Search

More people use the internet for medical information, especially YouTube. Nevertheless, no study has been conducted to analyze the quality of YouTube videos about tinnitus in Korea. This study aims to review the contents and quality of YouTube videos on tinnitus. The top 100 Korean YouTube videos on tinnitus were reviewed by a tinnitus expert. This study assessed video details: title, creator, length, and popularity indicators - subscribers, views, and likes. The contents of the video clips were analyzed to determine the relevant information they contained, and their understandability and actionability scores were also assessed. Out of 100 tinnitus videos, 27 were created by otolaryngologists, 25 by Oriental medicine doctors, 25 by other medical professionals, and 3 by lay persons. Sensorineural tinnitus was frequently dealt, and hearing loss, stress, and noise were introduced as main causes of tinnitus. Otolaryngologists' videos covered verified treatments, but others suggested unproven therapies including herbal medicine or acupressure. Otolaryngologists' videos had higher understandability than Oriental doctors (p = 0.003) and other medical professionals (p

Graph neural networks (GNNs) have demonstrated superior performance in various tasks on graphs. However, existing GNNs often suffer from weak-generalization due to sparsely labeled datasets. Here we propose a novel framework that learns to augment the input features using topological information and automatically controls the strength of augmentation. Our framework learns the augmentor to minimize GNNs’ loss on unseen labeled data while maximizing the consistency of GNNs’ predictions on unlabeled data. This can be formulated as a meta-learning problem and our framework alternately optimizes the augmentor and GNNs for a target task. Our extensive experiments demonstrate that the proposed framework is applicable to any GNNs and significantly improves the performance of graph neural networks on node classification. In particular, our method provides 5.78% improvement with Graph convolutional network (GCN) on average across five benchmark datasets.

Filter pruning is prevalent for pruning-based model compression. Most filter pruning methods have two main issues: 1) the pruned network capability depends on that of source pretrained models, and 2) they do not consider that filter weights follow a normal distribution. To address these issues, we propose a new pruning method employing both weight re-initialization and latent space clustering. For latent space clustering, we define filters and their feature maps as vertices and edges to be a graph, transformed into a latent space by graph convolution, alleviating to prune zero-near weight filters only. In addition, a part of filters is re-initialized with a constraint for enhancing filter diversity, and thus the pruned model is less dependent on the source network. This approach provides more robust accuracy even when pruned from the pretrained model with low accuracy. Extensive experimental results show our method decreases 56.6% and 84.6% of FLOPs and parameters of VGG16 with negligible loss of accuracy on CIFAR100, which is the state-of-the art performance. Furthermore, our method presents outperforming or comparable pruning results against state-of-the-art models on multiple datasets.

Learning under a continuously changing data distribution with incorrect labels is a desirable real-world problem yet challenging. A large body of continual learning (CL) methods, however, assumes data streams with clean labels, and online learning scenarios under noisy data streams are yet underexplored. We consider a more practical CL task setup of an online learning from blurry data stream with corrupted labels, where existing CL methods struggle. To address the task, we first argue the importance of both diversity and purity of examples in the episodic memory of continual learning models. To balance diversity and purity in the episodic memory, we propose a novel strategy to manage and use the memory by a unified approach of label noise aware diverse sampling and robust learning with semi-supervised learning. Our empirical validations on four real-world or synthetic noise datasets (CIFAR10 and 100, mini-WebVision, and Food-101N) exhibit that our method significantly outperforms prior arts in this realistic and challenging continual learning scenario. Code and data splits are available in https://github.com/clovaai/puridiver., Accepted paper at CVPR 2022

The retriever-reader pipeline has shown promising performance in open-domain QA but suffers from a very slow inference speed. Recently proposed question retrieval models tackle this problem by indexing question-answer pairs and searching for similar questions. These models have shown a significant increase in inference speed, but at the cost of lower QA performance compared to the retriever-reader models. This paper proposes a two-step question retrieval model, SQuID (Sequential Question-Indexed Dense retrieval) and distant supervision for training. SQuID uses two bi-encoders for question retrieval. The first-step retriever selects top-k similar questions, and the second-step retriever finds the most similar question from the top-k questions. We evaluate the performance and the computational efficiency of SQuID. The results show that SQuID significantly increases the performance of existing question retrieval models with a negligible loss on inference speed., Comment: ACL2022-Findings

