cbp.builder¶

cbp.builder.base_builder¶

class cbp.builder.base_builder.BaseBuilder(dim, policy, rand_seed=1)[source]¶

Bases: abc.ABC

add_branch(head_node=None, is_constrained=False, prob=None, is_conv=False)[source]¶

add_constrained_node(probability=None)[source]¶

add_factor(name_list, is_conv=False)[source]¶

add_factor_different(name_list, is_conv=False)[source]¶

add_trivial_node(dim=None)[source]¶

abstract init_graph()[source]¶

cbp.builder.hmm_builder¶

class cbp.builder.hmm_builder.HMMBuilder(length, node_dim, policy, rand_seed=1)[source]¶

Bases: cbp.builder.base_builder.BaseBuilder

init_graph()[source]¶

step(time_stamp=None)[source]¶

hmm go forward a step, time + 1

Parameters: time_stamp (int) – timer for the current add node

cbp.builder.hmm_zero_builder¶

class cbp.builder.hmm_zero_builder.HMMZeroBuilder(length, d, policy, rand_seed=1)[source]¶

Bases: cbp.builder.hmm_builder.HMMBuilder

add_constrained_node(probability=None)[source]¶

add_factor(name_list, is_conv=False)[source]¶

cbp.builder.hmm_sim_builder¶

class cbp.builder.hmm_sim_builder.HMMSimBuilder(length, simulator, policy, random_seed=1)[source]¶

Bases: cbp.builder.hmm_builder.HMMBuilder

add_branch(head_node=None, is_constrained=False, prob=None, is_conv=False)[source]¶

add_constrained_node(probability=None)[source]¶

add_factor(name_list, is_conv=False)[source]¶

add factor to hmm graph

Parameters

name_list (list) – connected node
is_conv (bool, optional) – is emit or not , defaults to False– transition

Returns

FactorNode

Return type

cbp.FactorNode

example(num_sample)[source]¶

fix_initpotential(potential=None)[source]¶

cbp.builder.hmm_simulator¶

class cbp.builder.hmm_simulator.HMMSimDate(time_step, state_num, obser_num, *args, **kwargs)[source]¶

Bases: dict

cal_theory(verbose)[source]¶

property sensor¶

property traj¶

class cbp.builder.hmm_simulator.HMMSimulator(time_step, dim_states, dim_observations, random_seed, is_theorem=False)[source]¶

Bases: object

get_emission_potential()[source]¶

get_fix_margin(time_step=None)[source]¶

return observation marginal

Parameters: time_step – if int then return a specific time distribution

otherwise all distributions as matrxi, defaults to None :type time_step: int, optional :return: array for single time_step or a matrix :rtype: ndarray

get_gt_joint()[source]¶

get_hidden_margin(time_step=None)[source]¶

return ground truth marginal

Parameters: time_step – if int then return a specific time distribution

otherwise all distributions as matrix, defaults to None :type time_step: int, optional :return: array for single time_step or a matrix :rtype: ndarray

get_init_potential()[source]¶

get_precious(time_step=None, verbose=False)[source]¶

return precious marginal

Parameters: time_step – if int then return a specific time distribution

otherwise all distributions as matrix, defaults to None :type time_step: int, optional :param verbose: whether or not ouput difference between simulation and

theorical margin

Returns: array for single time_step or a matrix
Return type: ndarray

get_sensor()[source]¶

get_traj()[source]¶

get_transition_potential()[source]¶

classmethod load(sim_name: str)[source]¶

load a simulator from pkl file

Parameters: sim_name (str) – name for the simulator
Returns: simulator
Return type: simulator

property name¶

observe(state)[source]¶

do a emission sample

Parameters: state (int) – current states
Returns: the observation of cur states
Return type: int

observe_traj(traj)[source]¶

register_potential(ptype, potential)[source]¶

register potential for simulation

Parameters

ptype (PotentialType) – potential type
potential (ndarray) – [description]

reset()[source]¶

sample(num_sample)[source]¶

sample_engine(state, dim, potential)[source]¶

sample according to conditional prob table

Parameters

state (int) – cur_state
dim (int) – range of next state or observation
potential (ndarray) – conditional prob table

