sumu.mcmc module

class sumu.mcmc.MC3(chains, mode, **params)

Bases: Describable

adapt_temperatures()

describe()

Return a dict of stats as simple key:value pairs.

E.g., ‘p_local’: sum(self.local_history) / len(self.local_history)

static get_inv_temperatures(heating, M, step=1)

Returns the inverse temperatures in descending order.

static get_swap_acceptance_prob(chains, i, j)

Compute the swap acceptance probability between chains in indices i and j.

inverse_temperatures()

static make_swap(chains, i, j)

sample()

class sumu.mcmc.PartitionMCMC(C, score, d, inv_temp=1.0, move_weights={'DAG_edge_reversal': 2, 'R_split_merge': 1, 'R_swap_node_pair': 1}, R=None)

Bases: Describable

Partition-MCMC sampler [1] with efficient scoring.

DAG_edge_reversal(**kwargs)

R_split_merge(**kwargs)

R_swap_node_pair(**kwargs)

describe()

Return a dict of stats as simple key:value pairs.

E.g., ‘p_local’: sum(self.local_history) / len(self.local_history)

sample()

[1]

Jack Kuipers and Giusi Moffa. Partition MCMC for inference on acyclic digraphs. Journal of the American Statistical Association, 112(517):282–299, January 2017. arXiv:1504.05006, doi:10.1080/01621459.2015.1133426.