22, 24, 41, 59] have greatly reduced the gap between unsupervised and supervised learning. Specifically, instance discrimination features a Single Instance Multi-view (SIM) paradigm to separate different instances. It seeks to narrow the distance among multiple views of the same instance (e.g. an image), which are typically yielded from vanilla data augmentation policies like color jittering, cropping, resizing, applying Gaussian noise.