bayesian model selection and multi-target tracking presenters: xingqiu zhao and nikki hu joint work...

26
Bayesian Model Selection Bayesian Model Selection and Multi-target and Multi-target Tracking Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun University of Alberta Supported by NSERC, MITACS, PIMS Lockheed Martin Naval Electronics and Surveillance System Lockheed Martin Canada, APR. Inc

Post on 18-Dec-2015

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

Bayesian Model Selection and Bayesian Model Selection and Multi-target TrackingMulti-target Tracking

Presenters: Xingqiu Zhao and Nikki Hu

Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

University of Alberta

Supported by NSERC, MITACS, PIMS Lockheed Martin Naval Electronics and Surveillance System

Lockheed Martin Canada, APR. Inc

Page 2: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

Outline Outline

• Introduction

• Simulation Studies

• Filtering Equations

• Markov Chain Approximations

• Model Selection

• Future Work

Page 3: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

1. Introduction1. Introduction• Motivation: Submarine tracking and fish farming

• Model:

- Signal:

(1)

d

Page 4: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

- Observation:

(2)

• Goal: to find the best estimation for the number of targets and the location of each target.

Page 5: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

2. Simulation Studies2. Simulation Studies

Page 6: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

3. filtering equations3. filtering equations

• Notations : the space of bounded continuous functions on ; : the set of all cadlag functions from into ; : the spaces of probability measures; : the spaces of positive finite measures on ; : state space of .

Page 7: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

Let , , and .

Define

Page 8: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• The generator of Let

where .

Page 9: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

For any ,

we define

where

and

Page 10: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• Conditions:

C1. and satisfy the Lipschitz conditions.

C2.

C3.

C4.

Page 11: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• Theorem 1. The equation (1) has a unique solution

a.s.,

which is an -valued Markov process.

Page 12: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• Bayes formula and filtering equations

Theorem 2. Suppose that C1-C3 hold. Then

(i)

(ii)

where

is the innovation process.

(iii)

Page 13: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• Uniqueness

Theorem 3. Suppose that C1-C4 hold. Let be an -

adapted cadlag process which is a solution of the Kushner-FKK equation

where

Then , for all a.s.

Page 14: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

Theorem 4 Suppose that C1-C4 hold. If is an - adapted

-valued cadlag process satisfying

and

Then , for all a.s.

Page 15: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

4. 4. Markov chain approximationsMarkov chain approximations

• Step 1: Constructing smooth approximation

of the observation process

• Step 2: Dividing D and

Let ,

For , let

For , let

Page 16: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

Note that if is a rearrangement

of . Let

then . For , let .

For , with 1 in the i-th coordinate.

Page 17: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• Step 3: Constructing the Markov chain approximations

— Method 1:Method 1:

Let .

Set . One can find that

and

Page 18: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

Define as

and for , define as

let

Page 19: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

──Method 2Method 2:: Let and ,

Then

and

Define as for .

(μ ) (μ ) μNNA F L f k k

Page 20: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• Let as , take denote the integer part,

set

and let satisfy

Page 21: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

Then, the Markov chain approximation is given by

Theorem 5.

in probability on

for almost every sample path of .

Page 22: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

5. Model selection5. Model selection • Assume that the possible number of targets is , .

Model k: , . Which model is better?

• Bayesian FactorsBayesian Factors

Define the filter ratio processes as

Page 23: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• The Evolution of Bayesian Factors Let and be independent and Y be Brownian

motion on some probability space.

Theorem 3.

Let be the generator of , . Suppose that is continuous. Then is the unique measure-valued pair solution of the following system of SDEs,

(3)

Page 24: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

for , and

(4)

for , where is the optimal filter for model k, and

Page 25: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

• Markov chain approximations Applying the method in Section 3, one can construct

Markov chain approximations to equations (3) and (4).

Page 26: Bayesian Model Selection and Multi-target Tracking Presenters: Xingqiu Zhao and Nikki Hu Joint work with M. A. Kouritzin, H. Long, J. McCrosky, W. Sun

6. Future work6. Future work

• Number of targets is a random variable

• Number of Targets is a random process