Cheng Zhang

Markov chain Monte Carlo methods have been widely used in Bayesian inference involving intractable probabilistic models. However, simple MCMC algorithms are notorious for their lack of computational efficiency in complex, high-dimensional models and poor scaling to large data sets. In recent years, many advanced MCMC methods have been proposed that utilize geometrical and statistical quantities from the model in order to explore the target distribution more effectively. The gain in the efficacy of exploration, however, often comes at a significant computational cost which hinders their application to problems with large data sets or complex likelihoods. In practice, it remains challenging to design scalable MCMC algorithms that can balance computational complexity and exploration efficacy well. To address this issue, some recent algorithms rely on stochastic gradient methods by approximating full data gradients using mini-batches of data. In contrast, this thesis focuses on accelerating the computation of MCMC samplers based on various surrogate methods via exploring the regularity of the target distribution.We start with a precomputing strategy that can be used to build efficient surrogates in relatively low-dimension parameter spaces. We then propose a random network surrogate architecture which can effectively capture the collective properties of large data sets or complex models with scalability, flexibility and efficiency. Finally, we provide a variational perspective for our random network surrogate methods and propose an approximate inference framework that combines the advantages of both variational Bayes and Markov chain Monte Carlo methods. The properties and efficiency of our proposed methods are demonstrated on a variety of synthetic and real-world data problems.

Physical literacy has received considerable attention in the field of physical education and physical activity worldwide. According to recent studies, the level of physical fitness among Chinese university students is gradually decreasing. This study aims to examine the impact of the PF level on PA and PL, as well as the relationships among PF, PA, and PL, in Chinese university students. Participants comprised 798 university students in Chongqing, China. Participants completed the tests of vital capacity, cardiorespiratory fitness, muscular strength, and flexibility, according to the National Physical Fitness Measurement Standards Manual, as well as questionnaires on PA and PL. The independent t-tests were conducted to examine sex differences in the evaluated variables, and the Pearson’s correlation between each PF test and PL attributes and PA was calculated according to sex. In addition, the independent t-tests were conducted to determine whether the PF level had an effect on PL attributes and PA at various intensities. Significant sex differences were found in the PF domains of vital capacity, muscular strength, and aerobic fitness, with higher scores in men than in women, but not in the PF domain of flexibility and total PL score. Furthermore, the PF domains of muscular strength and aerobic fitness were significantly and positively correlated with the PL attributes of confidence and physical competence in both men and women, while the PF domains of vital capacity and aerobic fitness were significantly and positively correlated with the PL attribute of motivation in men. In addition, PL was significantly and positively associated with cardiorespiratory fitness, vital capacity, muscular strength, and flexibility among participants in the fit group. These findings support advocating for increased participation in PA in university students and using PL as a tool to improve PF components.

In order to provide a consistent explanation for Aristotelian modal syllogistic, this paper reveals the reductions between the Aristotelian modal syllogism ◇I□A◇I-3 and the other valid modal syllogisms. Specifically, on the basis of formalizing Aristotelian modal syllogisms, this paper proves the validity of ◇I□A◇I-3 by means of the truth value definition of (modal) categorical propositions. Then in line with classical propositional logic and modal logic, generalized quantifier theory and set theory, this paper deduces the other 47 valid Aristotelian modal syllogisms from the modal syllogism ◇I□A◇I-3. This study shows that the reasons why these syllogisms are reducible are: 1) any of Aristotelian quantifier can be defined by the other three Aristotelian quantifiers; 2) the Aristotelian quantifiers some and no have symmetry; 3) the possible modality ◇ and necessary modality □ can be mutually defined. This formal study of Aristotelian modal syllogistic not only conforms to the needs of formalization transformation of various information in the era of artificial intelligence, but also provides a unified mathematical research paradigm for other kinds of syllogistic.

Previous studies have shown that the motion intention recognition for lower limb prosthesis mainly focused on the identification of performed gait. However, the bionic prosthesis needs to know the next movement at the beginning of a new gait, especially in complex operation environments. In this paper, an upcoming locomotion prediction scheme via multilevel classifier fusion was proposed for the complex operation. At first, two motion states, including steady state and transient state, were defined. Steady-state recognition was backtracking of a completed gait, which would be used as prior knowledge of motion prediction. In steady-state recognition, surface electromyographic and inertial sensors were fused to improve recognition accuracy; five typical locomotion modes were recognized by random forest classifier with over 97.8% accuracy. The transient state was defined as an observation period at the initial stage of upcoming movement, in which only the sEMG signal was recorded due to the limitation of sliding window length. LightGBM classifier was validated to outperform other methods in the accuracy and prediction time of transient-state recognition. Finally, a simplified HMM model based on prior knowledge and observation result was constructed to predict upcoming locomotion. The results indicated that the locomotion prediction was over 91% accuracy. The proposed scheme implements the locomotion prediction at the initial stage of each gait and provides critical information for the gait control of lower limb prosthesis.