Catalogue Search | MBRL
Are you sure you want to remove the book from the shelf?
{{itemTitle}}
5,377
result(s) for
"Fu, Yun"
Sort by:
Human Action Recognition and Prediction: A Survey
Derived from rapid advances in computer vision and machine learning, video analysis tasks have been moving from inferring the present state to predicting the future state. Vision-based action recognition and prediction from videos are such tasks, where action recognition is to infer human actions (present state) based upon complete action executions, and action prediction to predict human actions (future state) based upon incomplete action executions. These two tasks have become particularly prevalent topics recently because of their explosively emerging real-world applications, such as visual surveillance, autonomous driving vehicle, entertainment, and video retrieval, etc. Many attempts have been devoted in the last a few decades in order to build a robust and effective framework for action recognition and prediction. In this paper, we survey the complete state-of-the-art techniques in action recognition and prediction. Existing models, popular algorithms, technical difficulties, popular action databases, evaluation protocols, and promising future directions are also provided with systematic discussions.
Journal Article
Stairway Plot 2: demographic history inference with folded SNP frequency spectra
Inferring the demographic histories of populations has wide applications in population, ecological, and conservation genomics. We present Stairway Plot 2, a cross-platform program package for this task using SNP frequency spectra. It is based on a nonparametric method with the capability of handling folded SNP frequency spectra (that is, when the ancestral alleles of the SNPs are unknown) of thousands of samples produced with genotyping-by-sequencing technologies; therefore, it is particularly suitable for nonmodel organisms.
Journal Article
Exploring population size changes using SNP frequency spectra
Xiaoming Liu and Yun-Xin Fu present a model-flexible method for inferring changes in population size over time on the basis of the composite likelihood of SNP frequencies. They apply the method to 1000 Genomes Project data to infer changes in human population size on the timescale of 10,000 to 200,000 years ago.
Inferring demographic history is an important task in population genetics. Many existing inference methods are based on predefined simplified population models, which are more suitable for hypothesis testing than exploratory analysis. We developed a novel model-flexible method called stairway plot, which infers changes in population size over time using SNP frequency spectra. This method is applicable for whole-genome sequences of hundreds of individuals. Using extensive simulation, we demonstrate the usefulness of the method for inferring demographic history, especially recent changes in population size. We apply the method to the whole-genome sequence data of 9 populations from the 1000 Genomes Project and show a pattern of fluctuations in human populations from 10,000 to 200,000 years ago.
Journal Article
Self-directed online machine learning for topology optimization
Topology optimization by optimally distributing materials in a given domain requires non-gradient optimizers to solve highly complicated problems. However, with hundreds of design variables or more involved, solving such problems would require millions of Finite Element Method (FEM) calculations whose computational cost is huge and impractical. Here we report Self-directed Online Learning Optimization (SOLO) which integrates Deep Neural Network (DNN) with FEM calculations. A DNN learns and substitutes the objective as a function of design variables. A small number of training data is generated dynamically based on the DNN’s prediction of the optimum. The DNN adapts to the new training data and gives better prediction in the region of interest until convergence. The optimum predicted by the DNN is proved to converge to the true global optimum through iterations. Our algorithm was tested by four types of problems including compliance minimization, fluid-structure optimization, heat transfer enhancement and truss optimization. It reduced the computational time by 2 ~ 5 orders of magnitude compared with directly using heuristic methods, and outperformed all state-of-the-art algorithms tested in our experiments. This approach enables solving large multi-dimensional optimization problems.
Topology optimization, relevant for materials design and engineering, requires solving of challenging high-dimensional problems. The authors introduce a self-directed online learning approach, as embedding of deep learning in optimization methods, that accelerates the training and optimization processes.
Journal Article
Human mesenchymal stem cell treatment of premature ovarian failure: new challenges and opportunities
Premature ovarian failure (POF) is one of the common disorders found in women leading to 1% female infertility. Clinical features of POF are hypoestrogenism or estrogen deficiency, increased gonadotropin level, and, most importantly, amenorrhea. With the development of regenerative medicine, human mesenchymal stem cell (hMSC) therapy brings new prospects for POF. This study aimed to describe the types of MSCs currently available for POF therapy, their biological characteristics, and their mechanism of action. It reviewed the latest findings on POF to provide the theoretical basis for further investigation and clinical therapy.
Journal Article
Generalized Transfer Subspace Learning Through Low-Rank Constraint
It is expensive to obtain labeled real-world visual data for use in training of supervised algorithms. Therefore, it is valuable to leverage existing databases of labeled data. However, the data in the source databases is often obtained under conditions that differ from those in the new task. Transfer learning provides techniques for transferring learned knowledge from a
source
domain to a
target
domain by finding a mapping between them. In this paper, we discuss a method for projecting both source and target data to a generalized subspace where each target sample can be represented by some combination of source samples. By employing a low-rank constraint during this transfer, the structure of source and target domains are preserved. This approach has three benefits. First, good alignment between the domains is ensured through the use of only relevant data in some subspace of the source domain in reconstructing the data in the target domain. Second, the discriminative power of the source domain is naturally passed on to the target domain. Third, noisy information will be filtered out during knowledge transfer. Extensive experiments on synthetic data, and important computer vision problems such as face recognition application and visual domain adaptation for object recognition demonstrate the superiority of the proposed approach over the existing, well-established methods.
Journal Article
On Non-Penalization SEMDOT Using Discrete Variable Sensitivities
This work proposes a non-penalization Smooth-Edged Material Distribution for Optimizing Topology (SEMDOT) algorithm, which is a typical elemental volume fraction-based topology optimization method, by adopting discrete variable sensitivities for solid, void, and assumed boundary elements instead of the continuous variable sensitivities used in the penalization one. In the proposed non-penalized SEMDOT algorithm, the material penalization scheme is eliminated. The efficiency, effectiveness, and general applicability of the proposed non-penalized algorithm are demonstrated in three case studies containing compliance minimization, compliant mechanism design, and heat conduction problems, as well as thorough comparisons with the penalized algorithm. In addition, the length scale control approach is used to solve the discontinuous boundary issue observed in thin and long structural features. The numerical results show that the convergency of the newly proposed non-penalization algorithm is stronger than the penalization algorithm, and improved results can be obtained by the non-penalized algorithm.
Journal Article
PyCIL: a Python toolbox for class-incremental learning
Conclusion
We have presented PyCIL, a classincremental learning toolbox written in Python. It contains implementations of a number of founding studies of CIL, but also provides current state-of-the-art algorithms that can be used to conduct novel fundamental research. Code consistency makes it an easy tool for research purposes, teaching, and industrial applications.
Journal Article