Variational Autoencoder (VAE) is widely used as a generative model to approximate a model’s posterior on latent variables by combining the amortized variational inference and deep neural networks. However, when paired with strong autoregressive decoders, VAE often converges to a degenerated local optimum known as “posterior collapse”. Previous approaches consider the Kullback–Leibler divergence (KL) individual for each datapoint. We propose to let the KL follow a distribution across the whole dataset, and analyze that it is sufficient to prevent posterior collapse by keeping the expectation of the KL’s distribution positive. Then we propose Batch Normalized-VAE (BN-VAE), a simple but effective approach to set a lower bound of the expectation by regularizing the distribution of the approximate posterior’s parameters. Without introducing any new model component or modifying the objective, our approach can avoid the posterior collapse effectively and efficiently. We further show that the proposed BN-VAE can be extended to conditional VAE (CVAE). Empirically, our approach surpasses strong autoregressive baselines on language modeling, text classification and dialogue generation, and rivals more complex approaches while keeping almost the same training time as VAE.
In this paper, we focus on natural language video localization: localizing (ie, grounding) a natural language description in a long and untrimmed video sequence. All currently published models for addressing this problem can be categorized into two types: (i) top-down approach: it does classification and regression for a set of pre-cut video segment candidates; (ii) bottom-up approach: it directly predicts probabilities for each video frame as the temporal boundaries (ie, start and end time point). However, both two approaches suffer several limitations: the former is computation-intensive for densely placed candidates, while the latter has trailed the performance of the top-down counterpart thus far. To this end, we propose a novel dense bottom-up framework: DEnse Bottom-Up Grounding (DEBUG). DEBUG regards all frames falling in the ground truth segment as foreground, and each foreground frame regresses the unique distances from its location to bi-directional ground truth boundaries. Extensive experiments on three challenging benchmarks (TACoS, Charades-STA, and ActivityNet Captions) show that DEBUG is able to match the speed of bottom-up models while surpassing the performance of the state-of-the-art top-down models.
Discovering the latent topics within texts has been a fundamental task for many applications. However, conventional topic models suffer different problems in different settings. The Latent Dirichlet Allocation (LDA) may not work well for short texts due to the data sparsity (i.e. the sparse word co-occurrence patterns in short documents). The Biterm Topic Model (BTM) learns topics by modeling the word-pairs named biterms in the whole corpus. This assumption is very strong when documents are long with rich topic information and do not exhibit the transitivity of biterms. In this paper, we propose a novel way called GraphBTM to represent biterms as graphs and design a Graph Convolutional Networks (GCNs) with residual connections to extract transitive features from biterms. To overcome the data sparsity of LDA and the strong assumption of BTM, we sample a fixed number of documents to form a mini-corpus as a sample. We also propose a dataset called All News extracted from 15 news publishers, in which documents are much longer than 20 Newsgroups. We present an amortized variational inference method for GraphBTM. Our method generates more coherent topics compared with previous approaches. Experiments show that the sampling strategy improves performance by a large margin.