In the real world, normal and abnormal behavior patterns vary depending on a given environment, which means that the abnormal behavior detection model should be customized. To address this issue, in this paper, we emp...
详细信息
ISBN:
(纸本)9781728116648
In the real world, normal and abnormal behavior patterns vary depending on a given environment, which means that the abnormal behavior detection model should be customized. To address this issue, in this paper, we employ OS-ELM (Online Sequential Extreme Learning Machine) and autoencoder for adaptive abnormal behavior detection. First, state-transition probability tables of a target during an initial learning period are learned as normal behaviors. Then, autoencoder-based anomaly detection is performed for the state-transition probability tables of subsequent time frames. The abnormal behavior detection model is updated by using OS-ELM algorithm every time a new probability table or behavior comes. The number of abnormal behavior detection instances is dynamically tuned to reflect the recent normal patterns or modes. Also, the table is compressed to reduce the computation cost. Evaluation results using a driving dataset of cars show that the proposed abnormal behavior detection accurately identifies normal and aggressive driving patterns with the optimal number of the abnormal behavior detection instances.
The quality of the interaction between two individuals depends upon not only exchange (i.e. understanding partner's intention and reacting to it), but also on how personalized is the interaction. In this work, we ...
详细信息
ISBN:
(纸本)9781728100647
The quality of the interaction between two individuals depends upon not only exchange (i.e. understanding partner's intention and reacting to it), but also on how personalized is the interaction. In this work, we have set out to accomplish these objectives for Human Robot Interaction. For this, we have developed a distributed and multimodal data acquisition and interaction manager architecture aiming to enable personalized Human-Robot Interactions. In the proposed approach, high-level perceptual capabilities (i.e. recognizing human activity and engagement) are performed by an autoencoder, which is a Deep Learning and Unsupervised Learning method. This autoencoder module is integrated with a facial recognition and a dialog manager (speech recognition and speech generation) to enable personalized interaction. We discuss the advantages of autoencoders over Supervised Learning methods, and how our proposed architecture can be used to increase the duration of interaction with a robot during unscripted scenarios. Experimental validations are also performed in real Human-Robot interactions using a humanoid robot.
Traditional portfolio theory divides stocks into different categories using indicators such as industry, market value, and liquidity, and then selects representative stocks according to them. In this paper, we propose...
详细信息
Traditional portfolio theory divides stocks into different categories using indicators such as industry, market value, and liquidity, and then selects representative stocks according to them. In this paper, we propose a novel portfolio learning approach based on deep learning and apply it to China's stock market. Specifically, this method is based on the similarity of deep features extracted from candlestick charts. First, we obtained whole stock information from Tushare, a professional financial data interface. These raw time series data are then plotted into candlestick charts to make an image dataset for studying the stock market. Next, the method extracts high-dimensional features from candlestick charts through an autoencoder. After that, K-means is used to cluster these high-dimensional features. Finally, we choose one stock from each category according to the Sharpe ratio and a low-risk, high-return portfolio is obtained. Extensive experiments are conducted on stocks in the Chinese stock market for evaluation. The results demonstrate that the proposed portfolio outperforms the market's leading funds and the Shanghai Stock Exchange Composite Index (SSE Index) in a number of metrics.
Background: Sepsis is a heterogenous syndrome and individualized management strategy is the key to successful treatment. Genome wide expression profiling has been utilized for identifying subclasses of sepsis, but the...
