Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation...
Transcript of Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation...
![Page 1: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/1.jpg)
Markov Models and Hidden Markov Models
Robert PlattNortheastern University
Some images and slides are used from:1. CS188 UC Berkeley2. RN, AIMA
![Page 2: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/2.jpg)
Markov Models
We have already seen that an MDP provides a useful framework for modeling stochastic control problems.
Markov Models: model any kind of temporally dynamic system.
![Page 3: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/3.jpg)
Probability again: Independence
Two random variables, x and y, are independent when:
The outcomes of two different coin flips are usually independent of each other
Image: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 4: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/4.jpg)
Probability again: Independence
If:
Then:
Why?
![Page 5: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/5.jpg)
Are T and W independent?
T W P
hot sun 0.4
hot rain 0.1
cold sun 0.2
cold rain 0.3
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 6: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/6.jpg)
Are T and W independent?
T W P
hot sun 0.4
hot rain 0.1
cold sun 0.2
cold rain 0.3
T P
hot 0.5
cold 0.5
W P
sun 0.6
rain 0.4
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 7: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/7.jpg)
Are T and W independent?
T W P
hot sun 0.4
hot rain 0.1
cold sun 0.2
cold rain 0.3
T W P
hot sun 0.3
hot rain 0.2
cold sun 0.3
cold rain 0.2
T P
hot 0.5
cold 0.5
W P
sun 0.6
rain 0.4
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 8: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/8.jpg)
Conditional independence
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
Independence:
Conditional independence:
Equivalent statements of conditional independence:
![Page 9: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/9.jpg)
Conditional independence: example
cavity
toothache catch
P(toothache, catch | cavity) = P(toothache | cavity) = P(catch | cavity)
P(toothache | cavity) = P(toothache | cavity, catch)
P(catch | cavity) = P(catch | cavity, toothache)
or...
![Page 10: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/10.jpg)
Conditional independence: example
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
What about this domain:
Traffic Umbrella Raining
![Page 11: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/11.jpg)
Conditional independence: example
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
What about this domain:
Fire Smoke Alarm
![Page 12: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/12.jpg)
Markov Processes
transitions
State at time=1State at time=2
![Page 13: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/13.jpg)
Markov Processes
transitions
State at time=1State at time=2
Since this is a Markov process, we assume transitions are Markov:
Markov assumption:
Process model:
![Page 14: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/14.jpg)
Markov Processes
How do we calculate:
![Page 15: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/15.jpg)
Markov Processes
How do we calculate:
![Page 16: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/16.jpg)
Markov Processes
How do we calculate:
![Page 17: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/17.jpg)
Markov Processes
How do we calculate:
Can we simplify this expression?
![Page 18: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/18.jpg)
Markov Processes
How do we calculate:
![Page 19: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/19.jpg)
Markov Processes
How do we calculate:
![Page 20: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/20.jpg)
Markov Processes
How do we calculate:
In general:
![Page 21: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/21.jpg)
Markov Processes
How do we calculate:
In general:
Process model
![Page 22: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/22.jpg)
Markov Processes: example
Two new ways of representing the same CPT
sun
rain
sun
rain
0.1
0.9
0.7
0.3
States: X = {rain, sun}
rain sun0.9
0.7
0.3
0.1
Xt-1 Xt P(Xt|Xt-1)
sun sun 0.9
sun rain 0.1
rain sun 0.3
rain rain 0.7
Initial distribution: 1.0 sun
Process model: P(Xt | Xt-1):
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 23: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/23.jpg)
Simulating dynamics forward
Joint distribution:
But, suppose we want to predict the state at time T, given a prior distribution at time 1?
...