The problem of generating a set of diverse paraphrase sentences while (1) not compromising the original meaning of the original sentence, and (2) imposing diversity in various semantic aspects, such as a lexical or syntactic structure, is examined. Existing work on paraphrase generation has focused more on the former, and the latter was trained as a fixed style transfer, such as transferring from positive to negative sentiments, even at the cost of losing semantics. In this work, we consider style transfer as a means of imposing diversity, with a paraphrasing correctness constraint that the target sentence must remain a paraphrase of the original sentence. However, our goal is to maximize the diversity for a set of k generated paraphrases, denoted as the diversified paraphrase (DP) problem. Our key contribution is deciding the style guidance at generation towards the direction of increasing the diversity of output with respect to those generated previously. As pre-materializing training data for all style decisions is impractical, we train with biased data, but with debiasing guidance. Compared to state-of-the-art methods, our proposed model can generate more diverse and yet semantically consistent paraphrase sentences. That is, our model, trained with the MSCOCO dataset, achieves the highest embedding scores, .94/.95/.86, similar to state-of-the-art results, but with a lower mBLEU score (more diverse) by 8.73%.

Continual learning is a realistic learning scenario for AI models. Prevalent scenario of continual learning, however, assumes disjoint sets of classes as tasks and is less realistic rather artificial. Instead, we focus on 'blurry' task boundary; where tasks shares classes and is more realistic and practical. To address such task, we argue the importance of diversity of samples in an episodic memory. To enhance the sample diversity in the memory, we propose a novel memory management strategy based on per-sample classification uncertainty and data augmentation, named Rainbow Memory (RM). With extensive empirical validations on MNIST, CIFAR10, CIFAR100, and ImageNet datasets, we show that the proposed method significantly improves the accuracy in blurry continual learning setups, outperforming state of the arts by large margins despite its simplicity. Code and data splits will be available in https://github.com/clovaai/rainbow-memory., Accepted paper at CVPR 2021

In multi-hop QA, answering complex questions entails iterative document retrieval for finding the missing entity of the question. The main steps of this process are sub-question detection, document retrieval for the sub-question, and generation of a new query for the final document retrieval. However, building a dataset that contains complex questions with sub-questions and their corresponding documents requires costly human annotation. To address the issue, we propose a new method for weakly supervised multi-hop retriever pre-training without human efforts. Our method includes 1) a pre-training task for generating vector representations of complex questions, 2) a scalable data generation method that produces the nested structure of question and sub-question as weak supervision for pre-training, and 3) a pre-training model structure based on dense encoders. We conduct experiments to compare the performance of our pre-trained retriever with several state-of-the-art models on end-to-end multi-hop QA as well as document retrieval. The experimental results show that our pre-trained retriever is effective and also robust on limited data and computational resources., Comment: ACL-Findings 2021

Assessing advertisements, specifically on the basis of user preferences and ad quality, is crucial to the marketing industry. Although recent studies have attempted to use deep neural networks for this purpose, these studies have not utilized image-related auxiliary attributes, which include embedded text frequently found in ad images. We, therefore, investigated the influence of these attributes on ad image preferences. First, we analyzed large-scale real-world ad log data and, based on our findings, proposed a novel multi-step modality fusion network (M2FN) that determines advertising images likely to appeal to user preferences. Our method utilizes auxiliary attributes through multiple steps in the network, which include conditional batch normalization-based low-level fusion and attention-based high-level fusion. We verified M2FN on the AVA dataset, which is widely used for aesthetic image assessment, and then demonstrated that M2FN can achieve state-of-the-art performance in preference prediction using a real-world ad dataset with rich auxiliary attributes., Comment: published in Applied Soft Computing