详细信息
Background: Sepsis is a heterogenous syndrome and individualized management strategy is the key to successful treatment. Genome wide expression profiling has been utilized for identifying subclasses of sepsis, but the clinical utility of these subclasses was limited because of the classification instability, and the lack of a robust class prediction model with extensive external validation. The study aimed to develop a parsimonious class model for the prediction of class membership and validate the model for its prognostic and predictive capability in external datasets. Methods: The Gene Expression Omnibus (GEO) and ArrayExpress databases were searched from inception to April 2020. Datasets containing whole blood gene expression profiling in adult sepsis patients were included. autoencoder was used to extract representative features for k-means clustering. Genetic algorithms (GA) were employed to derive a parsimonious 5-gene class prediction model. The class model was then applied to external datasets (n = 780) to evaluate its prognostic and predictive performance. Findings: A total of 12 datasets involving 1613 patients were included. Two classes were identified in the discovery cohort (n = 685). Class 1 was characterized by immunosuppression with higher mortality than class 2 (21.8% [70/321] vs. 12.1% [44/364];p < 0.01 for Chi-square test). A 5-gene class model (C14orf159, AKNA, PILRA, STOM and USP4) was developed with GA. In external validation cohorts, the 5-gene class model (AUC: 0.707;95% CI: 0.664 - 0.750) performed better in predicting mortality than sepsis response signature (SRS) endotypes (AUC: 0.610;95% CI: 0.521 - 0.700), and performed equivalently to the APACHE II score (AUC: 0.681;95% CI: 0.595 - 0.767). In the dataset E-MTAB-7581, the use of hydrocortisone was associated with increased risk of mortality (OR: 3.15 [1.13, 8.82];p = 0.029) in class 2. The effect was not statistically significant in class 1 (OR: 1.88 [0.70, 5.09];p = 0.211). Interpretati
The automatic recognition of chemical structure diagrams from the literature is an indispensable component of workflows to re-discover information about chemicals and to make it available in open-access databases. Her...
详细信息
The automatic recognition of chemical structure diagrams from the literature is an indispensable component of workflows to re-discover information about chemicals and to make it available in open-access databases. Here we report preliminary findings in our development of Deep lEarning for Chemical ImagE Recognition (DECIMER), a deep learning method based on existing show-and-tell deep neural networks, which makes very few assumptions about the structure of the underlying problem. It translates a bitmap image of a molecule, as found in publications, into a SMILES. The training state reported here does not yet rival the performance of existing traditional approaches, but we present evidence that our method will reach a comparable detection power with sufficient training time. Training success of DECIMER depends on the input data representation: DeepSMILES are superior over SMILES and we have a preliminary indication that the recently reported SELFIES outperform DeepSMILES. An extrapolation of our results towards larger training data sizes suggests that we might be able to achieve near-accurate prediction with 50 to 100 million training structures. This work is entirely based on open-source software and open data and is available to the general public for any purpose.
This paper introduces the notion of soft bits to address the rate-distortion optimization for learning-based image compression. Recent methods for such compression train an autoencoder end-to-end with an objective to ...
详细信息
ISBN:
(纸本)9781538662496
This paper introduces the notion of soft bits to address the rate-distortion optimization for learning-based image compression. Recent methods for such compression train an autoencoder end-to-end with an objective to strike a balance between distortion and rate. They are faced with the zero gradient issue due to quantization and the difficulty of estimating the rate accurately. Inspired by soft quantization, we represent quantization indices of feature maps with differentiable soft bits. This allows us to couple tightly the rate estimation with context-adaptive binary arithmetic coding. It also provides a differentiable distortion objective function. Experimental results show that our approach achieves the state-of-the-art compression performance among the learning-based schemes in terms of MS-SSIM and PSNR.
Glaucoma is the leading causes of blindness in the world. We develop a convolutional neural network for glaucoma diagnosis based on visual fields (NE), which is the gold standard to show functional damages of optic ne...
详细信息
ISBN:
(纸本)9781538636411
Glaucoma is the leading causes of blindness in the world. We develop a convolutional neural network for glaucoma diagnosis based on visual fields (NE), which is the gold standard to show functional damages of optic nerve. However, we have to deal with two major problems common in medical imaging domains. 1) It is difficult and expensive to label a large amount of data, while most modern deep learning methods require it. 2) Severe data imbalance makes the classifier easily over-fitting. In this work, for the first problem, we train an autoencoder with all the data (labeled and unlabeled) to obtain good features and introduce an active learning (AL) scheme to select and annotate a few most valuable samples from the unlabeled date set for model training. Then, we address the second problem by augmenting negative samples generated by a deep convolutional generative adversarial network (DCGAN). Experiments on our dataset (738 Samples) suggest the effectiveness of the proposed approach.