![Page 24: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/24.jpg)
Markov Processes: example
Initial distribution: 1.0 sun
What is the probability distribution after one step?
rain sun0.9
0.7
0.3
0.1
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 25: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/25.jpg)
Simulating dynamics forward
From initial observation of sun
From initial observation of rain
From yet another initial distribution P(X1):
P(X1) P(X2) P(X3) P(X)P(X4)
P(X1) P(X2) P(X3) P(X)P(X4)
P(X1) P(X)
…
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 26: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/26.jpg)
Simulating dynamics forward
From initial observation of sun
From initial observation of rain
From yet another initial distribution P(X1):
P(X1) P(X2) P(X3) P(X)P(X4)
P(X1) P(X2) P(X3) P(X)P(X4)
P(X1) P(X)
…
This is called the stationary distribution
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 27: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/27.jpg)
Hidden Markov Models (HMMs)
Hidden Markov Models: markov models applied to estimation problems
– speech to text– tracking in computer vision– robot localization
![Page 28: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/28.jpg)
Hidden Markov Models (HMMs)
State, , is assumed to be unobserved
However, you get to make one observation, , on each timestep.
Called an “emission”
![Page 29: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/29.jpg)
Hidden Markov Models (HMMs)
Sensor Markov Assumption: the current observation depends only on current state:
![Page 30: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/30.jpg)
HMM example
Rt Rt+1 P(Rt+1|Rt)
+r +r 0.7
+r -r 0.3
-r +r 0.3
-r -r 0.7
Rt Ut P(Ut|Rt)
+r +u 0.9
+r -u 0.1
-r +u 0.2
-r -u 0.8
An HMM is defined by:
Initial distribution: Transitions: Emissions:
Umbrellat-1
Umbrellat
Umbrellat+1
Raint-1 Raint Raint+1
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 31: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/31.jpg)
Real world HMM applications
Speech recognition HMMs: Observations are acoustic signals (continuous valued) States are specific positions in specific words (so, tens of
thousands)
Machine translation HMMs: Observations are words (tens of thousands) States are translation options
Robot tracking: Observations are range readings (continuous) States are positions on a map (continuous)
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 32: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/32.jpg)
HMM Filtering
Filtering, or monitoring, is the task of tracking the distribution Bt(X) = Pt(Xt | e1, …, et) (the belief state) over time
We start with B1(X) in an initial setting, usually uniform
As time passes, or we get observations, we update B(X)
The Kalman filter was invented in the 60’s and first implemented as a method of trajectory estimation for the Apollo program
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 33: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/33.jpg)
HMM Filtering
Given a prior distribution, , and a series of observations, , calculate the posterior distribution:
Two steps:
Process update Observation update
![Page 34: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/34.jpg)
HMM Filtering
Given a prior distribution, , and a series of observations, , calculate the posterior distribution:
Two steps:
Process update Observation update
![Page 35: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/35.jpg)
HMM Filtering
Given a prior distribution, , and a series of observations, , calculate the posterior distribution:
Two steps:
Process update Observation update
“Beliefs”
![Page 36: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/36.jpg)
Process update
This is just forward simulation of the Markov Model
![Page 37: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/37.jpg)
Process update: example
As time passes, uncertainty “accumulates”
T = 1 T = 2 T = 5
(Transition model: ghosts usually go clockwise)
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 38: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/38.jpg)
Observation update
Where is a normalization factor
![Page 39: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/39.jpg)
Observation update
As we get observations, beliefs get reweighted, uncertainty “decreases”
Before observation After observation
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 40: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/40.jpg)
Robot localization example
10
Observation model: can read in which directions there is a wall, never more than 1 mistake
Process model: may not execute action with small prob.
