Struct sampling::entropic_sampling::EntropicSampling

pub struct EntropicSampling<Hist, R, E, S, Res, Energy> { /* private fields */ }

Entropic sampling made easy

J. Lee, “New Monte Carlo algorithm: Entropic sampling,” Phys. Rev. Lett. 71, 211–214 (1993), DOI: 10.1103/PhysRevLett.71.211

Implementations

impl<Hist, R, E, S, Res, T> EntropicSampling<Hist, R, E, S, Res, T>

pub fn ensemble(&self) -> &E

Current state of the Ensemble

pub fn energy(&self) -> &T

Energy of ensemble

• assuming energy_fn (see self.entropic_step etc.) is deterministic and will allways give the same result for the same ensemble, this returns the energy of the current ensemble

pub fn set_step_goal(&mut self, step_goal: usize)

Number of entropic steps to be performed

• set the number of steps to be performed by entropic sampling

pub fn step_size(&self) -> usize

Smallest possible markov step (m_steps of MarkovChain trait) by entropic step

pub fn fraction_accepted_total(&self) -> f64

Fraction of steps accepted since the creation of self

• total_steps_accepted / total_steps
• NaN if no steps were performed yet

pub fn log_density_estimate(&self) -> &Vec<f64>

• returns the (non normalized) log_density estimate log(P(E)), with which the simulation was started
• if you created this from a WangLandau simulation, this is the result of the WangLandau Simulation

pub fn hist(&self) -> &Hist

Return current state of histogram

impl<Hist, R, E, S, Res, T> EntropicSampling<Hist, R, E, S, Res, T>

where Hist: Histogram,

pub fn log_density_refined(&self) -> Vec<f64>

calculates the (non normalized) log_density estimate log(P(E)) according to the paper

pub fn refine_estimate(&mut self) -> Vec<f64>

Calculates self.log_density_refined and uses that as estimate for a the entropic sampling simulation

• returns old estimate

prepares self for a new entropic simulation

• sets new estimate for log(P(E))
• resets statistic gathering
• resets step_count

impl<Hist, R, E, S, Res, T> EntropicSampling<Hist, R, E, S, Res, T>

where Hist: Histogram, R: Rng,

pub fn from_wl( wl: WangLandau1T<Hist, R, E, S, Res, T> ) -> Result<Self, EntropicErrors>

Creates Entropic from a WangLandauAdaptive state

• WangLandauAdaptive state needs to be valid, i.e., you must have called one of the init* methods

• this ensures, that the members old_energy and old_bin are not None

pub fn from_wl_adaptive( wl: WangLandauAdaptive<Hist, R, E, S, Res, T> ) -> Result<Self, EntropicErrors>

Creates Entropic from a WangLandauAdaptive state

• WangLandauAdaptive state needs to be valid, i.e., you must have called one of the init* methods

• this ensures, that the members old_energy and old_bin are not None

impl<Hist, R, E, S, Res, T> EntropicSampling<Hist, R, E, S, Res, T>

where Hist: Histogram + HistogramVal<T>, R: Rng, E: MarkovChain<S, Res>, T: Clone,

pub fn entropic_sampling_while<F, G, W>( &mut self, energy_fn: F, print_fn: G, condition: W )

where F: FnMut(&E) -> Option<T>, G: FnMut(&Self), W: FnMut(&Self) -> bool,

Entropic sampling

• performs self.entropic_step(energy_fn) until condition is false
• Note: you have access to the current step_count (self.step_count())

Parameter

• energy_fn function calculating Some(energy) of the system or rather the Parameter of which you wish to obtain the probability distribution. If there are any states, for which the calculation is invalid, None should be returned
• steps resulting in ensembles for which energy_fn(&mut ensemble) is None will always be rejected
• Important energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!
• print_fn: see below

