Movatterモバイル変換

Cluster Assignment

Description

Assign sequences to cluster models that give the highest sequence-to-hmmlikelihood. Used inhmm.clust.

Usage

assign.clusters(partition, memberships, sequences, smoothing = 1e-04)

Arguments

Value

The updated matrix with cluster memberships for each sequence.

See Also

Used in main function for the DBHC algorithmhmm.clust.

HMM BIC

Description

Compute the BIC of a single HMM given a threshold epsilon for countingparameters. Auxiliary function used insize.search.

Usage

cluster.bic(hmm, eps = 0.001)

Arguments

Value

The BIC ofhmm.

See Also

Used insize.search.

Count HMM Parameters

Description

Count the number of parameters in an HMM larger than a small number epsilon.Auxiliary function used inpartition.bic andcluster.bic.

Usage

count.parameters(hmm, eps = 0.001)

Arguments

Value

The number of parameters larger thaneps.

See Also

Used inpartition.bic andcluster.bic.

Heatmap Emission Probabilities

Description

Plots a heatmap of an HMM's emission probabilities.

Usage

emission.heatmap(emission, base_size = 10)

Arguments

See Also

Seehmm.clust for an example.

DBHC Algorithm

Description

Implementation of the DBHC algorithm, an HMM clustering algorithm that findsa mixture of discrete-output HMMs. The algorithm uses heuristics based on BICto search for the optimal number of hidden states in each HMM and the optimalnumber of clusters.

Usage

hmm.clust(  sequences,  id = NULL,  smoothing = 1e-04,  eps = 0.001,  init.size = 2,  alphabet = NULL,  K.max = NULL,  log_space = FALSE,  print = FALSE,  seed.size = 3)

Arguments

Value

A list with components:

sequences

Anstslist object of sequences with discrete observations.

id

A vector with ids that identify the sequences insequences.

cluster

A vector with found clustermemberships for the sequences.

partition

A list object withthe partition, a mixture of HMMs. Each element in the list is anhmmobject.

memberships

A matrix with cluster memberships foreach sequence.

n.clusters

Numerical, the found number ofclusters.

sizes

A vector with the number of HMM states foreach cluster model.

bic

A vector with the BICs for eachcluster model.

Examples

## Simulated datalibrary(seqHMM)output.labels <-  c("H", "T")# HMM 1states.1 <- c("A", "B", "C")transitions.1 <- matrix(c(0.8,0.1,0.1,0.1,0.8,0.1,0.1,0.1,0.8), nrow = 3)rownames(transitions.1) <- states.1colnames(transitions.1) <- states.1emissions.1 <- matrix(c(0.5,0.75,0.25,0.5,0.25,0.75), nrow = 3)rownames(emissions.1) <- states.1colnames(emissions.1) <- output.labelsinitials.1 <- c(1/3,1/3,1/3)# HMM 2states.2 <- c("A", "B")transitions.2 <- matrix(c(0.75,0.25,0.25,0.75), nrow = 2)rownames(transitions.2) <- states.2colnames(transitions.2) <- states.2emissions.2 <- matrix(c(0.8,0.6,0.2,0.4), nrow = 2)rownames(emissions.2) <- states.2colnames(emissions.2) <- output.labelsinitials.2 <- c(0.5,0.5)# Simulatehmm.sim.1 <- simulate_hmm(n_sequences = 100,                          initial_probs = initials.1,                          transition_probs = transitions.1,                          emission_probs = emissions.1,                          sequence_length = 25)hmm.sim.2 <- simulate_hmm(n_sequences = 100,                          initial_probs = initials.2,                          transition_probs = transitions.2,                          emission_probs = emissions.2,                          sequence_length = 25)sequences <- rbind(hmm.sim.1$observations, hmm.sim.2$observations)n <- nrow(sequences)# Clustering algorithmid <- paste0("K-", 1:n)rownames(sequences) <- idsequences <- sequences[sample(1:n, n),]res <- hmm.clust(sequences, id = rownames(sequences))############################################################################### Swiss Household Datadata("biofam", package = "TraMineR")# Clustering algorithmnew.alphabet <- c("P", "L", "M", "LM", "C", "LC", "LMC", "D")sequences <- seqdef(biofam[,10:25], alphabet = 0:7, states = new.alphabet)## Not run: res <- hmm.clust(sequences)# Heatmapscluster <- 1  # display heatmaps for cluster 1transition.heatmap(res$partition[[cluster]]$transition_probs,                   res$partition[[cluster]]$initial_probs)emission.heatmap(res$partition[[cluster]]$emission_probs)## End(Not run)## A smaller example, which takes less time to runsubset <- sequences[sample(1:nrow(sequences), 20, replace = FALSE),]# Clustering algorithm, limiting number of clusters to 2res <- hmm.clust(subset, K.max = 2)# Number of clustersprint(res$n.clusters)# Table of cluster membershipstable(res$memberships[,"cluster"])# BIC for each number of clustersprint(res$bic)# Heatmapscluster <- 1  # display heatmaps for cluster 1transition.heatmap(res$partition[[cluster]]$transition_probs,                   res$partition[[cluster]]$initial_probs)emission.heatmap(res$partition[[cluster]]$emission_probs)

