Approximation properties of Gaussian-binary restricted Boltzmann machines and Gaussian-binary deep belief networks.

Despite the successful use of Gaussian-binary restricted Boltzmann machines (GB-RBMs) and Gaussian-binary deep belief networks (GB-DBNs), little is known about their theoretical approximation capabilities to represent distributions of continuous random variables. In this paper, we address the expressive properties of GB-RBMs and GB-DBNs, contributing theoretical insights to the optimal number of hidden variables. We first treat the GB-RBM's unnormalized log-likelihood as a sum of a special two-layer feedforward neural network and a negative quadratic term.

Fusing 2D and 3D convolutional neural networks for the segmentation of aorta and coronary arteries from CT images.

Automated segmentation of three-dimensional medical images is of great importance for the detection and quantification of certain diseases such as stenosis in the coronary arteries. Many 2D and 3D deep learning models, especially deep convolutional neural networks (CNNs), have achieved state-of-the-art segmentation performance on 3D medical images. Yet, there is a trade-off between the field of view and the utilization of inter-slice information when using pure 2D or 3D CNNs for 3D segmentation, which compromises the segmentation accuracy.

Modulating the biofunctionality of enzyme-MOF nanobiocatalyst through structure-switching aptamer for continuous degradation of BPA.

Encapsulating enzyme within MOF (enzyme-MOF) gives rise to new opportunity to improve the fragility of enzyme, but practical application of enzyme-MOF composite is far from being realized. The development of a novel enzyme-MOF composite system should simultaneously guarantee the enhanced activity and controllably complete recycling, and only in this way can we efficiently and economically utilize the enzyme-MOF composite. Herein, we addressed all these fundamental limitations of current enzyme-MOF composite by establishing aptamer-functionalized enzyme-MOF composite (HRP-ZIF-8@P1).

Refinements of Approximation Results of Conditional Restricted Boltzmann Machines.

Conditional restricted Boltzmann machines (CRBMs) are the conditional variant of restricted Boltzmann machines (RBMs), which are used to simulate conditional probability distributions. While promising for practical applications, there is a lack of theoretical studies on the approximation ability of CRBMs. In this article, by contributing analysis tools, especially designed for the conditional models, we improve the results of the representational power of CRBMs based on existing work on RBMs.

On the Representational Power of Restricted Boltzmann Machines for Symmetric Functions and Boolean Functions.

Restricted Boltzmann machines (RBMs) are used to build deep-belief networks that are widely thought to be one of the first effective deep learning neural networks. This paper studies the ability of RBMs to represent distributions over {0,1} via softplus/hardplus RBM networks. It is shown that any distribution whose density depends on the number of 1's in their input can be approximated with arbitrarily high accuracy by an RBM of size 2n+1 , which improves the result of a previous study by reducing the size from n to 2n+1 .

[Analysis of the relevance of age and toe out angle of normal adults' gait].

Due to the decline of motor ability and the impact of the diseases, abnormalities in gait is common in the elderly population, which will raise the risk of fall and cause serious injury. This study focuses on the analysis of the gait kinematics parameters of normal adults' gait, aiming to investigate the characteristics of gait parameters in different age groups and to explore the role of gait parameters in motor function assessment and clinical diagnosis. Based on the gait data gained by electronic walkway, the relationship among the toe out angles and their correlation with age and gender etc.

[Real-time Gait Training System with Embedded Functional Electrical Stimulation].

To solve the problem that mostly gait analysis is independent from the treatment, this work proposes a system that integrates the functions of gait training and assessment for foot drop treatment. The system uses a set of sensors to collect gait parameters and designes multi-mode functional electrical stimulators as actuator. Body area network technology is introduced to coordinate the data communication and execution of the sensors and stimulators, synchronize the gait analysis and foot drop treatment.