Analyzing the associations between genotypic changes and phenotypic traits on a genome-wide scale can contribute to understanding the functional roles of distinct genetic variations during breed development. We performed a whole-genome analysis of Angus and Jersey cattle breeds using conditional mutual information, which is an information-theoretic method estimating the conditional independency among multiple factor variables. The proposed conditional mutual information-based approach allows breed-discriminative genetic variations to be explicitly identified from tens of millions of SNP (single nucleotide polymorphism) positions on a genome-wide scale while minimizing the usage of prior knowledge. Using this data-driven approach, we identified biologically relevant functional genes, including breed-specific variants for cattle traits such as beef and dairy production. The identified lipid-related genes were shown to be significantly associated with lipid and triglyceride metabolism, fat cell differentiation, and muscle development. In addition, we confirmed that milk-related genes are involved in mammary gland development, lactation, and mastitis-associated processes. Our results provide the distinct properties of Angus and Jersey cattle at a genome-wide level. Moreover, this study offers important insights into discovering unrevealed genetic variants for breed-specific traits and the identification of genetic signatures of diverse cattle breeds with respect to target breed-specific properties.

A good image-to-image translation model should learn a mapping between different visual domains while satisfying the following properties: 1) diversity of generated images and 2) scalability over multiple domains. Existing methods address either of the issues, having limited diversity or multiple models for all domains. We propose StarGAN v2, a single framework that tackles both and shows significantly improved results over the baselines. Experiments on CelebA-HQ and a new animal faces dataset (AFHQ) validate our superiority in terms of visual quality, diversity, and scalability. To better assess image-to-image translation models, we release AFHQ, high-quality animal faces with large inter- and intra-domain differences. The code, pretrained models, and dataset can be found at https://github.com/clovaai/stargan-v2., Accepted to CVPR 2020

Recent advances in pre-trained language models have significantly improved neural response generation. However, existing methods usually view the dialogue context as a linear sequence of tokens and learn to generate the next word through token-level self-attention. Such token-level encoding hinders the exploration of discourse-level coherence among utterances. This paper presents DialogBERT, a novel conversational response generation model that enhances previous PLM-based dialogue models. DialogBERT employs a hierarchical Transformer architecture. To efficiently capture the discourse-level coherence among utterances, we propose two training objectives, including masked utterance regression and distributed utterance order ranking in analogy to the original BERT training. Experiments on three multi-turn conversation datasets show that our approach remarkably outperforms the baselines, such as BART and DialoGPT, in terms of quantitative evaluation. The human evaluation suggests that DialogBERT generates more coherent, informative, and human-like responses than the baselines with significant margins., Comment: Published as a conference paper at AAAI 2021

Language model pre-training has shown promising results in various downstream tasks. In this context, we introduce a cross-modal pre-trained language model, called Speech-Text BERT (ST-BERT), to tackle end-to-end spoken language understanding (E2E SLU) tasks. Taking phoneme posterior and subword-level text as an input, ST-BERT learns a contextualized cross-modal alignment via our two proposed pre-training tasks: Cross-modal Masked Language Modeling (CM-MLM) and Cross-modal Conditioned Language Modeling (CM-CLM). Experimental results on three benchmarks present that our approach is effective for various SLU datasets and shows a surprisingly marginal performance degradation even when 1% of the training data are available. Also, our method shows further SLU performance gain via domain-adaptive pre-training with domain-specific speech-text pair data., Comment: ICASSP 2021; 5 pages

Extractive QA models have shown very promising performance in predicting the correct answer to a question for a given passage. However, they sometimes result in predicting the correct answer text but in a context irrelevant to the given question. This discrepancy becomes especially important as the number of occurrences of the answer text in a passage increases. To resolve this issue, we propose \textbf{BLANC} (\textbf{BL}ock \textbf{A}ttentio\textbf{N} for \textbf{C}ontext prediction) based on two main ideas: context prediction as an auxiliary task in multi-task learning manner, and a block attention method that learns the context prediction task. With experiments on reading comprehension, we show that BLANC outperforms the state-of-the-art QA models, and the performance gap increases as the number of answer text occurrences increases. We also conduct an experiment of training the models using SQuAD and predicting the supporting facts on HotpotQA and show that BLANC outperforms all baseline models in this zero-shot setting., Comment: EMNLP 2020

