研究会 (2019 年 06 月 08 日)
共催:SICE 九州支部 制御理論と応用に関する研究会
協力:SICE 制御部門 超スマート社会実現のためのシステム制御技術調査研究会
日時: 6/08(月) 14:00〜16:10
場所:
北九州市立大学 小倉サテライトキャンパス
(小倉駅(小倉城口)に直結しているアミュプラザ小倉の7階)
講演1: A 2-step algorithm for the estimation of time-varying single particle
tracking models using Maximum Likelihood
(Prof. Boris Godoy, Boston University; 14:00〜15:00)
講演2: Sparse optimization for dynamical systems
(池田卓矢, 北九州市立大学; 15:10〜16:10; 発表は日本語の予定)
懇親会: 16:30〜
小倉漁介酒場 魚衛門
(小倉駅直結アミュプラザ小倉1階(小倉宿 駅から三十歩横丁)
tel:050-5355-5402
参加者: Godoy(Boston Univ.), Celikovsky(Czech Academy of Sciences),
佐伯(広大), 西村(鹿児島大), 佐藤(佐賀大),
加藤(福工大), 畑田(福大), 永原, 池田(以上北九州大),
川邊, 山本(九大), 福井, 瀬部(九工大)
(以上敬称略)
問合せ先: 永原正章
(nagahara[a]kitakyu-u.ac.jp)
概要
1.Single particle tracking (SPT) is a powerful class of methods for
studying the dynamics of biomolecules inside living cells. The
techniques reveal both trajectories of individual particles, with a
resolution well below the diffraction limit of light, and the
parameters defining the motion model, such as diffusion coefficients
and confinement lengths. Existing algorithms assume these
parameters are constant throughout an experiment. However, it has
been demonstrated that they often vary with time as the tracked
particles move through different regions in the cell or as
conditions inside the cell change in response to stimuli. In this
work we apply the method of local Maximum Likelihood (ML) estimation
to the SPT application combined with change detection. Local ML uses
a sliding window over the data, estimating the model parameters in
each window. Once we have found the values for the parameters before
and after the change, we apply offline change detection to know the
exact time of the change. Then, we reestimate these parameters and
show that there is an improvement in the estimation of key
parameters found in SPT. Preliminary results using simulated data
with a basic diffusion model with additive Gaussian noise show that
our proposed algorithm is able to track abrupt changes in the
parameters as they evolve during a trajectory.
2.Optimization problems in which control variables are penalized via
the L0 norm have attracted a renewed interest due to its connection to
sparsity. Since the penalty cost is defined as the length of the
support of controls, the optimization tends to make the control input
identically zero on a set with positive measures and the optimal
control is switched off completely on parts of the time domain. This
is why the problem is referred to as sparse optimal control. In this
talk, I introduce some characterizations of the sparse optimal
control.
Firstly, we investigate the sparse optimal control for linear systems
with a final state constraint. In the light of the fact that the Lp
optimization problems with p ∈ (0, 1] have been seen in the context of
the sparse optimization, we show the relationship among the Lp
optimizations with p ∈ [0,1]. A sufficient condition under which the
sparse optimization problem is exactly solved by the associated L1
optimization problem is obtained. In addition, an equivalence theorem
among the Lp optimal solutions with p ∈ [0, 1) is shown.
Secondary, we investigate the sparse optimal control for nonlinear
systems with a terminal cost. In contrast to the linear case, I
directly deal with the underlying non-smooth and non-convex L0 cost
without the aid of any Lp relaxations and derive the sparse optimal
feedback map. For this purpose, the dynamic programming approach is
adopted and the value function is analyzed. Due to the non-smoothness
of the L0 cost functional, in general, the value function is not
differentiable in the domain. Then, we characterize the value function
as a viscosity solution to the associated Hamilton-Jacobi-Bellman
(HJB) equation. Based on the result, we derive a sufficient and
necessary condition for the L0 optimality, which immediately gives the
optimal feedback map. In addition, we show an equivalence with the
associated L1 optimal control problem.
Last modified: Thu Jun 13 21:02:00 JST 2019