Yusuke Fujimoto

Informative input design for Kernel-Based system identification

This paper discusses the design of input sequence for Kernel-Based system identification. From the Bayesian point of view, the kernel reflects a priori information about the target system, which implies that the information obtained from I/O data differs over kernels. This paper focuses on finding an input sequence which maximizes the information obtained through an observation according to the kernel which is given in advance. As an appropriate measure of such information, the mutual information is adopted. For the given kernel, a concrete procedure is proposed to find the input sequence maximizing the mutual information subject to the input energy constraints. Numerical examples are given to illustrate the effectiveness of the proposed input design. Furthermore, it is shown analytically that the impulse input is optimal for a special class of kernels.

Metabolic changes in adipose tissues in response to β3-adrenergic receptor activation in mice: FUJIMOTO et al.

Brown and beige adipocytes dissipate energy as heat. Thus, the activation of brown adipocytes and the emergence of beige adipocytes in white adipose tissue (WAT) are suggested to be useful for preventing and treating obesity. Although β3‐adrenergic receptor activation is known to stimulate lipolysis and activation of brown and beige adipocytes, fat depot–dependent changes in metabolite concentrations are not fully elucidated. The current study examined the effect of treatment with CL‐316,243, a β3‐adrenergic receptor agonist, on the relative abundance of metabolites in interscapular brown adipose tissue (iBAT), inguinal WAT (ingWAT), and epididymal WAT (epiWAT). Intraperitoneal injection of CL‐316,243 (1 mg/kg) for 3 consecutive days increased the relative abundance of several glycolysis‐related metabolites in all examined fat depots. The cellular concentrations of metabolites involved in the citric acid cycle and of free amino acids were also increased in epiWAT by CL‐316,243. CL‐316,243 increased the expression levels of several enzymes and transporters related to glucose metabolism and amino acid catabolism in ingWAT and iBAT but not in epiWAT. CL‐316,243 also induced the emergence of more beige adipocytes in ingWAT than in epiWAT. Furthermore, adipocytes surrounded by macrophages were detected in the epiWAT of mice given CL‐316,243. The current study reveals the fat depot–dependent modulation of cellular metabolites in CL‐316,243‐treated mice, presumably resulting from differential regulation of cell metabolism in different cell populations.

On a relationship between integral compensation and stochastic gradient descent

Disturbance rejection is a fundamental problem in control engineering, and there are many methods to achieve a good disturbance rejection. One of the standard methods for the disturbance rejection is to employ an integral compensator. This compensator integrates the error between the reference signal and the output, and use the integrated value with a specific coefficient as a compensating signal. In this work, we discuss the relationship between the machine learning theory and the integral compensation. We focus on single-input-single-output discrete-time time-invariant systems, and show that the integral compensation can be understood in the context of the machine learning in this case. In particular, the integral compensation is identical to the online optimization with the standard stochastic gradient descent [1]. The above idea gives two suggestions. First, the integral compensation may become faster by employing other types of stochastic gradient descent. Many algorithms of stochastic gradient descent have been proposed in the machine learning literature, and such algorithms may improve the classical integral compensation. Second, it may become possible to reject the time-invariant state-dependent disturbance. The modeling error is a typical example of such disturbance. By learning such a disturbance, the control performance for repetitive motions will be improved compared to the integral compensation. This presentation discusses the pros and cons of regarding the integral compensation as the stochastic gradient descent optimization.

Synthesis of non-parametric type controllers based on I/O data from closed-loop experiment

This paper proposes a data-driven control method with non-parametric map. The proposed method yields the input for tracking control by minimizing a stochastic cost function which consists of I/O data set of the plant. One of the advantages of the proposed method is that it does not need a prior information about the nonlinear system such as their structure. In order to cope with the difficulty in handling the closed-loop I/O data, a penalty term is introduced in the cost function for computing the optimal control input. Experimental validation using a DC motor is performed to show the effectiveness of the proposed controllers.

Error bound analysis and optimal construction of non-parametric PWA models

This paper discusses non-parametric PWA models which are recently developed for modelling a broad class of nonlinear systems. First, as for the accuracy of such models, an upper bound of output errors between the original system and the PWA models is given. Then, a construction method of such models is proposed, which minimizes the error upper bound. Furthermore, simulation results are given to show the effectiveness of the proposed construction method.