Constructor Detail

Method Detail

• mapFromTrace

  @JvmOverloads() final static MAPFitResult mapFromTrace(DoubleArray trace, IntArray orders, Integer maxIter, Double stopCond, Array<Matrix> initialGuess, String resultFormat)

  Performs MAP fitting using the EM algorithm (ErCHMM).

  This implements the Erlang-based Continuous Hidden Markov Model algorithm for fitting a Markovian Arrival Process to trace data, as described in:

  • Okamura, Hiroyuki, and Tadashi Dohi. "Faster maximum likelihood estimation algorithms for Markovian arrival processes." QEST 2009.

  • Horvath, Gabor, and Hiroyuki Okamura. "A Fast EM Algorithm for Fitting Marked Markovian Arrival Processes with a New Special Structure." Computer Performance Engineering, Springer 2013.

  When orders contains the explicit Erlang branch orders (e.g., 1,2,3), the function performs EM fitting with those branches. When orders contains a single integer N, all possible branch-number / order combinations summing to N are tested and the best (highest log-likelihood) result is returned.

  The algorithm:

  • Initialize transition matrix P between Erlang branches and rate parameters

  • Run EM iterations:

  • Convert the best ErCHMM to MAP(D0, D1) representation

  Parameters:

  trace - The samples of the trace

  orders - The Erlang branch orders.

  maxIter - Maximum number of EM iterations (default 200)

  stopCond - The algorithm stops when relative log-likelihood improvement falls below this threshold (default 1e-7)

  initialGuess - Optional initial guess as array D0, D1 of MAP matrices.

  Returns:

  A Pair of (D0, D1) matrices representing the fitted MAP

• mapFromTrace

  @JvmOverloads() final static MAPFitResult mapFromTrace(DoubleArray trace, IntArray orders, Integer maxIter, Double stopCond, Array<Matrix> initialGuess)

  Performs MAP fitting using the EM algorithm (ErCHMM).

  This implements the Erlang-based Continuous Hidden Markov Model algorithm for fitting a Markovian Arrival Process to trace data, as described in:

  • Okamura, Hiroyuki, and Tadashi Dohi. "Faster maximum likelihood estimation algorithms for Markovian arrival processes." QEST 2009.

  • Horvath, Gabor, and Hiroyuki Okamura. "A Fast EM Algorithm for Fitting Marked Markovian Arrival Processes with a New Special Structure." Computer Performance Engineering, Springer 2013.

  When orders contains the explicit Erlang branch orders (e.g., 1,2,3), the function performs EM fitting with those branches. When orders contains a single integer N, all possible branch-number / order combinations summing to N are tested and the best (highest log-likelihood) result is returned.

  The algorithm:

  • Initialize transition matrix P between Erlang branches and rate parameters

  • Run EM iterations:

  • Convert the best ErCHMM to MAP(D0, D1) representation

  Parameters:

  trace - The samples of the trace

  orders - The Erlang branch orders.

  maxIter - Maximum number of EM iterations (default 200)

  stopCond - The algorithm stops when relative log-likelihood improvement falls below this threshold (default 1e-7)

  initialGuess - Optional initial guess as array D0, D1 of MAP matrices.

  Returns:

  A Pair of (D0, D1) matrices representing the fitted MAP

• mapFromTrace

  @JvmOverloads() final static MAPFitResult mapFromTrace(DoubleArray trace, IntArray orders, Integer maxIter, Double stopCond)

  Performs MAP fitting using the EM algorithm (ErCHMM).

  This implements the Erlang-based Continuous Hidden Markov Model algorithm for fitting a Markovian Arrival Process to trace data, as described in:

  • Okamura, Hiroyuki, and Tadashi Dohi. "Faster maximum likelihood estimation algorithms for Markovian arrival processes." QEST 2009.

  • Horvath, Gabor, and Hiroyuki Okamura. "A Fast EM Algorithm for Fitting Marked Markovian Arrival Processes with a New Special Structure." Computer Performance Engineering, Springer 2013.

  When orders contains the explicit Erlang branch orders (e.g., 1,2,3), the function performs EM fitting with those branches. When orders contains a single integer N, all possible branch-number / order combinations summing to N are tested and the best (highest log-likelihood) result is returned.

  The algorithm:

  • Initialize transition matrix P between Erlang branches and rate parameters

  • Run EM iterations:

  • Convert the best ErCHMM to MAP(D0, D1) representation

  Parameters:

  trace - The samples of the trace

  orders - The Erlang branch orders.

  maxIter - Maximum number of EM iterations (default 200)

  stopCond - The algorithm stops when relative log-likelihood improvement falls below this threshold (default 1e-7)

  Returns:

  A Pair of (D0, D1) matrices representing the fitted MAP

• mapFromTrace

  @JvmOverloads() final static MAPFitResult mapFromTrace(DoubleArray trace, IntArray orders, Integer maxIter)

  Performs MAP fitting using the EM algorithm (ErCHMM).

  This implements the Erlang-based Continuous Hidden Markov Model algorithm for fitting a Markovian Arrival Process to trace data, as described in:

  • Okamura, Hiroyuki, and Tadashi Dohi. "Faster maximum likelihood estimation algorithms for Markovian arrival processes." QEST 2009.

  • Horvath, Gabor, and Hiroyuki Okamura. "A Fast EM Algorithm for Fitting Marked Markovian Arrival Processes with a New Special Structure." Computer Performance Engineering, Springer 2013.

  When orders contains the explicit Erlang branch orders (e.g., 1,2,3), the function performs EM fitting with those branches. When orders contains a single integer N, all possible branch-number / order combinations summing to N are tested and the best (highest log-likelihood) result is returned.

  The algorithm:

  • Initialize transition matrix P between Erlang branches and rate parameters

  • Run EM iterations:

  • Convert the best ErCHMM to MAP(D0, D1) representation

  Parameters:

  trace - The samples of the trace

  orders - The Erlang branch orders.

  maxIter - Maximum number of EM iterations (default 200)

  Returns:

  A Pair of (D0, D1) matrices representing the fitted MAP

• mapFromTrace

  @JvmOverloads() final static MAPFitResult mapFromTrace(DoubleArray trace, IntArray orders)

  Performs MAP fitting using the EM algorithm (ErCHMM).

  This implements the Erlang-based Continuous Hidden Markov Model algorithm for fitting a Markovian Arrival Process to trace data, as described in:

  • Okamura, Hiroyuki, and Tadashi Dohi. "Faster maximum likelihood estimation algorithms for Markovian arrival processes." QEST 2009.

  • Horvath, Gabor, and Hiroyuki Okamura. "A Fast EM Algorithm for Fitting Marked Markovian Arrival Processes with a New Special Structure." Computer Performance Engineering, Springer 2013.

  When orders contains the explicit Erlang branch orders (e.g., 1,2,3), the function performs EM fitting with those branches. When orders contains a single integer N, all possible branch-number / order combinations summing to N are tested and the best (highest log-likelihood) result is returned.

  The algorithm:

  • Initialize transition matrix P between Erlang branches and rate parameters

  • Run EM iterations:

  • Convert the best ErCHMM to MAP(D0, D1) representation

  Parameters:

  trace - The samples of the trace

  orders - The Erlang branch orders.

  Returns:

  A Pair of (D0, D1) matrices representing the fitted MAP