mass.thermo.conc_sampling.conc_optgp
Provides concentration sampling through an OptGP sampler.
Based on sampling implementations in cobra.sampling.optgp
Module Contents
Classes
A parallel optimized sampler. |
- class mass.thermo.conc_sampling.conc_optgp.ConcOptGPSampler(concentration_solver, processes=None, thinning=100, nproj=None, seed=None)[source]
Bases:
mass.thermo.conc_sampling.conc_hr_sampler.ConcHRSampler
A parallel optimized sampler.
A parallel sampler with fast convergence and parallel execution [MHM14].
Notes
The sampler is very similar to artificial centering where each process samples its own chain. The implementation used here is the similar as in the Python
cobra
package.Initial points are chosen randomly from the warmup points followed by a linear transformation that pulls the points a little bit towards the center of the sampling space.
If the number of processes used is larger than the one requested, number of samples is adjusted to the smallest multiple of the number of processes larger than the requested sample number. For instance, if you have 3 processes and request 8 samples you will receive 9.
Memory usage is roughly in the order of:
(number included reactions + number included metabolites)^2
due to the required nullspace matrices and warmup points. So large models easily take up a few GB of RAM. However, most of the large matrices are kept in shared memory. So the RAM usage is independent of the number of processes.
- Parameters
concentration_solver (ConcSolver) – The
ConcSolver
to use in generating samples.thinning (int) – The thinning factor for the generated sampling chain as a positive
int
> 0. A thinning factor of 10 means samples are returned every 10 steps.processes (int or None) –
The number of processes used to generate samples. If
None
the number of processes specified in theMassConfiguration
is utilized. Only valid formethod='optgp'
.Default is
None
.nproj (int or None) –
A positive
int
> 0 indicating how often to reporject the sampling point into the feasibility space. Avoids numerical issues at the cost of lower samplimg. IfNone
then the value is determined via the following:nproj = int(min(len(self.concentration_solver.variables)**3, 1e6))
Default is
None
seed (int or None) –
A positive
int
> 0 indiciating random number seed that should be used. IfNone
provided, the current time stamp is used.Default is
None
.
- concentration_solver
The
ConcSolver
used to generate samples.- Type
- retries
The overall of sampling retries the sampler has observed. Larger values indicate numerical instabilities.
- Type
- problem
A
namedtuple
whose attributes define the entire sampling problem in matrix form. See docstring ofProblem
for more information.- Type
collections.namedtuple
- warmup
A matrix of with as many columns as variables in the model of the
ConcSolver
and more than 3 rows containing a warmup sample in each row.None
if no warmup points have been generated yet.- Type
- sample(n, concs=True)[source]
Generate a set of samples.
This is the basic sampling function for all hit-and-run samplers.
Notes
Performance of this function linearly depends on the number of metabolites in your model and the thinning factor.
If the number of processes is larger than one, computation is split across as the CPUs of your machine. This may shorten computation time.
However, there is also overhead in setting up parallel computation so it is recommended to calculate large numbers of samples at once (
n
> 1000).- Parameters
n (int) – The number of samples that are generated at once.
concs (boolean) – Whether to return concentrations or the internal solver variables. If
False
will return a variable for each metabolite and reaction Keq as well as all additional variables that may have been defined in the model.
- Returns
A matrix with
n
rows, each containing a concentration sample.- Return type