membership.py - cMonkey cluster membership functionality
This module captures the microarray-specific scoring component
of cMonkey.
 
This file is part of cMonkey Python. Please see README and LICENSE for
more information and licensing details.

Modules
		array cPickle cmonkey.datamatrix logging math numpy random rpy2.robjects sqlite3 sys cmonkey.util

Classes
		OrigMembership   class OrigMembership     This is an implementation of a membership data structure that more closely resembles the R original. It is much simpler than ClusterMembership, with a smaller memory footprint     Methods defined here: __init__(self, row_names, col_names, row_is_member_of, col_is_member_of, config_params, row_indexes=None, col_indexes=None) identical constructor to ClusterMembership add_cluster_to_column(self, col, cluster, force=False) add_cluster_to_row(self, row, cluster, force=False) clusters_for_column(self, column) determine the clusters for the specified column clusters_for_row(self, row) determine the clusters for the specified row clusters_not_in_column(self, col, clusters) clusters_not_in_row(self, row, clusters) columns_for_cluster(self, cluster) free_slots_for_column(self, col) free_slots_for_row(self, row) is_column_in_cluster(self, col, cluster) is_row_in_cluster(self, row, cluster) max_changes_per_col(self) returns the maximum number of changes per column max_changes_per_row(self) returns the maximum number of changes per row max_cluster_rows_allowed(self) returns the maximum number of rows that should be in a cluster min_cluster_columns_allowed(self) returns the minimum number of columns that should be in a cluster min_cluster_rows_allowed(self) returns the minimum number of rows that should be in a cluster num_clusters(self) returns the number of clusters num_clusters_for_column(self, column) returns the number of clusters for the column num_clusters_for_row(self, row) returns the number of clusters for the row num_clusters_per_column(self) returns the number of clusters per row num_clusters_per_row(self) returns the number of clusters per row num_column_members(self, cluster) num_row_members(self, cluster) pickle_path(self) returns the function-specific pickle-path probability_seeing_col_change(self) returns the probability for seeing a row change probability_seeing_row_change(self) returns the probability for seeing a row change replace_column_cluster(self, col, index, new) replace_row_cluster(self, row, index, new) rows_for_cluster(self, cluster) update(self, matrix, row_scores, column_scores, num_iterations, iteration_result) top-level update method write_column_members(self, filename) Mostly for debugging, write out the current column membership state into a TSV file write_row_members(self, filename) Mostly for debugging, write out the current row membership state into a TSV file

Functions
		adjust_cluster(membership, cluster, rowscores, cutoff, limit) adjust a single cluster compensate_size(membership, matrix, rd_scores, cd_scores) size compensation function create_membership(matrix, seed_row_memberships, seed_column_memberships, config_params) create instance of ClusterMembership using the provided seeding algorithms fuzzify(membership, row_scores, column_scores, num_iterations, iteration_result, add_fuzz) Provide an iteration-specific fuzzification get_best_clusters(scores, n, sort=False) retrieve the n best scored clusters for the given row/column score matrix get_cc_scores(membership, scores, bandwidth, cluster) calculate the density scores for the given column score values in the specified cluster get_col_density_scores(membership, col_scores) get_density_scores(membership, row_scores, col_scores) get_row_density_scores(membership, row_scores) getting density scores improves small clusters get_rr_scores(membership, rowscores, bandwidth, cluster) calculate the density scores for the given row score values in the specified cluster make_db_column_seeder(outdb) make_db_row_seeder(outdb) make_file_column_seeder(filename, sep=' ') make_file_seeder(filename, sep=' ') uses a TSV file to seed row membership make_kmeans_row_seeder(num_clusters) creates a row seeding function based on k-means old_fuzzy_coefficient(iteration, num_iterations) standard fuzzy coefficient as defined in cMonkey postadjust(membership, rowscores, cutoff=0.33, limit=100) adjusting the cluster memberships after the main iterations have been done Returns true if the function changed the membership, false if not replace_delta_column_member(membership, col, cm, cd_scores) replace_delta_row_member(membership, row, rm, rd_scores) seeing_change(prob) returns true if the update is seeing the change std_fuzzy_coefficient(iteration, num_iterations) standard fuzzy coefficient as defined in cMonkey update_for_cols(membership, cd_scores, multiprocessing) updating column memberships according to cd_scores update_for_rows(membership, rd_scores, multiprocessing) generically updating row memberships according to  rd_scores

Data
		KEY_CLUSTERS_PER_COL = 'memb.clusters_per_col' KEY_CLUSTERS_PER_ROW = 'memb.clusters_per_row' KEY_COL_IS_MEMBER_OF = 'memb.col_is_member_of' KEY_MAX_CHANGES_PER_COL = 'memb.max_changes_per_col' KEY_MAX_CHANGES_PER_ROW = 'memb.max_changes_per_row' KEY_MAX_CLUSTER_ROWS_ALLOWED = 'memb.max_cluster_rows_allowed' KEY_MIN_CLUSTER_ROWS_ALLOWED = 'memb.min_cluster_rows_allowed' KEY_NUM_CLUSTERS = 'num_clusters' KEY_PROB_COL_CHANGE = 'memb.prob_col_change' KEY_PROB_ROW_CHANGE = 'memb.prob_row_change' KEY_ROW_IS_MEMBER_OF = 'memb.row_is_member_of' MAX_ADJUST_TRIES = 50 UPDATE_MEMBERSHIP = None