Automatic speech recognition (ASR) via call is essential for various applications, including AI for contact center (AICC) services. Despite the advancement of ASR, however, most publicly available call-based speech corpora such as Switchboard are old-fashioned. Also, most existing call corpora are in English and mainly focus on open domain dialog or general scenarios such as audiobooks. Here we introduce a new large-scale Korean call-based speech corpus under a goal-oriented dialog scenario from more than 11,000 people, i.e., ClovaCall corpus. ClovaCall includes approximately 60,000 pairs of a short sentence and its corresponding spoken utterance in a restaurant reservation domain. We validate the effectiveness of our dataset with intensive experiments using two standard ASR models. Furthermore, we release our ClovaCall dataset and baseline source codes to be available via https://github.com/ClovaAI/ClovaCall., Comment: 5 pages, 2 figures, 4 tables, The first two authors equally contributed to this work

A 7-MHz integrated peak-current-mode buck regulator with a charge-recycling technique and a fast response current-sensing circuit is proposed. The proposed charge-recycling technique contributes to the improvement of the light load efficiency by up to 1.7% for the proposed buck regulator. Moreover, the proposed buck regulator achieves a fast response time of less than 10 ns with a fast response current-sensing circuit. Applying the proposed charge-recycling technique and the current-sensing circuit, the peak-current-mode buck regulator for the mobile application is realized with an operating frequency of 7 MHz, an output voltage of 0.8 V, a maximum load current of 500 mA, and a peak efficiency of over 77%.

Recent style transfer models have provided promising artistic results. However, given a photograph as a reference style, existing methods are limited by spatial distortions or unrealistic artifacts, which should not happen in real photographs. We introduce a theoretically sound correction to the network architecture that remarkably enhances photorealism and faithfully transfers the style. The key ingredient of our method is wavelet transforms that naturally fits in deep networks. We propose a wavelet corrected transfer based on whitening and coloring transforms (WCT$^2$) that allows features to preserve their structural information and statistical properties of VGG feature space during stylization. This is the first and the only end-to-end model that can stylize a $1024\times1024$ resolution image in 4.7 seconds, giving a pleasing and photorealistic quality without any post-processing. Last but not least, our model provides a stable video stylization without temporal constraints. Our code, generated images, and pre-trained models are all available at https://github.com/ClovaAI/WCT2., Comment: Accepted to ICCV 2019, Code and data: https://github.com/ClovaAI/WCT2

Generating SQL codes from natural language questions (NL2SQL) is an emerging research area. Existing studies have mainly focused on clear scenarios where specified information is fully given to generate a SQL query. However, in developer forums such as Stack Overflow, questions cover more diverse tasks including table manipulation or performance issues, where a table is not specified. The SQL query posted in Stack Overflow, Pseudo-SQL (pSQL), does not usually contain table schemas and is not necessarily executable, is sufficient to guide developers. Here we describe a new NL2pSQL task to generate pSQL codes from natural language questions on under-specified database issues, NL2pSQL. In addition, we define two new metrics suitable for the proposed NL2pSQL task, Canonical-BLEU and SQL-BLEU, instead of the conventional BLEU. With a baseline model using sequence-to-sequence architecture integrated by denoising autoencoder, we confirm the validity of our task. Experiments show that the proposed NL2pSQL approach yields well-formed queries (up to 43% more than a standard Seq2Seq model). Our code and datasets will be publicly released.

Recent studies have shown remarkable success in image-to-image translation for two domains. However, existing approaches have limited scalability and robustness in handling more than two domains, since different models should be built independently for every pair of image domains. To address this limitation, we propose StarGAN, a novel and scalable approach that can perform image-to-image translations for multiple domains using only a single model. Such a unified model architecture of StarGAN allows simultaneous training of multiple datasets with different domains within a single network. This leads to StarGAN's superior quality of translated images compared to existing models as well as the novel capability of flexibly translating an input image to any desired target domain. We empirically demonstrate the effectiveness of our approach on a facial attribute transfer and a facial expression synthesis tasks., Accepted to CVPR 2018 (Oral)

