This paper studies unsupervised/self-supervised whole-graph representation learning, which is critical in many tasks such as molecule properties prediction in drug and material discovery. Existing methods mainly focus on preserving the local similarity structure between different graph instances but fail to discover the global semantic structure of the entire data set. In this paper, we propose a unified framework called Local-instance and Global-semantic Learning (GraphLoG) for self-supervised whole-graph representation learning. Specifically, besides preserving the local similarities, GraphLoG introduces the hierarchical prototypes to capture the global semantic clusters. An efficient online expectation-maximization (EM) algorithm is further developed for learning the model. We evaluate GraphLoG by pre-training it on massive unlabeled graphs followed by fine-tuning on downstream tasks. Extensive experiments on both chemical and biological benchmark data sets demonstrate the effectiveness of the proposed approach.
Label Smoothing (LS) is an effective regularizer to improve the generalization of state-of-the-art deep models. For each training sample the LS strategy smooths the one-hot encoded training signal by distributing its distribution mass over the non ground-truth classes, aiming to penalize the networks from generating overconfident output distributions. This paper introduces a novel label smoothing technique called Pairwise Label Smoothing (PLS). The PLS takes a pair of samples as input. Smoothing with a pair of ground-truth labels enables the PLS to preserve the relative distance between the two truth labels while further soften that between the truth labels and the other targets, resulting in models producing much less confident predictions than the LS strategy. Also, unlike current LS methods, which typically require to find a global smoothing distribution mass through cross-validation search, PLS automatically learns the distribution mass for each input pair during training. We empirically show that PLS significantly outperforms LS and the baseline models, achieving up to 30% of relative classification error reduction. We also visually show that when achieving such accuracy gains the PLS tends to produce very low winning softmax scores.
SkipGram word embedding models with negative sampling, or SGN in short, is an elegant family of word embedding models. In this paper, we formulate a framework for word embedding, referred to as Word-Context Classification (WCC), that generalizes SGN to a wide family of models. The framework, utilizing some "noise examples", is justified through a theoretical analysis. The impact of noise distribution on the learning of the WCC embedding models is studied experimentally, suggesting that the best noise distribution is in fact the data distribution, in terms of both the embedding performance and the speed of convergence during training. Along our way, we discover several novel embedding models that outperform the existing WCC models.
A fundamental problem in computational chemistry is to find a set of reactants to synthesize a target molecule, a.k.a. retrosynthesis prediction. Existing state-of-the-art methods rely on matching the target molecule with a large set of reaction templates, which are very computationally expensive and also suffer from the problem of coverage. In this paper, we propose a novel template-free approach called G2Gs by transforming a target molecular graph into a set of reactant molecular graphs. G2Gs first splits the target molecular graph into a set of synthons by identifying the reaction centers, and then translates the synthons to the final reactant graphs via a variational graph translation framework. Experimental results show that G2Gs significantly outperforms existing template-free approaches by up to 63% in terms of the top-1 accuracy and achieves a performance close to that of state-of-the-art template based approaches, but does not require domain knowledge and is much more scalable.
We consider the problem of learning a neural network classifier. Under the information bottleneck (IB) principle, we associate with this classification problem a representation learning problem, which we call "IB learning". We show that IB learning is, in fact, equivalent to a special class of the quantization problem. The classical results in rate-distortion theory then suggest that IB learning can benefit from a "vector quantization" approach, namely, simultaneously learning the representations of multiple input objects. Such an approach assisted with some variational techniques, result in a novel learning framework, "Aggregated Learning", for classification with neural network models. In this framework, several objects are jointly classified by a single neural network. The effectiveness of this framework is verified through extensive experiments on standard image recognition and text classification tasks.
As proteins with similar structures often have similar functions, analysis of protein structures can help predict protein functions and is thus important. We consider the problem of protein structure classification, which computationally classifies the structures of proteins into pre-defined groups. We develop a weighted network that depicts the protein structures, and more importantly, we propose the first graphlet-based measure that applies to weighted networks. Further, we develop a deep neural network (DNN) composed of both convolutional and recurrent layers to use this measure for classification. Put together, our approach shows dramatic improvements in performance over existing graphlet-based approaches on 36 real datasets. Even comparing with the state-of-the-art approach, it almost halves the classification error. In addition to protein structure networks, our weighted-graphlet measure and DNN classifier can potentially be applied to classification of other weighted networks in computational biology as well as in other domains.
Distant supervision for relation extraction enables one to effectively acquire structured relations out of very large text corpora with less human efforts. Nevertheless, most of the prior-art models for such tasks assume that the given text can be noisy, but their corresponding labels are clean. Such unrealistic assumption is contradictory with the fact that the given labels are often noisy as well, thus leading to significant performance degradation of those models on real-world data. To cope with this challenge, we propose a novel label-denoising framework that combines neural network with probabilistic modelling, which naturally takes into account the noisy labels during learning. We empirically demonstrate that our approach significantly improves the current art in uncovering the ground-truth relation labels.
In this work, we explain the working mechanism of MixUp in terms of adversarial training. We introduce a new class of adversarial training schemes, which we refer to as directional adversarial training, or DAT. In a nutshell, a DAT scheme perturbs a training example in the direction of another example but keeps its original label as the training target. We prove that MixUp is equivalent to a special subclass of DAT, in that it has the same expected loss function and corresponds to the same optimization problem asymptotically. This understanding not only serves to explain the effectiveness of MixUp, but also reveals a more general family of MixUp schemes, which we call Untied MixUp. We prove that the family of Untied MixUp schemes is equivalent to the entire class of DAT schemes. We establish empirically the existence of Untied Mixup schemes which improve upon MixUp.
Mixup, a recent proposed data augmentation method through linearly interpolating inputs and modeling targets of random samples, has demonstrated its capability of significantly improving the predictive accuracy of the state-of-the-art networks for image classification. However, how this technique can be applied to and what is its effectiveness on natural language processing (NLP) tasks have not been investigated. In this paper, we propose two strategies for the adaption of Mixup on sentence classification: one performs interpolation on word embeddings and another on sentence embeddings. We conduct experiments to evaluate our methods using several benchmark datasets. Our studies show that such interpolation strategies serve as an effective, domain independent data augmentation approach for sentence classification, and can result in significant accuracy improvement for both CNN and LSTM models.
MixUp, a data augmentation approach through mixing random samples, has been shown to be able to significantly improve the predictive accuracy of the current art of deep neural networks. The power of MixUp, however, is mostly established empirically and its working and effectiveness have not been explained in any depth. In this paper, we develop a theoretical understanding for MixUp as a form of out-of-manifold regularization, which constrains the model on the input space beyond the data manifold. This analytical study also enables us to identify MixUp's limitation caused by manifold intrusion, where synthetic samples collide with real examples of the manifold. Such intrusion gives rise to over regularization and thereby under-fitting. To address this issue, we further propose a novel regularizer, where mixing policies are adaptively learned from the data and a manifold intrusion loss is embraced as to avoid collision with the data manifold. We empirically show, using several benchmark datasets, our regularizer's effectiveness in terms of over regularization avoiding and accuracy improvement upon current art of deep classification models and MixUp.