Recently, deep learning has been exploited in many fields with revolutionary breakthroughs. In the light of this, deep learning-assisted communication systems have also attracted much attention in recent years and have potential to break down the conventional design rule for communication systems. In this work, we propose two kinds of neural network-based equalizers to exploit different characteristics between convolutional neural networks and recurrent neural networks. The equalizer in conventional block-based design may destroy the code structure and degrade the capacity of coding gain for decoder. On the contrary, our proposed approach not only eliminates channel fading, but also exploits the code structure with utilization of coding gain in advance, which can effectively increase the overall utilization of coding gain with more than 1.5 dB gain.
This paper proposed a novel approach for the detection and reconstruction of dysarthric speech. The encoder-decoder model factorizes speech into a low-dimensional latent space and encoding of the input text. We showed that the latent space conveys interpretable characteristics of dysarthria, such as intelligibility and fluency of speech. MUSHRA perceptual test demonstrated that the adaptation of the latent space let the model generate speech of improved fluency. The multi-task supervised approach for predicting both the probability of dysarthric speech and the mel-spectrogram helps improve the detection of dysarthria with higher accuracy. This is thanks to a low-dimensional latent space of the auto-encoder as opposed to directly predicting dysarthria from a highly dimensional mel-spectrogram.
Predicting distant future trajectories of agents in a dynamic scene is not an easy problem because the future trajectory of an agent is affected by not only his/her past trajectory but also the scene contexts. To tackle this problem, we propose a model based on recurrent neural networks (RNNs) and a novel method for training the model. The proposed model is based on an encoder-decoder architecture where the encoder encodes inputs (past trajectories and scene context information) while the decoder produces a trajectory from the context vector given by the encoder. We train the networks of the proposed model to produce a future trajectory, which is the closest to the true trajectory, while maximizing a reward from a reward function. The reward function is also trained at the same time to maximize the margin between the rewards from the ground-truth trajectory and its estimate. The reward function plays the role of a regularizer for the proposed model so the trained networks are able to better utilize the scene context information for the prediction task. We evaluated the proposed model on several public datasets. Experimental results show that the prediction performance of the proposed model is much improved by the regularization, which outperforms the-state-of-the-arts in terms of accuracy. The implementation codes are available at https://github.com/d1024choi/traj-pred-irl/.
The application of handwritten text recognition to historical works is highly dependant on accurate text line retrieval. A number of systems utilizing a robust baseline detection paradigm have emerged recently but the advancement of layout analysis methods for challenging scripts is held back by the lack of well-established datasets including works in non-Latin scripts. We present a dataset of 400 annotated document images from different domains and time periods. A short elaboration on the particular challenges posed by handwriting in Arabic script for layout analysis and subsequent processing steps is given. Lastly, we propose a method based on a fully convolutional encoder-decoder network to extract arbitrarily shaped text line images from manuscripts.
Multiple sequence to sequence models were used to establish an end-to-end multi-turns proactive dialogue generation agent, with the aid of data augmentation techniques and variant encoder-decoder structure designs. A rank-based ensemble approach was developed for boosting performance. Results indicate that our single model, in average, makes an obvious improvement in the terms of F1-score and BLEU over the baseline by 18.67% on the DuConv dataset. In particular, the ensemble methods further significantly outperform the baseline by 35.85%.
Predicting the future trajectories of multiple interacting agents in a scene has become an increasingly important problem for many different applications ranging from control of autonomous vehicles and social robots to security and surveillance. This problem is compounded by the presence of social interactions between humans and their physical interactions with the scene. While the existing literature has explored some of these cues, they mainly ignored the multimodal nature of each human's future trajectory. In this paper, we present Social-BiGAT, a graph-based generative adversarial network that generates realistic, multimodal trajectory predictions by better modelling the social interactions of pedestrians in a scene. Our method is based on a graph attention network (GAT) that learns reliable feature representations that encode the social interactions between humans in the scene, and a recurrent encoder-decoder architecture that is trained adversarially to predict, based on the features, the humans' paths. We explicitly account for the multimodal nature of the prediction problem by forming a reversible transformation between each scene and its latent noise vector, as in Bicycle-GAN. We show that our framework achieves state-of-the-art performance comparing it to several baselines on existing trajectory forecasting benchmarks.