Place awareness is an essential for location-based services that are widely provided to smartphone users. However, traditional GPS-based methods are only valid outdoors where the GPS signal is strong and also require symbolic place information of the physical location. In this paper, environmental sounds and images are used to recognize important aspects of each place. The proposed method extracts feature vectors from visual, auditory and location data recorded by a smartphone with built-in camera, microphone and GPS sensors modules. The heterogeneous feature vectors were then learned by an ensemble learning method that learns each group of feature vectors for each classifier respectively and votes to produce the highest weighted result. The proposed method is evaluated for place recognition using a data group of 3000 samples in six places and the experimental results show a remarkably improved recognition accuracy when using all kinds of sensory data comparing to results using data from a single sensor or audio-visual integrated data only.

Higher-order representation is suitable for the complicated relationships among many factors. However, existing higher-order classification models have difficulties in learning from high-dimensional data due to their large combinatorial hypothesis spaces. The interpretability of models is also significant for causality analysis. Here we propose a Bayesian evolutionary method to learn a higher-order graphical model for high-dimensional data, called Bayesian evolutionary hypernetwork (BEHN). Our method represents the combinatorial feature space using a generalized graph, hypernetwork. A hypernetwork contains a large population of hyperedges encoding higher-order relationships among feature variables, and is optimized by an evolutionary algorithm formulated as sequential Bayesian sampling. This Bayesian evolutionary approach allows for probabilistic search through the higher-order feature space while satisfying soft constraints defined by the priors. We show that two information-theoretic and complexity-related priors are effective to balance model accuracy and parsimony. Also, BEHN provides interpretable representations to investigate feature interactions. Using two benchmarking and three real-world datasets we demonstrate that BEHN outperforms baseline classification models while tackling large-scale data of dimensionality up to O( 1 0 4 ). We also analyze the stability and the scalability of the proposed method with respect to accuracy, computational cost, and the interpretability of the model structures.

We propose Dual Attention Networks (DANs) which jointly leverage visual and textual attention mechanisms to capture fine-grained interplay between vision and language. DANs attend to specific regions in images and words in text through multiple steps and gather essential information from both modalities. Based on this framework, we introduce two types of DANs for multimodal reasoning and matching, respectively. The reasoning model allows visual and textual attentions to steer each other during collaborative inference, which is useful for tasks such as Visual Question Answering (VQA). In addition, the matching model exploits the two attention mechanisms to estimate the similarity between images and sentences by focusing on their shared semantics. Our extensive experiments validate the effectiveness of DANs in combining vision and language, achieving the state-of-the-art performance on public benchmarks for VQA and image-text matching.

In this study, direct reduction of using liquid calcium was investigated. The influence of molar ratio of Ca and , reaction time and temperature on the reduction behavior of was studied. Experiments were conducted in a closed stainless steel chamber under Ar atmosphere during 5 to 60 minutes. Most of the was reduced to Zr in 5 minutes at 1223 K and 3 Ca/ molar ratio. The minimum oxygen content in reduced metal Zr was obtained about 0.66 wt% at 1373 K after 30 minutes and 4 Ca/ molar ratio. The morphology of zirconium powder obtained was highly affected by the reaction temperature and reaction time.

Learning mutually-grounded vision-language knowledge is a foundational task for cognitive systems and human-level artificial intelligence. Most of knowledge-learning techniques are focused on single modal representations in a static environment with a fixed set of data. Here, we explore an ecologically more-plausible setting by using a stream of cartoon videos to build vision-language concept hierarchies continuously. This approach is motivated by the literature on cognitive development in early childhood. We present the model of deep concept hierarchy (DCH) that enables the progressive abstraction of concept knowledge in multiple levels. We develop a stochastic method for graph construction, i.e. a graph Monte Carlo algorithm, to search efficiently the huge compositional space of the vision-language concepts. The concept hierarchies are built incrementally and can handle concept drift, allowing for being deployed in lifelong learning environments. Using a series of approximately 200 episodes of educational cartoon videos we demonstrate the emergence and evolution of the concept hierarchies as the video stories unfold. We also present the application of the deep concept hierarchies for context-dependent translation between vision and language, i.e. the transcription of a visual scene into text and the generation of visual imagery from text.

