1. QUIZ!!  In HMMs...  T/F:... the emissions are hidden. FALSE  T/F:... observations are independent given no evidence. FALSE  T/F:... each variable X i has its own (different) CPT. FALSE  T/F:... typically none of the variables X i are observed. TRUE  T/F:... can be solved with variable elimination up to any time t. TRUE  T/F: In the forward algorithm, you must normalize each iteration. FALSE 1 X2X2 E1E1 X1X1 X3X3 X4X4 E2E2 E3E3 E4E4

2. CSE 511a: Artificial Intelligence Spring 2013 Lecture 20: HMMs Particle Filters 04/15/2012 Robert Pless via Kilian Q. Weinberger w/ slides from Dan Klein – UC Berkeley

3. Announcements  Project 4 is up!  Due in one week!!  Contest Coming Soon 3

4. Recap: Reasoning Over Time  Stationary Markov models  Hidden Markov models X2X2 X1X1 X3X3 X4X4 rainsun 0.7 0.3 X5X5 X2X2 E1E1 X1X1 X3X3 X4X4 E2E2 E3E3 E4E4 E5E5 XEP rainumbrella0.9 rainno umbrella0.1 sunumbrella0.2 sunno umbrella0.8

5. Exact Inference 5

6. Just like Variable Elimination 6 X2X2 e1e1 X1X1 X3X3 X4X4 ¬e 2 e3e3 e4e4 XEP (E|X) rainumbrella0.9 rainno umbrella0.1 sunumbrella0.2 sunno umbrella0.8 X1X1 P( X 1 ) rain0.1 sun0.9 XP (e|X) rain0.9 sun0.2 evidence X1X1 P( X 1,e 1 ) rain0.09 sun0.18 Emission probability:

7. Just like Variable Elimination 7 X2X2 X 1,e 1 X3X3 X4X4 ¬e2¬e2 e3e3 e4e4 X1X1 P( X 1,e 1 ) rain0.09 sun0.18 XtXt X t+1 P (X t+1 |X t ) rain 0.9 rainsun0.1 sunrain0.2 sun 0.8 Transition probability: X1X1 X2X2 P (X 2,X 1,e 1 ) rain 0.081 rainsun0.009 sunrain0.036 sun 0.144

8. Just like Variable Elimination 8 X 1,X 2,e 1 X3X3 X4X4 ¬e2¬e2 e3e3 e4e4 Marginalize out X 1... X1X1 X2X2 P (X 2,X 1, e 1 ) rain 0.081 rainsun0.009 sunrain0.036 sun 0.144 X2X2 P( X 2,e 1 ) rain0.117 sun0.153

9. Just like Variable Elimination 9 X 2,e 1 X3X3 X4X4 ¬e2¬e2 e3e3 e4e4 X2X2 P( X 2,e 1 ) rain0.117 sun0.153 XEP (E|X) rainumbrella0.9 rainno umbrella0.1 sunumbrella0.2 sunno umbrella0.8 Emission probability:

10. Just like Variable Elimination 10 X 2,e 1, ¬e 2 X3X3 X4X4 e3e3 e4e4

11. The Forward Algorithm  We are given evidence at each time and want to know  We can derive the following updates We can normalize as we go if we want to have P(x|e) at each time step, or just once at the end…

12. Online Belief Updates  Every time step, we start with current P(X | evidence)  We update for time:  We update for evidence:  The forward algorithm does both at once (and doesn’t normalize)  Problem: space is |X| and time is |X| 2 per time step X2X2 X1X1 X2X2 E2E2

13. Best Explanation Queries  Query: most likely seq: X5X5 X2X2 E1E1 X1X1 X3X3 X4X4 E2E2 E3E3 E4E4 E5E5 13

14. State Path Trellis  State trellis: graph of states and transitions over time  Each arc represents some transition  Each arc has weight  Each path is a sequence of states  The product of weights on a path is the seq’s probability  Can think of the Forward algorithm as computing sums of all paths in this graph sun rain sun rain sun rain sun rain 14

15. Viterbi Algorithm sun rain sun rain sun rain sun rain 15

16. Example 16

17. Filtering  Filtering is the inference process of finding a distribution over X T given e 1 through e T : P( X T | e 1:t )  We first compute P( X 1 | e 1 ):  For each t from 2 to T, we have P( X t-1 | e 1:t-1 )  Elapse time: compute P( X t | e 1:t-1 )  Observe: compute P(X t | e 1:t-1, e t ) = P( X t | e 1:t ) 17

18. Recap: Reasoning Over Time  Stationary Markov models  Hidden Markov models X2X2 X1X1 X3X3 X4X4 rainsun 0.7 0.3 X5X5 X2X2 E1E1 X1X1 X3X3 X4X4 E2E2 E3E3 E4E4 E5E5 XEP rainumbrella0.9 rainno umbrella0.1 sunumbrella0.2 sunno umbrella0.8