Prob
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 41: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/41.jpg)
Robot localization example
10
Lighter grey: was possible to get the reading, but less likely b/c required 1 mistake
Prob
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 42: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/42.jpg)
Robot localization example
10Prob
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 43: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/43.jpg)
Robot localization example
10Prob
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 44: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/44.jpg)
Robot localization example
10Prob
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 45: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/45.jpg)
Robot localization example
10Prob
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 46: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/46.jpg)
Weather HMM example
RtRt+1 P(Rt+1|Rt)
+r +r 0.7
+r -r 0.3
-r +r 0.3
-r -r 0.7
Rt Ut P(Ut|Rt)
+r +u 0.9
+r -u 0.1
-r +u 0.2
-r -u 0.8Umbrella1 Umbrella2
Rain0 Rain1 Rain2
B(+r) = 0.5B(-r) = 0.5
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 47: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/47.jpg)
Weather HMM example
Rt Rt+1 P(Rt+1|Rt)
+r +r 0.7
+r -r 0.3
-r +r 0.3
-r -r 0.7
Rt Ut P(Ut|Rt)
+r +u 0.9
+r -u 0.1
-r +u 0.2
-r -u 0.8Umbrella1 Umbrella2
Rain0 Rain1 Rain2
B(+r) = 0.5B(-r) = 0.5
B’(+r) = 0.5B’(-r) = 0.5
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 48: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/48.jpg)
Weather HMM example
Rt Rt+1 P(Rt+1|Rt)
+r +r 0.7
+r -r 0.3
-r +r 0.3
-r -r 0.7
Rt Ut P(Ut|Rt)
+r +u 0.9
+r -u 0.1
-r +u 0.2
-r -u 0.8Umbrella1 Umbrella2
Rain0 Rain1 Rain2
B(+r) = 0.5B(-r) = 0.5
B’(+r) = 0.5B’(-r) = 0.5
B(+r) = 0.818B(-r) = 0.182
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 49: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/49.jpg)
Weather HMM example
Rt Rt+1 P(Rt+1|Rt)
+r +r 0.7
+r -r 0.3
-r +r 0.3
-r -r 0.7
Rt Ut P(Ut|Rt)
+r +u 0.9
+r -u 0.1
-r +u 0.2
-r -u 0.8Umbrella1 Umbrella2
Rain0 Rain1 Rain2
B(+r) = 0.5B(-r) = 0.5
B’(+r) = 0.5B’(-r) = 0.5
B(+r) = 0.818B(-r) = 0.182
B’(+r) = 0.627B’(-r) = 0.373
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 50: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/50.jpg)
Weather HMM example
Rt Rt+1 P(Rt+1|Rt)
+r +r 0.7
+r -r 0.3
-r +r 0.3
-r -r 0.7
Rt Ut P(Ut|Rt)
+r +u 0.9
+r -u 0.1
-r +u 0.2
-r -u 0.8Umbrella1 Umbrella2
Rain0 Rain1 Rain2
B(+r) = 0.5B(-r) = 0.5
B’(+r) = 0.5B’(-r) = 0.5
B(+r) = 0.818B(-r) = 0.182
B’(+r) = 0.627B’(-r) = 0.373
B(+r) = 0.883B(-r) = 0.117
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 51: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/51.jpg)
Particle Filtering
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 52: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/52.jpg)
Representation: Particles
Our representation of P(X) is now a list of N particles (samples) Generally, N << |X| Storing map from X to counts would defeat the
point
P(x) approximated by number of particles with value x So, many x may have P(x) = 0! More particles, more accuracy
For now, all particles have a weight of 1
Particles: (3,3) (2,3) (3,3) (3,2) (3,3) (3,2) (1,2) (3,3) (3,3) (2,3)
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 53: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/53.jpg)
Particle Filtering: Elapse Time
Each particle is moved by sampling its next position from the transition model
This is like prior sampling – samples’ frequencies reflect the transition probabilities
Here, most samples move clockwise, but some move in another direction or stay in place
This captures the passage of time If enough samples, close to exact values
before and after (consistent)
