1CS 188: Artificial IntelligenceFall 2011Lecture 18: HMMs: Intro and Filtering11/2/2011Dan Klein --- UC BerkeleyPresented by Woody HoburgAnnouncements Midterm back today solutions online grades also in glookup P4 out Thursday22Reasoning over Time Often, we want to reason about a sequence of observations Robot localization Medical monitoring Speech recognition Vehicle control Need to introduce time into our models Basic approach: hidden Markov models (HMMs)3[VIDEO]Outline Markov Models(last lecture) Hidden Markov Models (HMMs)RepresentationInferenceForward algorithm (special case of variable elimination)Particle filtering (next lecture)43Markov Models: recap A Markov model is a chain-structured BN Each node is identically distributed (stationarity) Value of X at a given time is called the state As a BN: Parameters: called transition probabilities or dynamics, specify how the state evolves over time (also, initial probs)X2X1X3X4Conditional Independence Basic conditional independence: Past and future independent of the present Each time step only depends on the previous This is called the (first order) Markov property Note that the chain is just a (growing) BN We can always use generic BN reasoning on it if we truncate the chain at a fixed lengthX2X1X3X464Example: Markov Chain Weather: States: X = {rain, sun} Transitions: Initial distribution: 1.0 sun What’s the probability distribution after one step?rain sun0.90.90.10.1This are two new representations of a CPT, not BNs!7sunrainsunrain0.10.90.90.1Mini-Forward Algorithm Question: What’s P(X) on some day t? An instance of variable elimination! (In order X1, X2, … )sunrainsunrainsunrainsunrainForward simulation85Example From initial observation of sun From initial observation of rainP(X1) P(X2) P(X3) P(X∞)P(X1) P(X2) P(X3) P(X∞)9Outline Markov Models(last lecture) Hidden Markov Models (HMMs)RepresentationInferenceForward algorithm (special case of variable elimination)Particle filtering (next lecture)106Hidden Markov Models Markov chains not so useful for most agents Eventually you don’t know anything anymore Need observations to update your beliefs Hidden Markov models (HMMs) Underlying Markov chain over states S You observe outputs (effects) at each time step As a Bayes’ net:X5X2E1X1X3X4E2E3E4E5Example An HMM is defined by: Initial distribution: Transitions: Emissions:7Conditional Independence HMMs have two important independence properties: Markov hidden process, future depends on past via the present Current observation independent of all else given current state Quiz: does this mean that observations are independent given no evidence? [No, correlated by the hidden state]X5X2E1X1X3X4E2E3E4E5Real HMM Examples 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)8Filtering / Monitoring Filtering, or monitoring, is the task of tracking the distribution B(X) (the belief state) over time We start with B(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 programExample: Robot Localizationt=0Sensor model: can read in which directions there is a wall, never more than 1 mistakeMotion model: may not execute action with small prob.10ProbExample from Michael Pfeiffer9Example: Robot Localizationt=1Lighter grey: was possible to get the reading, but less likely b/c required 1 mistake10ProbExample: Robot Localizationt=210Prob10Example: Robot Localizationt=310ProbExample: Robot Localizationt=410Prob11Example: Robot Localizationt=510ProbInference: Base Cases Observation Given: P(X1), P(e1| X1) Query: P(x1| e1) ∀ x1E1X1X2X1 Passage of Time Given: P(X1), P(X2| X1) Query: P(x2) ∀ x212Passage of Time Assume we have current belief P(X | evidence to date) Then, after one time step passes: Or, compactly: Basic idea: beliefs get “pushed” through the transitions With the “B” notation, we have to be careful about what time step t the belief is about, and what evidence it includesX2X1Example: Passage of Time As time passes, uncertainty “accumulates”T = 1 T = 2 T = 5Transition model: ghosts usually go clockwise13Observation Assume we have current belief P(X | previous evidence): Then: Or: Basic idea: beliefs reweighted by likelihood of evidence Unlike passage of time, we have to renormalizeE1X1Example: Observation As we get observations, beliefs get reweighted, uncertainty “decreases”Before observation After observation14Example HMMThe Forward Algorithm We are given evidence at each time and want to know We can derive the following updates = exactly variable elimination in order X1, X2, …We can normalize as we go if we want to have P(x|e) at each time step, or just once at the end…15Online 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|2per time
View Full Document
Unlocking...