Get HMM Log Likelihood

Description

Get the log likelihood of an HMM object and check if it is feasible (i.e.,contains no illegal emissions). Auxiliary function used inpartition.bic.

Usage

model.ll(hmm)

Arguments

Value

The log likelihood of thehmm object, print warning if modelis infeasible (i.e., if the log likelihood is evaluated for a sequence thatcontains emissions that are assigned probability 0 in thehmmobject).

See Also

Used inpartition.bic.

Partition BIC

Description

Compute the BIC of a partition given a threshold epsilon for countingparameters. Auxiliary function used inhmm.clust.

Usage

partition.bic(partition, eps = 0.001)

Arguments

Value

The BIC of the partition.

See Also

Used in main function for the DBHC algorithmhmm.clust.

Seed Selection Procedure

Description

Seed selection procedure of the DBHC algorithm, also invokes size searchalgorithm for seed insize.search. Used inhmm.clust.

Usage

select.seeds(  sequences,  log_space = FALSE,  K,  seed.size = 3,  init.size = 2,  print = FALSE,  smoothing = 1e-04)

Arguments

Value

A partition as a list object with HMMs for the selected seeds.

See Also

Used in main function for the DBHC algorithmhmm.clust.

Sequence-to-HMM Likelihood

Description

Compute the sequence-to-HMM likelihood of an HMM evaluated for a singlesequence and check if the sequence contains emissions that are not possibleaccording to the HMM. Auxiliary function used inselect.seedsandassign.clusters.

Usage

seq2hmm.ll(hmm)

Arguments

Value

The log likelihood of the sequence contained inhmm, valuewill be set to minus infinity if the sequence contains illegal emissions.

See Also

Used inselect.seeds andassign.clusters.

Size Search Algorithm

Description

The size search algorithm finds the optimal number of HMM states for a set ofsequences and returns both the optimalhmm object and thecorresponding number of hidden states. Used inselect.seeds.

Usage

size.search(sequences, log_space = FALSE, print = FALSE)

Arguments

Value

A list with the optimal number of HMM states and the optimalhmm object.

See Also

Used in the DBHC seed selection procedure inselect.seeds.

Smooth HMM Parameters

Description

Smooth the parameters of an HMM using absolute discounting given a thresholdepsilon. Auxiliary function used inselect.seeds,assign.clusters, andhmm.clust.

Usage

smooth.hmm(hmm, smoothing = 1e-04)

Arguments

Value

Anhmm object with smoothed probabilities.

See Also

Used inselect.seeds,assign.clusters,and main function for the DBHC algorithmhmm.clust.

Smooth Probabilities

Description

Smooth a vector of probabilities using absolute discounting. Auxiliaryfunction used insmooth.hmm.

Usage

smooth.probabilities(probs, smoothing = 1e-04)

Arguments

Value

A vector of smoothed probabilities.

See Also

Used insmooth.hmm.

Heatmap Transition Probabilities

Description

Plots a heatmap of an HMM's initial and transition probabilities.

Usage

transition.heatmap(transition, initial = NULL, base_size = 10)

Arguments

See Also

Seehmm.clust for an example.