Returns

next state or observation

Return type

int

save()[source]¶

step(state)[source]¶

do a states transition sample

Parameters: state (int) – current states
Returns: the next states
Return type: int

viz_emission_potential()[source]¶

viz_gt()[source]¶

viz_sensor()[source]¶

viz_trans_potential()[source]¶

class cbp.builder.hmm_simulator.PotentialType(value)[source]¶

Bases: str, enum.Enum

Potential Type Enum

EMISSION = 'Emission'¶

INIT = 'Init'¶

TRANSITION = 'Transition'¶

cbp.builder.hmm_simulator.cal_single(traj, col_num, state_num)[source]¶

cbp.builder.line_builder¶

class cbp.builder.line_builder.LineBuilder(num_node, d, policy, rand_seed=1)[source]¶

Bases: cbp.builder.base_builder.BaseBuilder

init_graph()[source]¶

cbp.builder.migr_simulator¶

class cbp.builder.migr_simulator.MigrSimulator(time_step, d_col, d_row, random_seed)[source]¶

Bases: cbp.builder.hmm_simulator.HMMSimulator

compile()[source]¶: register various potential for simulation

viz_emission_potential()[source]¶

viz_estm(estimated_marginal)[source]¶

viz_gt()[source]¶

viz_sensor()[source]¶

viz_trans_potential()[source]¶

cbp.builder.migr_visualizer¶

class cbp.builder.migr_visualizer.MigrVisualizer(d_row, d_col)[source]¶

Bases: object

ind2rowcol(index)[source]¶

migration(data, **kwargs)[source]¶

draw migration figure, heat map distribution 3 reservered key in kwargs: * fig_name: savefig * xlabel: xlabel * ylabel

Parameters: data – [i,j] record i-th particle position at j timestamp

potential_heatmap(data, **kwargs)[source]¶

plot potential heatmap and save png. 2 reserved keys:

title figure title
path figure path prefix

Parameters: data (ndarray) – i-th row represents the potential of d state

visualize_location(xx, yy, xy_size, **kwargs)[source]¶

plot grid location distribution. Origin point is in DownLeft. 3 reservered key in kwargs: * fig_name: savefig * xlabel: xlabel * ylabel

Parameters

xx (ndarray) – scatter plot x
yy (ndarray) – scatter plot y
xy_size (ndarray) – at position (x,y) the number of particles

visualize_map_bins(bins, **kwargs)[source]¶

converts the statistics bins data to the map figure. 3 reservered key in kwargs: * fig_name: savefig * xlabel: xlabel * ylabel

Parameters: bins (list or ndarray) – every element represents how many particles in the cell

cbp.builder.potential_utils¶

cbp.builder.potential_utils.diagonal_potential(d_1: int, d_2: int, rng: numpy.random.mtrand.RandomState) → numpy.ndarray[source]¶

cbp.builder.potential_utils.diagonal_potential_conv(d_1: int, d_2: int, rng: numpy.random.mtrand.RandomState) → numpy.ndarray[source]¶

cbp.builder.potential_utils.diagonal_potential_different(d_1: int, d_2: int, rng: numpy.random.mtrand.RandomState) → numpy.ndarray[source]¶

cbp.builder.potential_utils.identity_potential(d_1: int, d_2: int, rng: numpy.random.mtrand.RandomState) → numpy.ndarray[source]¶

cbp.builder.star_builder¶

class cbp.builder.star_builder.StarBuilder(num_node, d, policy, rand_seed)[source]¶

Bases: cbp.builder.base_builder.BaseBuilder

init_graph()[source]¶

cbp.builder.tree_builder¶

class cbp.builder.tree_builder.TreeBuilder(tree_depth, d, policy, rand_seed=1)[source]¶

Bases: cbp.builder.base_builder.BaseBuilder

init_graph()[source]¶

cbp.builder.wifi_hmm_builder¶

class cbp.builder.wifi_hmm_builder.WifiHMMBuilder(length, grid_d, policy, rand_seed=1, num_sensor=16, time_step=60)[source]¶

Bases: cbp.builder.hmm_sim_builder.HMMSimBuilder

builder with wifi-type simulator. Sparse observation. pass the wifi-simulator to the cbp.builder.HMMSimBuilder

cbp.builder.wifi_simulator¶

class cbp.builder.wifi_simulator.WifiSimulator(time_step, d_col, d_row, random_seed)[source]¶

Bases: cbp.builder.migr_simulator.MigrSimulator

simulate ensemble move, sensor is sparse - place sensors in some place - draw various data

compile()[source]¶: register various potential for simulation

draw_sensor()[source]¶: draw wifi sensor position in grid

fixed_sensors(space_d=1)[source]¶

random_sensor(num)[source]¶

register_hotspot(row, col)[source]¶

viz_sensor()[source]¶: draw raw sensor data