Particles: (3,3) (2,3) (3,3) (3,2) (3,3) (3,2) (1,2) (3,3) (3,3) (2,3)
Particles: (3,2) (2,3) (3,2) (3,1) (3,3) (3,2) (1,3) (2,3) (3,2) (2,2)
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 54: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/54.jpg)
Slightly trickier: Don’t sample observation, fix it
Similar to likelihood weighting, downweight samples based on the evidence
As before, the probabilities don’t sum to one, since all have been downweighted (in fact they now sum to (N times) an approximation of P(e))
Particle Filtering: Observe
Particles: (3,2) w=.9 (2,3) w=.2 (3,2) w=.9 (3,1) w=.4 (3,3) w=.4 (3,2) w=.9 (1,3) w=.1 (2,3) w=.2 (3,2) w=.9 (2,2) w=.4
Particles: (3,2) (2,3) (3,2) (3,1) (3,3) (3,2) (1,3) (2,3) (3,2) (2,2)
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 55: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/55.jpg)
Particle Filtering: Resample
Rather than tracking weighted samples, we resample
N times, we choose from our weighted sample distribution (i.e. draw with replacement)
This is equivalent to renormalizing the distribution
Now the update is complete for this time step, continue with the next one
Particles: (3,2) w=.9 (2,3) w=.2 (3,2) w=.9 (3,1) w=.4 (3,3) w=.4 (3,2) w=.9 (1,3) w=.1 (2,3) w=.2 (3,2) w=.9 (2,2) w=.4
(New) Particles: (3,2) (2,2) (3,2) (2,3) (3,3) (3,2) (1,3) (2,3) (3,2) (3,2)
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 56: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/56.jpg)
Recap: Particle Filtering Particles: track samples of states rather than an explicit
distribution
Particles: (3,3) (2,3) (3,3) (3,2) (3,3) (3,2) (1,2) (3,3) (3,3) (2,3)
Elapse Weight Resample
Particles: (3,2) (2,3) (3,2) (3,1) (3,3) (3,2) (1,3) (2,3) (3,2) (2,2)
Particles: (3,2) w=.9 (2,3) w=.2 (3,2) w=.9 (3,1) w=.4 (3,3) w=.4 (3,2) w=.9 (1,3) w=.1 (2,3) w=.2 (3,2) w=.9 (2,2) w=.4
(New) Particles: (3,2) (2,2) (3,2) (2,3) (3,3) (3,2) (1,3) (2,3) (3,2) (3,2)
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 57: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/57.jpg)
Robot Localization
In robot localization: We know the map, but not the robot’s position Observations may be vectors of range finder
readings State space and readings are typically
continuous (works basically like a very fine grid) and so we cannot store B(X)
Particle filtering is a main technique
Slide: Berkeley CS188 course notes (downloaded Summer 2015)
![Page 58: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/58.jpg)
Particle Filter Localization (Sonar)
![Page 59: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/59.jpg)
Particle Filter Localization (Laser)
![Page 60: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/60.jpg)
Dynamic Bayes Nets
![Page 61: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/61.jpg)
Dynamic Bayes Nets (DBNs)
We want to track multiple variables over time, using multiple sources of evidence
Idea: Repeat a fixed Bayes net structure at each time
Variables from time t can condition on those from t-1
Dynamic Bayes nets are a generalization of HMMs
G1a
E1a E1
b
G1b
G2a
E2a E2
b
G2b
t =1 t =2
G3a
E3a E3
b
G3b
t =3
![Page 62: Markov Models and Hidden Markov ModelsHidden Markov Models: markov models applied to estimation problems – speech to text – tracking in computer vision – robot localization.](https://reader034.fdocuments.net/reader034/viewer/2022051207/603e44d7b5cc8e42ac6f4284/html5/thumbnails/62.jpg)
DBN Particle Filters
A particle is a complete sample for a time step
Initialize: Generate prior samples for the t=1 Bayes net Example particle: G1
a = (3,3) G1b = (5,3)
Elapse time: Sample a successor for each particle Example successor: G2
a = (2,3) G2b = (6,3)
Observe: Weight each entire sample by the likelihood of the evidence conditioned on the sample Likelihood: P(E1
a |G1a ) * P(E1
b |G1b )
Resample: Select prior samples (tuples of values) in proportion to their likelihood