Existing methods in the Visual Storytelling field often suffer from the problem of generating general descriptions, while the image contains a lot of meaningful contents remaining unnoticed. The failure of informative story generation can be concluded to the model's incompetence of capturing enough meaningful concepts. The categories of these concepts include entities, attributes, actions, and events, which are in some cases crucial to grounded storytelling. To solve this problem, we propose a method to mine the cross-modal rules to help the model infer these informative concepts given certain visual input. We first build the multimodal transactions by concatenating the CNN activations and the word indices. Then we use the association rule mining algorithm to mine the cross-modal rules, which will be used for the concept inference. With the help of the cross-modal rules, the generated stories are more grounded and informative. Besides, our proposed method holds the advantages of interpretation, expandability, and transferability, indicating potential for wider application. Finally, we leverage these concepts in our encoder-decoder framework with the attention mechanism. We conduct several experiments on the VIsual StoryTelling~(VIST) dataset, the results of which demonstrate the effectiveness of our approach in terms of both automatic metrics and human evaluation. Additional experiments are also conducted showing that our mined cross-modal rules as additional knowledge helps the model gain better performance when trained on a small dataset.
We present improvements in automatic speech recognition (ASR) for Somali, a currently extremely under-resourced language. This forms part of a continuing United Nations (UN) effort to employ ASR-based keyword spotting systems to support humanitarian relief programmes in rural Africa. Using just 1.57 hours of annotated speech data as a seed corpus, we increase the pool of training data by applying semi-supervised training to 17.55 hours of untranscribed speech. We make use of factorised time-delay neural networks (TDNN-F) for acoustic modelling, since these have recently been shown to be effective in resource-scarce situations. Three semi-supervised training passes were performed, where the decoded output from each pass was used for acoustic model training in the subsequent pass. The automatic transcriptions from the best performing pass were used for language model augmentation. To ensure the quality of automatic transcriptions, decoder confidence is used as a threshold. The acoustic and language models obtained from the semi-supervised approach show significant improvement in terms of WER and perplexity compared to the baseline. Incorporating the automatically generated transcriptions yields a 6.55\% improvement in language model perplexity. The use of 17.55 hour of Somali acoustic data in semi-supervised training shows an improvement of 7.74\% relative over the baseline.
Head-driven phrase structure grammar (HPSG) enjoys a uniform formalism representing rich contextual syntactic and even semantic meanings. This paper makes the first attempt to formulate a simplified HPSG by integrating constituent and dependency formal representations into head-driven phrase structure. Then two parsing algorithms are respectively proposed for two converted tree representations, division span and joint span. As HPSG encodes both constituent and dependency structure information, the proposed HPSG parsers may be regarded as a sort of joint decoder for both types of structures and thus are evaluated in terms of extracted or converted constituent and dependency parsing trees. Our parser achieves new state-of-the-art performance for both parsing tasks on Penn Treebank (PTB) and Chinese Penn Treebank, verifying the effectiveness of joint learning constituent and dependency structures. In details, we report 96.33 F1 of constituent parsing and 97.20\% UAS of dependency parsing on PTB.
In this work, we investigate the effectiveness of two techniques for improving variational autoencoder (VAE) based voice conversion (VC). First, we reconsider the relationship between vocoder features extracted using the high quality vocoders adopted in conventional VC systems, and hypothesize that the spectral features are in fact F0 dependent. Such hypothesis implies that during the conversion phase, the latent codes and the converted features in VAE based VC are in fact source F0 dependent. To this end, we propose to utilize the F0 as an additional input of the decoder. The model can learn to disentangle the latent code from the F0 and thus generates converted F0 dependent converted features. Second, to better capture temporal dependencies of the spectral features and the F0 pattern, we replace the frame wise conversion structure in the original VAE based VC framework with a fully convolutional network structure. Our experiments demonstrate that the degree of disentanglement as well as the naturalness of the converted speech are indeed improved.
Reconstructing observed images from fMRI brain recordings is challenging. Unfortunately, acquiring sufficient "labeled" pairs of {Image, fMRI} (i.e., images with their corresponding fMRI responses) to span the huge space of natural images is prohibitive for many reasons. We present a novel approach which, in addition to the scarce labeled data (training pairs), allows to train fMRI-to-image reconstruction networks also on "unlabeled" data (i.e., images without fMRI recording, and fMRI recording without images). The proposed model utilizes both an Encoder network (image-to-fMRI) and a Decoder network (fMRI-to-image). Concatenating these two networks back-to-back (Encoder-Decoder & Decoder-Encoder) allows augmenting the training with both types of unlabeled data. Importantly, it allows training on the unlabeled test-fMRI data. This self-supervision adapts the reconstruction network to the new input test-data, despite its deviation from the statistics of the scarce training data.