This paper proposes a new use of the pseudospectral method to the eigenvalue analysis of the power transformer winding. The analysis method is based on the Chebyshev polynomials and collocated at the Gauss-Labatto collocation points. To verify the method, a helical-type winding of a power transformer is analyzed and numerical examples are provided for various boundary conditions. The computed results are in good agreement with finite-element method solutions. It shows that the method is straightforward and it has novel advantages in its conceptual simplicity

Predicting the clinical outcomes of cancer patients is a challenging task in biomedicine. A personalized and refined therapy based on predicting prognostic outcomes of cancer patients has been actively sought in the past decade. Accurate prognostic prediction requires higher-order representations of complex dependencies among genetic factors. However, identifying the co-regulatory roles and functional effects of genetic interactions on cancer prognosis is hindered by the complexity of the interactions. Here we propose a prognostic prediction model based on evolutionary learning that identifies higher-order prognostic biomarkers of cancer clinical outcomes. The proposed model represents the interactions of prognostic genes as a combinatorial space. It adopts a flexible hypergraph structure composed of a large population of hyperedges that encode higher-order relationships among many genetic factors. The hyperedge population is optimized by an evolutionary learning method based on sequential Bayesian sampling. The proposed learning approach effectively balances performance and parsimony of the model using information-theoretic dependency and complexity-theoretic regularization priors. Using MAQC-II project data, we demonstrate that our model can handle high-dimensional data more effectively than state-of-the-art classification models. We also identify potential gene interactions characterizing prognosis and recurrence risk in cancer.

Background Dysregulation of genetic factors such as microRNAs (miRNAs) and mRNAs has been widely shown to be associated with cancer progression and development. In particular, miRNAs and mRNAs cooperate to affect biological processes, including tumorigenesis. The complexity of miRNA-mRNA interactions presents a major barrier to identifying their co-regulatory roles and functional effects. Thus, by computationally modeling these complex relationships, it may be possible to infer the gene interaction networks underlying complicated biological processes. Results We propose a data-driven, hypergraph structural method for constructing higher-order miRNA-mRNA interaction networks from cancer genomic profiles. The proposed model explicitly characterizes higher-order relationships among genetic factors, from which cooperative gene activities in biological processes may be identified. The proposed model is learned by iteration of structure and parameter learning. The structure learning efficiently constructs a hypergraph structure by generating putative hyperedges representing complex miRNA-mRNA modules. It adopts an evolutionary method based on information-theoretic criteria. In the parameter learning phase, the constructed hypergraph is refined by updating the hyperedge weights using the gradient descent method. From the model, we produce biologically relevant higher-order interaction networks showing the properties of primary and metastatic prostate cancer, as candidates of potential miRNA-mRNA regulatory circuits. Conclusions Our approach focuses on potential cancer-specific interactions reflecting higher-order relationships between miRNAs and mRNAs from expression profiles. The constructed miRNA-mRNA interaction networks show oncogenic or tumor suppression characteristics, which are known to be directly associated with prostate cancer progression. Therefore, the hypergraph-based model can assist hypothesis formulation for the molecular pathogenesis of cancer.

Conventional methods for multimodal data retrieval use text-tag based or cross-modal approaches such as tag-image co-occurrence and canonical correlation analysis. Since there are differences of granularity in text and image features, however, approaches based on lower-order relationship between modalities may have limitations. Here, we propose a novel text and image keyword generation method by cross-modal associative learning and inference with multimodal queries. We use a modified hypernetwork model, i.e. layered hypernetworks (LHNs) which consists of the first (lower) layer and the second (upper) layer which has more than two modality-dependent hypernetworks and one modality-integrating hypernetwork, respectively. LHNs learn higher-order associative relationships between text and image modalities by training on an example set. After training, LHNs are used to extend multimodal queries by generating text and image keywords via cross-modal inference, i.e. text-to-image and image-to-text. The LHNs are evaluated on Korean magazine articles with images on women fashions and life-style. Experimental results show that the proposed method generates vision-language cross-modal keywords with high accuracy. The results also show that multimodal queries improve the accuracy of keyword generation compared with uni-modal ones.