19. Recap: Filtering  Elapse time: compute P( X t | e 1:t-1 ) Observe: compute P( X t | e 1:t ) X2X2 E1E1 X1X1 E2E2 Belief: Prior on X 1 Observe Elapse time Observe

20. Recap: Filtering  Elapse time: compute P( X t | e 1:t-1 ) Observe: compute P( X t | e 1:t ) X2X2 E1E1 X1X1 E2E2 Belief: Prior on X 1 Observe Elapse time Observe

21. Approximate Inference 21

22. Particle Filtering  Sometimes |X| is too big to use exact inference  |X| may be too big to even store B(X)  E.g. X is continuous  |X| 2 may be too big to do updates  Solution: approximate inference  Track samples of X, not all values  Samples are called particles  Time per step is linear in the number of samples  But: number needed may be large  In memory: list of particles, not states  This is how robot localization works in practice 0.00.1 0.0 0.2 0.00.20.5

23. 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 will have P(x) = 0!  More particles, more accuracy  For now, all particles have a weight of 1 23 Particles: (3,3) (2,3) (3,3) (3,2) (3,3) (3,2) (2,1) (3,3) (2,1)

24. 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 probs  This world is a pacman world with default clockwise motions, so here, most samples move clockwise, but some move in another direction or stay in place  This captures the passage of time  If we have enough samples, close to the exact values before and after (consistent)

25. Particle Filtering: Observe  Slightly trickier:  Don’t do rejection sampling (why not?)  We don’t sample the observation, we fix it  This is similar to likelihood weighting, so we downweight our samples based on the evidence  Note that, as before, the probabilities don’t sum to one, since most have been downweighted (in fact they sum to an approximation of P(e))

26. 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 Old Particles: (3,3) w=0.1 (2,1) w=0.9 (3,1) w=0.4 (3,2) w=0.3 (2,2) w=0.4 (1,1) w=0.4 (3,1) w=0.4 (2,1) w=0.9 (3,2) w=0.3 New Particles: (2,1) w=1 (3,2) w=1 (2,2) w=1 (2,1) w=1 (1,1) w=1 (3,1) w=1 (2,1) w=1 (1,1) w=1

27. 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  [Demos,Fox]DemosFox

28. P4: Ghostbusters 2.0 (beta)  Plot: Pacman's grandfather, Grandpac, learned to hunt ghosts for sport.  He was blinded by his power, but could hear the ghosts’ banging and clanging.  Transition Model: All ghosts move randomly, but are sometimes biased  Emission Model: Pacman knows a “noisy” distance to each ghost Noisy distance prob True distance = 8 [Demo]

29. 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  Discrete valued dynamic Bayes nets are also HMMs G1aG1a E1aE1a E1bE1b G1bG1b G2aG2a E2aE2a E2bE2b G2bG2b t =1t =2 G3aG3a E3aE3a E3bE3b G3bG3b t =3

30. Exact Inference in DBNs  Variable elimination applies to dynamic Bayes nets  Procedure: “unroll” the network for T time steps, then eliminate variables until P(X T |e 1:T ) is computed  Online belief updates: Eliminate all variables from the previous time step; store factors for current time only 30 G1aG1a E1aE1a E1bE1b G1bG1b G2aG2a E2aE2a E2bE2b G2bG2b G3aG3a E3aE3a E3bE3b G3bG3b t =1t =2t =3 G3bG3b

31. 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: G 1 a = (3,3) G 1 b = (5,3)  Elapse time: Sample a successor for each particle  Example successor: G 2 a = (2,3) G 2 b = (6,3)  Observe: Weight each entire sample by the likelihood of the evidence conditioned on the sample  Likelihood: P(E 1 a |G 1 a ) * P(E 1 b |G 1 b )  Resample: Select prior samples (tuples of values) in proportion to their likelihood 31

32. SLAM  SLAM = Simultaneous Localization And Mapping  We do not know the map or our location  Our belief state is over maps and positions!  Main techniques: Kalman filtering (Gaussian HMMs) and particle methods DP-SLAM, Ron Parr

33. Summary Filtering  Push Belief through time  Incorporate observations  Inference:  Forward Algorithm (/ Variable Elimination)  Particle Filter (/ Sampling)  Algorithms:  Forward Algorithm (Belief given evidence)  Viterbi Algorithm (Most likely sequence) 33

34. Filtering  Filtering is the inference process of finding a distribution over X T given e 1 through e T : P( X T | e 1:t )  We first compute P( X 1 | e 1 ):  For each t from 2 to T, we have P( X t-1 | e 1:t-1 )  Elapse time: compute P( X t | e 1:t-1 )  Observe: compute P(X t | e 1:t-1, e t ) = P( X t | e 1:t ) 34