Correlations

• if you want to measure correlations between “energy” and other measurable quantities, use print_fn, which will be called after each step - use this function to write to a file or whatever you desire
• Note: You do not have to recalculate the energy, if you need it in print_fn: just call self.energy()
• you have access to your ensemble with self.ensemble()
• if you do not need it, you can use |_|{} as print_fn

pub fn entropic_sampling_while_acc<F, G, W>( &mut self, energy_fn: F, print_fn: G, condition: W )

where F: FnMut(&E, &S, &mut T), G: FnMut(&Self), W: FnMut(&Self) -> bool,

Entropic sampling using an accumulating markov step

• performs self.entropic_step_acc(&mut energy_fn) until condition(self) == false

Parameter

• energy_fn function calculating the energy E of the system (or rather the Parameter of which you wish to obtain the probability distribution) during the markov steps, which can be more efficient.
• Important energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!
• print_fn: see below

Correlations

• if you want to measure correlations between “energy” and other measurable quantities, use print_fn, which will be called after each step - use this function to write to a file or whatever you desire
• Note: You do not have to recalculate the energy, if you need it in print_fn: just call self.energy()
• you have access to your ensemble with self.ensemble()
• if you do not need it, you can use |_|{} as print_fn

pub unsafe fn entropic_sampling_while_unsafe<F, G, W>( &mut self, energy_fn: F, print_fn: G, condition: W )

where F: FnMut(&mut E) -> Option<T>, G: FnMut(&mut Self), W: FnMut(&mut Self) -> bool,

Entropic sampling

• if possible, use entropic_sampling_while instead, as it is safer

Safety

• use this if you need mutable access to your ensemble while printing or calculating the condition. Note, that whatever you do there, should not change the energy of the current state. Otherwise this can lead to undefined behavior and the results of the entropic sampling cannot be trusted anymore!
• performs self.entropic_step(energy_fn) until condition is false
• Note: you have access to the current step_count (self.step_count())

Parameter

• energy_fn function calculating Some(energy) of the system or rather the Parameter of which you wish to obtain the probability distribution. If there are any states, for which the calculation is invalid, None should be returned
• steps resulting in ensembles for which energy_fn(&mut ensemble) is None will always be rejected
• Important energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!
• print_fn: see below

Correlations

• if you want to measure correlations between “energy” and other measurable quantities, use print_fn, which will be called after each step - use this function to write to a file or whatever you desire
• Note: You do not have to recalculate the energy, if you need it in print_fn: just call self.energy()
• you have mutable access to your ensemble with self.ensemble_mut()
• if you do not need it, you can use |_|{} as print_fn

pub fn entropic_sampling<F, G>(&mut self, energy_fn: F, print_fn: G)

where F: FnMut(&E) -> Option<T>, G: FnMut(&Self),

Entropic sampling

• performs self.entropic_step(energy_fn) until self.step_count == self.step_goal

Parameter

• energy_fn function calculating Some(energy) of the system or rather the Parameter of which you wish to obtain the probability distribution. If there are any states, for which the calculation is invalid, None should be returned
• steps resulting in ensembles for which energy_fn(&mut ensemble) is None will always be rejected
• Important energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!
• print_fn: see below

Correlations

• if you want to measure correlations between “energy” and other measurable quantities, use print_fn, which will be called after each step - use this function to write to a file or whatever you desire
• Note: You do not have to recalculate the energy, if you need it in print_fn: just call self.energy()
• you have access to your ensemble with self.ensemble()
• if you do not need it, you can use |_|{} as print_fn

pub fn entropic_sampling_acc<F, G>(&mut self, energy_fn: F, print_fn: G)

where F: FnMut(&E, &S, &mut T), G: FnMut(&Self),

Entropic sampling using an accumulating markov step

• performs self.entropic_step_acc(&mut energy_fn) until self.step_count >= self.step_goal

Parameter

• energy_fn function calculating the energy E of the system (or rather the Parameter of which you wish to obtain the probability distribution) during the markov steps, which can be more efficient.
• Important energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!
• print_fn: see below

Correlations

• if you want to measure correlations between “energy” and other measurable quantities, use print_fn, which will be called after each step - use this function to write to a file or whatever you desire
• Note: You do not have to recalculate the energy, if you need it in print_fn: just call self.energy()
• you have access to your ensemble with self.ensemble()
• if you do not need it, you can use |_|{} as print_fn

pub unsafe fn entropic_sampling_unsafe<F, G>( &mut self, energy_fn: F, print_fn: G )

where F: FnMut(&mut E) -> Option<T>, G: FnMut(&mut Self),

Entropic sampling

• if possible, use entropic_sampling instead, as it is safer

Safety

• NOTE You have mutable access to your ensemble (and to self, at least in the printing function). This makes this function unsafe. You should never change your ensemble in a way, that will effect the outcome of the energy function. Otherwise the results will just be wrong. This is intended for usecases, where the energycalculation is more efficient with mutable access, e.g., through using a buffer stored in the ensemble
• performs self.entropic_step(energy_fn) until self.step_count == self.step_goal

