Noise-Robust Sleep Staging via Adversarial Training With an Auxiliary Model.

Deep learning (DL)-based automatic sleep staging approaches have attracted much attention recently due in part to their outstanding accuracy. At the testing stage, however, the performance of these approaches is likely to be degraded, when applied in different testing environments, because of the problem of domain shift. This is because while a pre-trained model is typically trained on noise-free electroencephalogram (EEG) signals acquired from accurate medical equipment, deployment is carried out on consumer-level devices with undesirable noise. To alleviate this challenge, in this work, we propose an efficient training approach that is robust against unseen arbitrary noise. In particular, we propose to generate the worst-case input perturbations by means of adversarial transformation in an auxiliary model, to learn a wide range of input perturbations and thereby to improve reliability. Our approach is based on two separate training models: (i) an auxiliary model to generate adversarial noise and (ii) a target network to incorporate the noise signal to enhance robustness. Furthermore, we exploit novel class-wise robustness during the training of the target network to represent different robustness patterns of each sleep stage. Our experimental results demonstrated that our approach improved sleep staging performance on healthy controls, in the presence of moderate to severe noise levels, compared with competing methods. Our approach was able to effectively train and deploy a DL model to handle different types of noise, including adversarial, Gaussian, and shot noise.

Unsupervised Domain Adaptation for Segmentation with Black-box Source Model.

Unsupervised domain adaptation (UDA) has been widely used to transfer knowledge from a labeled source domain to an unlabeled target domain to counter the difficulty of labeling in a new domain. The training of conventional solutions usually relies on the existence of both source and target domain data. However, privacy of the large-scale and well-labeled data in the source domain and trained model parameters can become the major concern of cross center/domain collaborations. In this work, to address this, we propose a practical solution to UDA for segmentation with a black-box segmentation model trained in the source domain only, rather than original source data or a white-box source model. Specifically, we resort to a knowledge distillation scheme with exponential mixup decay (EMD) to gradually learn target-specific representations. In addition, unsupervised entropy minimization is further applied to regularization of the target domain confidence. We evaluated our framework on the BraTS 2018 database, achieving performance on par with white-box source model adaptation approaches.

Unsupervised Black-Box Model Domain Adaptation for Brain Tumor Segmentation.

Unsupervised domain adaptation (UDA) is an emerging technique that enables the transfer of domain knowledge learned from a labeled source domain to unlabeled target domains, providing a way of coping with the difficulty of labeling in new domains. The majority of prior work has relied on both source and target domain data for adaptation. However, because of privacy concerns about potential leaks in sensitive information contained in patient data, it is often challenging to share the data and labels in the source domain and trained model parameters in cross-center collaborations. To address this issue, we propose a practical framework for UDA with a black-box segmentation model trained in the source domain only, without relying on source data or a white-box source model in which the network parameters are accessible. In particular, we propose a knowledge distillation scheme to gradually learn target-specific representations. Additionally, we regularize the confidence of the labels in the target domain via unsupervised entropy minimization, leading to performance gain over UDA without entropy minimization. We extensively validated our framework on a few datasets and deep learning backbones, demonstrating the potential for our framework to be applied in challenging yet realistic clinical settings.

Transferring Structured Knowledge in Unsupervised Domain Adaptation of a Sleep Staging Network.

Automatic sleep staging based on deep learning (DL) has been attracting attention for analyzing sleep quality and determining treatment effects. It is challenging to acquire long-term sleep data from numerous subjects and manually labeling them even though most DL-based models are trained using large-scale sleep data to provide state-of-the-art performance. One way to overcome this data shortage is to create a pre-trained network with an existing large-scale dataset (source domain) that is applicable to small cohorts of datasets (target domain); however, discrepancies in data distribution between the domains prevent successful refinement of this approach. In this paper, we propose an unsupervised domain adaptation method for sleep staging networks to reduce discrepancies by re-aligning the domains in the same space and producing domain-invariant features. Specifically, in addition to a classical domain discriminator, we introduce local discriminators - subject and stage - to maintain the intrinsic structure of sleep data to decrease local misalignments while using adversarial learning to play a minimax game between the feature extractor and discriminators. Moreover, we present several optimization schemes during training because the conventional adversarial learning is not effective to our training scheme. We evaluate the performance of the proposed method by examining the staging performances of a baseline network compared with direct transfer (DT) learning in various conditions. The experimental results demonstrate that the proposed domain adaptation significantly improves the performance though it needs no labeled sleep data in target domain.

Learning to Detect Cracks on Damaged Concrete Surfaces Using Two-Branched Convolutional Neural Network.

