This paper studies Learning from Observations (LfO) for imitation learning with access to state-only demonstrations. In contrast to Learning from Demonstration (LfD) that involves both action and state supervision, LfO is more practical in leveraging previously inapplicable resources (e.g. videos), yet more challenging due to the incomplete expert guidance. In this paper, we investigate LfO and its difference with LfD in both theoretical and practical perspectives. We first prove that the gap between LfD and LfO actually lies in the disagreement of inverse dynamics models between the imitator and the expert, if following the modeling approach of GAIL. More importantly, the upper bound of this gap is revealed by a negative causal entropy which can be minimized in a model-free way. We term our method as Inverse-Dynamics-Disagreement-Minimization (IDDM) which enhances the conventional LfO method through further bridging the gap to LfD. Considerable empirical results on challenging benchmarks indicate that our method attains consistent improvements over other LfO counterparts.
Deep learning based medical image diagnosis has shown great potential in clinical medicine. However, it often suffers two major difficulties in practice: 1) only limited labeled samples are available due to expensive annotation costs over medical images; 2) labeled images may contain considerable label noises (e.g., mislabeling labels) due to diagnostic difficulties. In this paper, we seek to exploit rich labeled data from relevant domains to help the learning in the target task with unsupervised domain adaptation (UDA). Unlike most existing UDA methods which rely on clean labeled data or assume samples are equally transferable, we propose a novel Collaborative Unsupervised Domain Adaptation algorithm to conduct transferability-aware domain adaptation and conquer label noise in a cooperative way. Promising empirical results verify the superiority of the proposed method.
Designing effective architectures is one of the key factors behind the success of deep neural networks. Existing deep architectures are either manually designed or automatically searched by some Neural Architecture Search (NAS) methods. However, even a well-searched architecture may still contain many non-significant or redundant modules or operations (e.g., convolution or pooling), which may not only incur substantial memory consumption and computation cost but also deteriorate the performance. Thus, it is necessary to optimize the operations inside an architecture to improve the performance without introducing extra computation cost. Unfortunately, such a constrained optimization problem is NP-hard. To make the problem feasible, we cast the optimization problem into a Markov decision process (MDP) and seek to learn a Neural Architecture Transformer (NAT) to replace the redundant operations with the more computationally efficient ones (e.g., skip connection or directly removing the connection). Based on MDP, we learn NAT by exploiting reinforcement learning to obtain the optimization policies w.r.t. different architectures. To verify the effectiveness of the proposed strategies, we apply NAT on both hand-crafted architectures and NAS based architectures. Extensive experiments on two benchmark datasets, i.e., CIFAR-10 and ImageNet, demonstrate that the transformed architecture by NAT significantly outperforms both its original form and those architectures optimized by existing methods.
Towards the challenging problem of semi-supervised node classification, there have been extensive studies. As a frontier, Graph Neural Networks (GNNs) have aroused great interest recently, which update the representation of each node by aggregating information of its neighbors. However, most GNNs have shallow layers with a limited receptive field and may not achieve satisfactory performance especially when the number of labeled nodes is quite small. To address this challenge, we innovatively propose a graph few-shot learning (GFL) algorithm that incorporates prior knowledge learned from auxiliary graphs to improve classification accuracy on the target graph. Specifically, a transferable metric space characterized by a node embedding and a graph-specific prototype embedding function is shared between auxiliary graphs and the target, facilitating the transfer of structural knowledge. Extensive experiments and ablation studies on four real-world graph datasets demonstrate the effectiveness of our proposed model.
Molecule generation is to design new molecules with specific chemical properties and further to optimize the desired chemical properties. Following previous work, we encode molecules into continuous vectors in the latent space and then decode the vectors into molecules under the variational autoencoder (VAE) framework. We investigate the posterior collapse problem of current RNN-based VAEs for molecule sequence generation. For the first time, we find that underestimated reconstruction loss leads to posterior collapse, and provide both theoretical and experimental evidence. We propose an effective and efficient solution to fix the problem and avoid posterior collapse. Without bells and whistles, our method achieves SOTA reconstruction accuracy and competitive validity on the ZINC 250K dataset. When generating 10,000 unique valid SMILES from random prior sampling, it costs JT-VAE1450s while our method only needs 9s. Our implementation is at https://github.com/chaoyan1037/Re-balanced-VAE.
Automated machine learning aims to automate the whole process of machine learning, including model configuration. In this paper, we focus on automated hyperparameter optimization (HPO) based on sequential model-based optimization (SMBO). Though conventional SMBO algorithms work well when abundant HPO trials are available, they are far from satisfactory in practical applications where a trial on a huge dataset may be so costly that an optimal hyperparameter configuration is expected to return in as few trials as possible. Observing that human experts draw on their expertise in a machine learning model by trying configurations that once performed well on other datasets, we are inspired to speed up HPO by transferring knowledge from historical HPO trials on other datasets. We propose an end-to-end and efficient HPO algorithm named as Transfer Neural Processes (TNP), which achieves transfer learning by incorporating trials on other datasets, initializing the model with well-generalized parameters, and learning an initial set of hyperparameters to evaluate. Experiments on extensive OpenML datasets and three computer vision datasets show that the proposed model can achieve state-of-the-art performance in at least one order of magnitude less trials.
Existing Graph Convolutional Networks (GCNs) are shallow---the number of the layers is usually not larger than 2. The deeper variants by simply stacking more layers, unfortunately perform worse, even involving well-known tricks like weight penalizing, dropout, and residual connections. This paper reveals that developing deep GCNs mainly encounters two obstacles: \emph{over-fitting} and \emph{over-smoothing}. The over-fitting issue weakens the generalization ability on small graphs, while over-smoothing impedes model training by isolating output representations from the input features with the increase in network depth. Hence, we propose DropEdge, a novel technique to alleviate both issues. At its core, DropEdge randomly removes a certain number of edges from the input graphs, acting like a data augmenter and also a message passing reducer. More importantly, DropEdge enables us to recast a wider range of Convolutional Neural Networks (CNNs) from the image field to the graph domain; in particular, we study DenseNet and InceptionNet in this paper. Extensive experiments on several benchmarks demonstrate that our method allows deep GCNs to achieve promising performance, even when the number of layers exceeds 30---the deepest GCN that has ever been proposed.
Most state-of-the-art action localization systems process each action proposal individually, without explicitly exploiting their relations during learning. However, the relations between proposals actually play an important role in action localization, since a meaningful action always consists of multiple proposals in a video. In this paper, we propose to exploit the proposal-proposal relations using Graph Convolutional Networks (GCNs). First, we construct an action proposal graph, where each proposal is represented as a node and their relations between two proposals as an edge. Here, we use two types of relations, one for capturing the context information for each proposal and the other one for characterizing the correlations between distinct actions. Then we apply the GCNs over the graph to model the relations among different proposals and learn powerful representations for the action classification and localization. Experimental results show that our approach significantly outperforms the state-of-the-art on THUMOS14 (49.1% versus 42.8%). Moreover, augmentation experiments on ActivityNet also verify the efficacy of modeling action proposal relationships. Codes are available at https://github.com/Alvin-Zeng/PGCN.