binf.samplers package

Submodules

binf.samplers.gibbs module

Gibbs sampler implementations

class binf.samplers.gibbs.GibbsSampler(pdf, state, subsamplers)

Bases: csb.statistics.samplers.mc.singlechain.AbstractSingleChainMC

_abc_cache = <_weakrefset.WeakSet object>

_abc_negative_cache = <_weakrefset.WeakSet object>

_abc_negative_cache_version = 36

_abc_registry = <_weakrefset.WeakSet object>

_accept_proposal(proposal_state)

Accept the proposal state by setting it as the current state of the sampler object

@param proposal_state: The proposal state @type proposal_state: L{State}

_calc_pacc()

Not applicable

_checkstate(state)

Checks whether the state is a dict

_propose()

Not applicable

_setup_conditional_pdfs()

Sets up the conditional PDFs from the full PDF

_update_conditional_pdf_params()

Updates parameters of the conditional PDFs to values set in this object’s PDF

_update_state(**variables)

Updates this object’s state’s variables to new values

_update_statistics(accepted)

Update the sampling statistics.

@param accepted: Whether or not the proposal state has been accepted @type accepted: boolean

_update_subsampler_states()

Updates subsampler states, e.g., after a replica exchange swap

acceptance_rate

Acceptance rate.

energy

Negative log-likelihood of the current state. @rtype: float

last_draw_stats

Returns information about most recent move for each subsampler

Returns:information about most recent move for each subsampler Return type:dict

last_move_accepted

Information whether the last MC move was accepted or not.

pdf

Returns the PDF object this sampler samples from

Returns:PDF object Return type:AbstractBinfPDF

sample()

Performs one iteration of the Gibbs sampling scheme

Returns:a state Return type:BinfState

sampling_stats

Sampling statistics, consisting of sampler statistics of the subsamplers

Returns:all subsampler sampling statistics Return type:OrderedDict

state

Returns the current state of this sampler

Returns:current state Return type:BinfState

subsamplers

Returns the subsamplers the Gibbs sampler iterates over

Returns:subsamplers for each variable Return type:dict

temperature

update_samplers(**samplers)

Updates the subsamplers

binf.samplers.hmc module

HMC sampler implementations

class binf.samplers.hmc.HMCSampleStats(accepted, stepsize)

Bases: tuple

_asdict()

Return a new OrderedDict which maps field names to their values

_fields = ('accepted', 'stepsize')

classmethod _make(iterable, new=<built-in method __new__ of type object>, len=<built-in function len>)

Make a new HMCSampleStats object from a sequence or iterable

_replace(**kwds)

Return a new HMCSampleStats object replacing specified fields with new values

accepted

Alias for field number 0

count(value) → integer -- return number of occurrences of value

index(value[, start[, stop]]) → integer -- return first index of value.

Raises ValueError if the value is not present.

stepsize

Alias for field number 1

class binf.samplers.hmc.HMCSampler(pdf, state, timestep, nsteps, timestep_adaption_limit=0, adaption_uprate=1.05, adaption_downrate=0.95, variable_name=None)

Bases: object

_adapt_timestep()

Increases / decreasese the leap frog time step depending on whether the last move has been rejected / accepted.

_copy_state(state)

Copies a state

Parameters:state (numpy.ndarray) – variable value to copy

_leapfrog(q, p, timestep, nsteps)

Performs leap frog integration of Hamiltonian dynamics guided by the gradient of the negative log-probability

Parameters:

• q (numpy.ndarray) – initial ‘position’
• p (numpy.ndarray) – initial ‘momentum’
• timestep (float) – integration time step
• nsteps (int) – # of integration steps

Returns:

‘position’ and ‘momentum’ at the end of the approximated MD trajectory

Return type:

(numpy.ndarray, numpy.ndarray)

acceptance_rate

last_draw_stats

Returns information about the most recently performed move

This is usually used by a replica exchange scheme to log sampling statistics.

Returns:whether the last move has been accepted and the current time step in the shape of a named tuple in a dictionary. This contrived is needed for Gibbs sampling / replica exchange statistics. Return type:dict

last_move_accepted

Returns whether the last move has been accepted or not

Returns:whether the last move has been accepted or not Return type:bool

sample()

Draws a random sample

Returns:a sample Return type:numpy.ndarray

variable_name

Returns the name of the variable this sampler is supposed to draw random samples from

Returns:variable name Return type:str

Module contents

This module contains implementations of various MCMC samplers

class binf.samplers.BinfState(variables={}, momenta={})

Bases: object

momenta

I currently don’t use this

update_momenta(**momenta)

I currently don’t use this

update_variables(**variables)

Updates variables

Parameters:**variables (dict) – named and typed variables to update this state with

variables

Returns a copy (!) of the variables and their values held by this state

Returns:copy of variables Return type:dict