In the past years, significant improvements in the field of neural architecture search(NAS) have been made. However, it is still challenging to search for efficient networks due to the gap between the searched constraint and real inference time exist
s. To search for a high-performance network with low inference time, several previous works set a computational complexity constraint for the search algorithm. However, many factors affect the speed of inference(e.g., FLOPs, MACs). The correlation between a single indicator and the latency is not strong. Currently, some re-parameterization(Rep) techniques are proposed to convert multi-branch to single-path architecture which is inference-friendly. Nevertheless, multi-branch architectures are still human-defined and inefficient. In this work, we propose a new search space that is suitable for structural re-parameterization techniques. RepNAS, a one-stage NAS approach, is present to efficiently search the optimal diverse branch block(ODBB) for each layer under the branch number constraint. Our experimental results show the searched ODBB can easily surpass the manual diverse branch block(DBB) with efficient training. Code and models will be available sooner.
The content on the web is in a constant state of flux. New entities, issues, and ideas continuously emerge, while the semantics of the existing conversation topics gradually shift. In recent years, pre-trained language models like BERT greatly improv
ed the state-of-the-art for a large spectrum of content understanding tasks. Therefore, in this paper, we aim to study how these language models can be adapted to better handle continuously evolving web content. In our study, we first analyze the evolution of 2013 - 2019 Twitter data, and unequivocally confirm that a BERT model trained on past tweets would heavily deteriorate when directly applied to data from later years. Then, we investigate two possible sources of the deterioration: the semantic shift of existing tokens and the sub-optimal or failed understanding of new tokens. To this end, we both explore two different vocabulary composition methods, as well as propose three sampling methods which help in efficient incremental training for BERT-like models. Compared to a new model trained from scratch offline, our incremental training (a) reduces the training costs, (b) achieves better performance on evolving content, and (c) is suitable for online deployment. The superiority of our methods is validated using two downstream tasks. We demonstrate significant improvements when incrementally evolving the model from a particular base year, on the task of Country Hashtag Prediction, as well as on the OffensEval 2019 task.
Document layout comprises both structural and visual (eg. font-sizes) information that is vital but often ignored by machine learning models. The few existing models which do use layout information only consider textual contents, and overlook the exi
stence of contents in other modalities such as images. Additionally, spatial interactions of presented contents in a layout were never really fully exploited. To bridge this gap, we parse a document into content blocks (eg. text, table, image) and propose a novel layout-aware multimodal hierarchical framework, LAMPreT, to model the blocks and the whole document. Our LAMPreT encodes each block with a multimodal transformer in the lower-level and aggregates the block-level representations and connections utilizing a specifically designed transformer at the higher-level. We design hierarchical pretraining objectives where the lower-level model is trained similarly to multimodal grounding models, and the higher-level model is trained with our proposed novel layout-aware objectives. We evaluate the proposed model on two layout-aware tasks -- text block filling and image suggestion and show the effectiveness of our proposed hierarchical architecture as well as pretraining techniques.
Auto-ML pruning methods aim at searching a pruning strategy automatically to reduce the computational complexity of deep Convolutional Neural Networks(deep CNNs). However, some previous works found that the results of many Auto-ML pruning methods eve
n cannot surpass the results of the uniformly pruning method. In this paper, we first analyze the reason for the ineffectiveness of Auto-ML pruning. Subsequently, a stage-wise pruning(SP) method is proposed to solve the above problem. As with most of the previous Auto-ML pruning methods, SP also trains a super-net that can provide proxy performance for sub-nets and search the best sub-net who has the best proxy performance. Different from previous works, we split a deep CNN into several stages and use a full-net where all layers are not pruned to supervise the training and the searching of sub-nets. Remarkably, the proxy performance of sub-nets trained with SP is closer to the actual performance than most of the previous Auto-ML pruning works. Therefore, SP achieves the state-of-the-art on both CIFAR-10 and ImageNet under the mobile setting.