Research on style transfer and domain translation has clearly demonstrated the ability of deep learning-based algorithms to manipulate images in terms of artistic style. More recently, several attempts have been made to extend such approaches to music (both symbolic and audio) in order to enable transforming musical style in a similar manner. In this study, we focus on symbolic music with the goal of altering the 'style' of a piece while keeping its original 'content'. As opposed to the current methods, which are inherently restricted to be unsupervised due to the lack of 'aligned' data (i.e. the same musical piece played in multiple styles), we develop the first fully supervised algorithm for this task. At the core of our approach lies a synthetic data generation scheme which allows us to produce virtually unlimited amounts of aligned data, and hence avoid the above issue. In view of this data generation scheme, we propose an encoder-decoder model for translating symbolic music accompaniments between a number of different styles. Our experiments show that our models, although trained entirely on synthetic data, are capable of producing musically meaningful accompaniments even for real (non-synthetic) MIDI recordings.
Timely, accurate and automatic detection of pavement cracks is necessary for making cost-effective decisions concerning road maintenance. Conventional crack detection algorithms focus on the design of single or multiple crack features and classifiers. However, complicated topological structures, varying degrees of damage and oil stains make the design of crack features difficult. In addition, the contextual information around a crack is not investigated extensively in the design process. Accordingly, these design features have limited discriminative adaptability and cannot fuse effectively with the classifiers. To solve these problems, this paper proposes a deep learning network for pavement crack detection. Using the Encoder-Decoder structure, crack characteristics with multiple contexts are automatically learned, and end-to-end crack detection is achieved. Specifically, we first propose the Multi-Dilation (MD) module, which can synthesize the crack features of multiple context sizes via dilated convolution with multiple rates. The crack MD features obtained in this module can describe cracks of different widths and topologies. Next, we propose the SE-Upsampling (SEU) module, which uses the Squeeze-and-Excitation learning operation to optimize the MD features. Finally, the above two modules are integrated to develop the fast crack detection network, namely, FPCNet. This network continuously optimizes the MD features step-by-step to realize fast pixel-level crack detection. Experiments are conducted on challenging public CFD datasets and G45 crack datasets involving various crack types under different shooting conditions. The distinct performance and speed improvements over all the datasets demonstrate that the proposed method outperforms other state-of-the-art crack detection methods.
In this paper, we propose a new capsule network architecture called Attention Routing CapsuleNet (AR CapsNet). We replace the dynamic routing and squash activation function of the capsule network with dynamic routing (CapsuleNet) with the attention routing and capsule activation. The attention routing is a routing between capsules through an attention module. The attention routing is a fast forward-pass while keeping spatial information. On the other hand, the intuitive interpretation of the dynamic routing is finding a centroid of the prediction capsules. Thus, the squash activation function and its variant focus on preserving a vector orientation. However, the capsule activation focuses on performing a capsule-scale activation function. We evaluate our proposed model on the MNIST, affNIST, and CIFAR-10 classification tasks. The proposed model achieves higher accuracy with fewer parameters (x0.65 in the MNIST, x0.82 in the CIFAR-10) and less training time than CapsuleNet (x0.19 in the MNIST, x0.35 in the CIFAR-10). These results validate that designing a capsule-scale operation is a key factor to implement the capsule concept. Also, our experiment shows that our proposed model is transformation equivariant as CapsuleNet. As we perturb each element of the output capsule, the decoder attached to the output capsules shows global variations. Further experiments show that the difference in the capsule features caused by applying affine transformations on an input image is significantly aligned in one direction.
Anomaly detection is a classical but worthwhile problem, and many deep learning-based anomaly detection algorithms have been proposed, which can usually achieve better detection results than traditional methods. In view of reconstruct ability of the model and the calculation of anomaly score, this paper proposes a time series anomaly detection method based on Variational AutoEncoder model(VAE) with re-Encoder and Latent Constraint network(VELC). In order to modify reconstruct ability of the model to prevent it from reconstructing abnormal samples well, we add a constraint network in the latent space of the VAE to force it generate new latent variables that are similar with that of training samples. To be able to calculate anomaly score in two feature spaces, we train a re-encoder to transform the generated data to a new latent space. For better handling the time series, we use the LSTM as the encoder and decoder part of the VAE framework. Experimental results of several benchmarks show that our method outperforms state-of-the-art anomaly detection methods.