Image sensors are widely used for detecting cracks on concrete surfaces to help proactive and timely management of concrete structures. However, it is a challenging task to reliably detect cracks on damaged surfaces in the real world due to noise and undesired artifacts. In this paper, we propose an autonomous crack detection algorithm based on convolutional neural network (CNN) to solve the problem. To this aim, the proposed algorithm uses a two-branched CNN architecture, consisting of sub-networks named a crack-component-aware (CCA) network and a crack-region-aware (CRA) network. The CCA network is to learn gradient component regarding cracks, and the CRA network is to learn a region-of-interest by distinguishing critical cracks and noise such as scratches. Specifically, the two sub-networks are built on convolution-deconvolution CNN architectures, but also they are comprised of different functional components to achieve their own goals efficiently. The two sub-networks are trained in an end-to-end to jointly optimize parameters and produce the final output of localizing important cracks. Various crack image samples and learning methods are used for efficiently training the proposed network. In the experimental results, the proposed algorithm provides better performance in the crack detection than the conventional algorithms.

Machine Learning-Based Fast Angular Prediction Mode Decision Technique in Video Coding.

In this paper we propose a machine learning-based fast intra-prediction mode decision algorithm, using random forest that is an ensemble model of randomized decision trees. The random forest is used to estimate an intra-prediction mode from a prediction unit and to reduce encoding time significantly by avoiding the intensive Rate-Distortion optimization of a number of intra-prediction modes. To this aim, we develop a randomized tree model including parameterized split functions at nodes to learn directional block-based features. The feature uses only four pixels reflecting a directional property of a block, and, thus the evaluation is fast and efficient. To integrate the proposed technique into the conventional video coding standard frameworks, the intra-prediction mode derived from the proposed technique, called an inferred mode (IM), is used to shrink the pool of the candidate modes before carrying out the Rate-Distortion (R-D) optimization. The proposed technique is implemented into the High Efficiency Video Coding Test Model (HM) reference software of the state-of-the-art video coding standard and Joint Exploration Model (JEM) reference software, by integrating the random forest trained off-line into the codecs. Experimental results demonstrate that the proposed technique achieves significant encoding time reduction with only slight coding loss as compared the reference software models.

Combining random forest with multi-block local binary pattern feature selection for multiclass head pose estimation.

A new head pose estimation technique based on Random Forest (RF) and texture features for facial image analysis using a monocular camera is proposed in this paper, especially about how to efficiently combine the random forest and the features. In the proposed technique a randomized tree with useful attributes is trained to improve estimation accuracy and tolerance of occlusions and illumination. Specifically, a number of features including Multi-scale Block Local Block Pattern (MB-LBP) are extracted from an image, and random features such as the MB-LBP scale parameters, a block coordinate, and a layer of an image pyramid in the feature pool are used for training the tree. The randomized tree aims to maximize the information gain at each node while random samples traverse the nodes in the tree. To this aim, a split function considering the uniform property of the LBP feature is developed to move sample blocks to the left or the right children nodes. The trees are independently trained with random inputs, yet they are grouped to form a random forest so that the results collected from the trees are used for make the final decision. Precisely, we use a Maximum-A-Posteriori criterion in the decision. It is demonstrated with experimental results that the proposed technique provides significantly enhanced classification performance in the head pose estimation in various conditions of illumination, poses, expressions, and facial occlusions.

Intrusion Detection System Using Deep Neural Network for In-Vehicle Network Security.

A novel intrusion detection system (IDS) using a deep neural network (DNN) is proposed to enhance the security of in-vehicular network. The parameters building the DNN structure are trained with probability-based feature vectors that are extracted from the in-vehicular network packets. For a given packet, the DNN provides the probability of each class discriminating normal and attack packets, and, thus the sensor can identify any malicious attack to the vehicle. As compared to the traditional artificial neural network applied to the IDS, the proposed technique adopts recent advances in deep learning studies such as initializing the parameters through the unsupervised pre-training of deep belief networks (DBN), therefore improving the detection accuracy. It is demonstrated with experimental results that the proposed technique can provide a real-time response to the attack with a significantly improved detection ratio in controller area network (CAN) bus.

Sparse/DCT (S/DCT) two-layered representation of prediction residuals for video coding.

In this paper, we propose a cascaded sparse/DCT (S/DCT) two-layer representation of prediction residuals, and implement this idea on top of the state-of-the-art high efficiency video coding (HEVC) standard. First, a dictionary is adaptively trained to contain featured patterns of residual signals so that a high portion of energy in a structured residual can be efficiently coded via sparse coding. It is observed that the sparse representation alone is less effective in the R-D performance due to the side information overhead at higher bit rates. To overcome this problem, the DCT representation is cascaded at the second stage. It is applied to the remaining signal to improve coding efficiency. The two representations successfully complement each other. It is demonstrated by experimental results that the proposed algorithm outperforms the HEVC reference codec HM5.0 in the Common Test Condition.

Coding order decision of B frames for rate-distortion performance improvement in single-view video and multiview video coding.

The coding gain that can be achieved by improving the coding order of B frames in the H.264/AVC standard is investigated in this work. We first represent the coding order of B frames and their reference frames with a binary tree. We then formulate a recursive equation to find out the binary tree that provides a suboptimal, but very efficient, coding order. The recursive equation is efficiently solved using a dynamic programming method. Furthermore, we extend the coding order improvement technique to the case of multiview video sequences, in which the quadtree representation is used instead of the binary tree representation. Simulation results demonstrate that the proposed algorithm provides significantly better R-D performance than conventional prediction structures.