Search engines often follow a two-phase paradigm where in the first stage (the retrieval stage) an initial set of documents is retrieved and in the second stage (the re-ranking stage) the documents are re-ranked to obtain the final result list. While
deep neural networks were shown to improve the performance of the re-ranking stage in previous works, there is little literature about using deep neural networks to improve the retrieval stage. In this paper, we study the merits of combining deep neural network models and lexical models for the retrieval stage. A hybrid approach, which leverages both semantic (deep neural network-based) and lexical (keyword matching-based) retrieval models, is proposed. We perform an empirical study, using a publicly available TREC collection, which demonstrates the effectiveness of our approach and sheds light on the different characteristics of the semantic approach, the lexical approach, and their combination.
This paper presents a novel framework to build a voice conversion (VC) system by learning from a text-to-speech (TTS) synthesis system, that is called TTS-VC transfer learning. We first develop a multi-speaker speech synthesis system with sequence-to
-sequence encoder-decoder architecture, where the encoder extracts robust linguistic representations of text, and the decoder, conditioned on target speaker embedding, takes the context vectors and the attention recurrent network cell output to generate target acoustic features. We take advantage of the fact that TTS system maps input text to speaker independent context vectors, and reuse such a mapping to supervise the training of latent representations of an encoder-decoder voice conversion system. In the voice conversion system, the encoder takes speech instead of text as input, while the decoder is functionally similar to TTS decoder. As we condition the decoder on speaker embedding, the system can be trained on non-parallel data for any-to-any voice conversion. During voice conversion training, we present both text and speech to speech synthesis and voice conversion networks respectively. At run-time, the voice conversion network uses its own encoder-decoder architecture. Experiments show that the proposed approach outperforms two competitive voice conversion baselines consistently, namely phonetic posteriorgram and variational autoencoder methods, in terms of speech quality, naturalness, and speaker similarity.
Previous AutoML pruning works utilized individual layer features to automatically prune filters. We analyze the correlation for two layers from different blocks which have a short-cut structure. It is found that, in one block, the deeper layer has ma
ny redundant filters which can be represented by filters in the former layer so that it is necessary to take information from other layers into consideration in pruning. In this paper, a graph pruning approach is proposed, which views any deep model as a topology graph. Graph PruningNet based on the graph convolution network is designed to automatically extract neighboring information for each node. To extract features from various topologies, Graph PruningNet is connected with Pruned Network by an individual fully connection layer for each node and jointly trained on a training dataset from scratch. Thus, we can obtain reasonable weights for any size of sub-network. We then search the best configuration of the Pruned Network by reinforcement learning. Different from previous work, we take the node features from well-trained Graph PruningNet, instead of the hand-craft features, as the states in reinforcement learning. Compared with other AutoML pruning works, our method has achieved the state-of-the-art under same conditions on ImageNet-2012. The code will be released on GitHub.
We investigated the training of a shared model for both text-to-speech (TTS) and voice conversion (VC) tasks. We propose using an extended model architecture of Tacotron, that is a multi-source sequence-to-sequence model with a dual attention mechani
sm as the shared model for both the TTS and VC tasks. This model can accomplish these two different tasks respectively according to the type of input. An end-to-end speech synthesis task is conducted when the model is given text as the input while a sequence-to-sequence voice conversion task is conducted when it is given the speech of a source speaker as the input. Waveform signals are generated by using WaveNet, which is conditioned by using a predicted mel-spectrogram. We propose jointly training a shared model as a decoder for a target speaker that supports multiple sources. Listening experiments show that our proposed multi-source encoder-decoder model can efficiently achieve both the TTS and VC tasks.
So far, many of the deep learning approaches for voice conversion produce good quality speech by using a large amount of training data. This paper presents a Deep Bidirectional Long Short-Term Memory (DBLSTM) based voice conversion framework that can
work with a limited amount of training data. We propose to implement a DBLSTM based average model that is trained with data from many speakers. Then, we propose to perform adaptation with a limited amount of target data. Last but not least, we propose an error reduction network that can improve the voice conversion quality even further. The proposed framework is motivated by three observations. Firstly, DBLSTM can achieve a remarkable voice conversion by considering the long-term dependencies of the speech utterance. Secondly, DBLSTM based average model can be easily adapted with a small amount of data, to achieve a speech that sounds closer to the target. Thirdly, an error reduction network can be trained with a small amount of training data, and can improve the conversion quality effectively. The experiments show that the proposed voice conversion framework is flexible to work with limited training data and outperforms the traditional frameworks in both objective and subjective evaluations.