Deep Learning for Geometric Shape Understating has organized a challenge for extracting different kinds of skeletons from the images of different objects. This competition is organized in association with CVPR 2019. There are three different tracks of this competition. The present manuscript describes the method used to train the model for the dataset provided in the first track. The first track aims to extract skeleton pixels from the shape pixels of 89 different objects. For the purpose of extracting the skeleton, a U-net model which is comprised of an encoder-decoder structure has been used. In our proposed architecture, unlike the plain decoder in the traditional Unet, we have designed the decoder in the format of HED architecture, wherein we have introduced 4 side layers and fused them to one dilation convolutional layer to connect the broken links of the skeleton. Our proposed architecture achieved the F1 score of 0.77 on test data.
Reservoir computing (RC) is a special recurrent neural network which consists of a fixed high dimensional feature mapping and trained readout weights. In this paper, we introduce a new RC structure for multiple-input, multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) symbol detection, namely windowed echo state network (WESN). The theoretical analysis shows that adding buffers in input layers can bring an enhanced short-term memory (STM) to the underlying neural network. Furthermore, a unified training framework is developed for the WESN MIMO-OFDM symbol detector using both comb and scattered pilot patterns that are compatible with the structure adopted in 3GPP LTE/LTE-Advanced systems. Complexity analysis suggests the advantages of WESN based symbol detector over state-of-the-art symbol detectors such as the linear minimum mean square error (LMMSE) detection and the sphere decoder, when the system is employed with a large number of OFDM sub-carriers. Numerical evaluations illustrate the advantage of the introduced WESN-based symbol detector and demonstrate that the improvement of STM can significantly improve symbol detection performance as well as effectively mitigate model mismatch effects compared to existing methods.
Terahertz (THz) sensing is a promising imaging technology for a wide variety of different applications. Extracting the interpretable and physically meaningful parameters for such applications, however, requires solving an inverse problem in which a model function determined by these parameters needs to be fitted to the measured data. Since the underlying optimization problem is nonconvex and very costly to solve, we propose learning the prediction of suitable parameters from the measured data directly. More precisely, we develop a model-based autoencoder in which the encoder network predicts suitable parameters and the decoder is fixed to a physically meaningful model function, such that we can train the encoding network in an unsupervised way. We illustrate numerically that the resulting network is more than 140 times faster than classical optimization techniques while making predictions with only slightly higher objective values. Using such predictions as starting points of local optimization techniques allows us to converge to better local minima about twice as fast as optimization without the network-based initialization.
Inability of the naive users to formulate appropriate queries is a fundamental problem in web search engines. Therefore, assisting users to issue more effective queries is an important way to improve users' happiness. One effective approach is query reformulation, which generates new effective queries according to the current query issued by users. Previous researches typically generate words and phrases related to the original query. Since the definition of query reformulation is quite general, it is completely difficult to develop a uniform term-based approach for this problem. This paper uses readily available data, particularly over one billion anchor phrases in Clueweb09 corpus, in order to learn an end-to-end encoder-decoder model to automatically generate effective queries. Following successful researches in the field of sequence to sequence models, we employ a character-level convolutional neural network with max-pooling at encoder and an attention-based recurrent neural network at decoder. The whole model learned in an unsupervised end-to-end manner.Experiments on TREC collections show that the reformulated queries automatically generated by the proposed solution can significantly improve the retrieval performance.
Facial makeup transfer is a widely-used technology that aims to transfer the makeup style from a reference face image to a non-makeup face. Existing literature leverage the adversarial loss so that the generated faces are of high quality and realistic as real ones, but are only able to produce fixed outputs. Inspired by recent advances in disentangled representation, in this paper we propose DMT (Disentangled Makeup Transfer), a unified generative adversarial network to achieve different scenarios of makeup transfer. Our model contains an identity encoder as well as a makeup encoder to disentangle the personal identity and the makeup style for arbitrary face images. Based on the outputs of the two encoders, a decoder is employed to reconstruct the original faces. We also apply a discriminator to distinguish real faces from fake ones. As a result, our model can not only transfer the makeup styles from one or more reference face images to a non-makeup face with controllable strength, but also produce various outputs with styles sampled from a prior distribution. Extensive experiments demonstrate that our model is superior to existing literature by generating high-quality results for different scenarios of makeup transfer.