Parameter

• energy_fn function calculating Some(energy) of the system or rather the Parameter of which you wish to obtain the probability distribution. If there are any states, for which the calculation is invalid, None should be returned
• steps resulting in ensembles for which energy_fn(&mut ensemble) is None will always be rejected
• Important energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!
• print_fn: see below

Correlations

• if you want to measure correlations between “energy” and other measurable quantities, use print_fn, which will be called after each step - use this function to write to a file or whatever you desire
• Note: You do not have to recalculate the energy, if you need it in print_fn: just call self.energy()
• you have access to your ensemble with self.ensemble()
• if you do not need it, you can use |_|{} as print_fn

pub unsafe fn entropic_step_unsafe<F>(&mut self, energy_fn: F)

where F: FnMut(&mut E) -> Option<T>,

Entropic step

• if possible, use entropic_step instead
• performs a single step

Parameter

• energy_fn function calculating Some(energy) of the system or rather the Parameter of which you wish to obtain the probability distribution. If there are any states, for which the calculation is invalid, None should be returned
• steps resulting in ensembles for which energy_fn(&mut ensemble) is None will always be rejected

Important

• energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!

Safety

• While you do have mutable access to the ensemble, the energy function should not change the ensemble in a way, which affects the next calculation of the energy
• This is intended for usecases, where the energycalculation is more efficient with mutable access, e.g., through using a buffer stored in the ensemble

pub fn entropic_step<F>(&mut self, energy_fn: F)

where F: FnMut(&E) -> Option<T>,

Entropic step

• performs a single step

Parameter

• energy_fn function calculating Some(energy) of the system or rather the Parameter of which you wish to obtain the probability distribution. If there are any states, for which the calculation is invalid, None should be returned
• steps resulting in ensembles for which energy_fn(&mut ensemble) is None will always be rejected

Important

• energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!

pub fn entropic_step_acc<F>(&mut self, energy_fn: F)

where F: FnMut(&E, &S, &mut T),

Entropic sampling using an accumulating markov step

• performs self.entropic_step_acc(&mut energy_fn) until self.step_count == self.step_goal

Parameter

• energy_fn function calculating the energy E of the system (or rather the Parameter of which you wish to obtain the probability distribution) during the markov steps, which can be more efficient.
• Important energy_fn: should be the same as used for Wang Landau, otherwise the results will be wrong!
• print_fn: see below

Correlations

• if you want to measure correlations between “energy” and other measurable quantities, use print_fn, which will be called after each step - use this function to write to a file or whatever you desire
• Note: You do not have to recalculate the energy, if you need it in print_fn: just call self.energy()
• you have access to your ensemble with self.ensemble()
• if you do not need it, you can use |_|{} as print_fn

Trait Implementations

impl<Hist: Clone, R: Clone, E: Clone, S: Clone, Res: Clone, Energy: Clone> Clone for EntropicSampling<Hist, R, E, S, Res, Energy>

fn clone(&self) -> EntropicSampling<Hist, R, E, S, Res, Energy>

Returns a copy of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<Hist: Debug, R: Debug, E: Debug, S: Debug, Res: Debug, Energy: Debug> Debug for EntropicSampling<Hist, R, E, S, Res, Energy>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de, Hist, R, E, S, Res, Energy> Deserialize<'de> for EntropicSampling<Hist, R, E, S, Res, Energy>

where Hist: Deserialize<'de>, R: Deserialize<'de>, E: Deserialize<'de>, S: Deserialize<'de>, Energy: Deserialize<'de>,