We propose the Autoencoding Binary Classifiers (ABC), a novel supervised anomaly detector based on the autoencoder (AE). There are two main approaches in anomaly detection: supervised and unsupervised. The supervised ...
详细信息
ISBN:
(纸本)9783030299118;9783030299101
We propose the Autoencoding Binary Classifiers (ABC), a novel supervised anomaly detector based on the autoencoder (AE). There are two main approaches in anomaly detection: supervised and unsupervised. The supervised approach accurately detects the known anomalies included in training data, but it cannot detect the unknown anomalies. Meanwhile, the unsupervised approach can detect both known and unknown anomalies that are located away from normal data points. However, it does not detect known anomalies as accurately as the supervised approach. Furthermore, even if we have labeled normal data points and anomalies, the unsupervised approach cannot utilize these labels. The ABC is a probabilistic binary classifier that effectively exploits the label information, where normal data points are modeled using the AE as a component. By maximizing the likelihood, the AE in the proposed ABC is trained to minimize the reconstruction error for normal data points, and to maximize it for known anomalies. Since our approach becomes able to reconstruct the normal data points accurately and fails to reconstruct the known and unknown anomalies, it can accurately discriminate both known and unknown anomalies from normal data points. Experimental results show that the ABC achieves higher detection performance than existing supervised and unsupervised methods.
Modeling the knowledge of the student is very important in online tutoring systems. It helps the instructor to design effective exercises which vary in difficulty level and design curriculum according to the observed ...
详细信息
ISBN:
(纸本)9781538680759
Modeling the knowledge of the student is very important in online tutoring systems. It helps the instructor to design effective exercises which vary in difficulty level and design curriculum according to the observed knowledge state of students. But modeling the student knowledge has got its own challenges. The model which uses RNNs (Recurrent Neural Networks) which has proved to be effective in modeling the student knowledge but has got its own limitations. We found that there were some features that were very important in predicting the performance of student but were ignored in the original DKT (Deep knowledge Tracing) model [1]. In this paper, we attempt to consider three more features such as No. of hints accessed, Student First Response Time and Number of Attempts when compared to DKT and use the same model as in DKT where the only change is made in the input vector we provide to the model. We also use certain techniques like dimensionality reduction by using autoencoder which helps in reducing the dimensionality as we are taking extra features into consideration.
In many scenes, the frontal face image is the only criterion for judging the identity of a person. However, it is difficult to collect a standard frontal image in an uncontrolled environment. To get a clear frontal im...
详细信息
ISBN:
(纸本)9781538662434
In many scenes, the frontal face image is the only criterion for judging the identity of a person. However, it is difficult to collect a standard frontal image in an uncontrolled environment. To get a clear frontal image from a large variety of profile images, there are many studies on face frontalization. Some researches need three-dimension face data or prior pose information while others do not take into account the effect of pose information. And there are restrictions on the number of poses of input face images. Because of the ill-consideration of pose information, the authenticity of generated frontal face images is not high when we input multi-poses profile images. To resolve this problem, this paper proposes a Pose-weighted Generative Adversarial Network (PWGAN), which adds a pre-trained pose certification module to learn face pose information. For the single input image, PWGAN combines fusion features with pose features. And for multiple input images, PWGAN uses pose information to dynamic distribute weights when fusing feature maps. PWGAN makes full use of pose information to make the generation network learn more about facial features and get better-generating effect. Through contrastive experiments, this paper proves that PWGAN has a better effect on multi-poses face frontalization than the above methods.
暂无评论