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl<Hist, R, E, S, Res, T> Entropic for EntropicSampling<Hist, R, E, S, Res, T>

where Hist: Histogram, R: Rng,

fn step_counter(&self) -> usize

Number of entropic steps done until now

• will be reset by self.refine_estimate

fn step_goal(&self) -> usize

Number of entropic steps to be performed

• if self was created from WangLandauAdaptive, step_goal will be equal to the number of WangLandau steps, that were performed

fn log_density(&self) -> Vec<f64>

Current (non normalized) estimate of ln(P(E))

fn write_log<W: Write>(&self, w: W) -> Result<(), Error>

Writes Information about the simulation to a file. E.g. How many steps were performed.

fn total_steps_accepted(&self) -> usize

How many steps were accepted until now?

fn total_steps_rejected(&self) -> usize

How many steps were rejected until now?

fn steps_total(&self) -> usize

Counter

fn fraction_accepted_total(&self) -> f64

Calculate, which fraction of steps were accepted

fn fraction_rejected_total(&self) -> f64

Calculate, which fraction of steps were rejected

fn is_finished(&self) -> bool

Checks wang landau threshold

fn log_density_base10(&self) -> Vec<f64>

Current (non normalized) estimate of log10(P(E))

fn log_density_base(&self, base: f64) -> Vec<f64>

Current (non normalized) estimate of log_base(P(E))

impl<Hist, R, E, S, Res, Energy> EntropicEnergy<Energy> for EntropicSampling<Hist, R, E, S, Res, Energy>

where Hist: Histogram, R: Rng,

fn energy(&self) -> &Energy

Energy of ensemble

• assuming energy_fn (see self.entropic_step etc.) is deterministic and will allways give the same result for the same ensemble, this returns the energy of the current ensemble

impl<Hist, R, E, S, Res, Energy> EntropicEnsemble<E> for EntropicSampling<Hist, R, E, S, Res, Energy>

where Hist: Histogram, R: Rng,

fn ensemble(&self) -> &E

return reference to current state of ensemble

unsafe fn ensemble_mut(&mut self) -> &mut E

returns mutable reference to ensemble

impl<Hist, R, E, S, Res, Energy> EntropicHist<Hist> for EntropicSampling<Hist, R, E, S, Res, Energy>

where Hist: Histogram, R: Rng,

fn hist(&self) -> &Hist

returns current histogram

impl<Hist, R, E, S, Res, Energy> GlueAble<Hist> for EntropicSampling<Hist, R, E, S, Res, Energy>

where Hist: Clone + Histogram,

fn push_glue_entry_ignoring( &self, job: &mut GlueJob<Hist>, ignore_idx: &[usize] )

fn push_glue_entry(&self, job: &mut GlueJob<H>)

impl<Hist, R, E, S, Res, Energy> HasRng<R> for EntropicSampling<Hist, R, E, S, Res, Energy>

where R: Rng,

fn rng(&mut self) -> &mut R

Access RNG

fn swap_rng(&mut self, rng: &mut R)

If you need to exchange the internal rng

impl<Hist, R, E, S, Res, Energy> Serialize for EntropicSampling<Hist, R, E, S, Res, Energy>

where Hist: Serialize, R: Serialize, E: Serialize, S: Serialize, Energy: Serialize,

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl<Hist, R, E, S, Res, Energy> TryFrom<WangLandau1T<Hist, R, E, S, Res, Energy>> for EntropicSampling<Hist, R, E, S, Res, Energy>

where Hist: Histogram, R: Rng,

type Error = EntropicErrors

The type returned in the event of a conversion error.

fn try_from( wl: WangLandau1T<Hist, R, E, S, Res, Energy> ) -> Result<Self, Self::Error>

Performs the conversion.

impl<Hist, R, E, S, Res, T> TryFrom<WangLandauAdaptive<Hist, R, E, S, Res, T>> for EntropicSampling<Hist, R, E, S, Res, T>

where Hist: Histogram, R: Rng,

fn try_from( wl: WangLandauAdaptive<Hist, R, E, S, Res, T> ) -> Result<Self, Self::Error>

Uses as stepsize: first entry of bestof. If bestof is empty, it uses wl.min_step_size() + (wl.max_step_size() - wl.max_step_size()) / 2

type Error = EntropicErrors

The type returned in the event of a conversion error.