nest::WeightOptimizerCommonProperties Class Reference

Base class implementing common properties of a weight optimizer model. More...

#include <weight_optimizer.h>

Inheritance diagram for nest::WeightOptimizerCommonProperties:

Public Member Functions
	WeightOptimizerCommonProperties ()
	Default constructor.
virtual	~WeightOptimizerCommonProperties ()
	Destructor.
	WeightOptimizerCommonProperties (const WeightOptimizerCommonProperties &)
	Copy constructor.
WeightOptimizer &	operator= (const WeightOptimizer &)=delete
	Assignment operator.
virtual void	get_status (Dictionary &d) const
	Get parameter dictionary.
virtual void	set_status (const Dictionary &d)
	Update parameters in parameter dictionary.
virtual WeightOptimizerCommonProperties *	clone () const =0
	Clone constructor.
virtual WeightOptimizer *	get_optimizer () const =0
	Get optimizer.
double	get_Wmin () const
	Get minimal value for synaptic weight.
double	get_Wmax () const
	Get maximal value for synaptic weight.
virtual std::string	get_name () const =0
	Get optimizer name.

Public Attributes
size_t	batch_size_
	Size of an optimization batch.
double	eta_
	Common learning rate for all synapses.
double	eta_first_change_
	First non-default learning rate.
long	eta_change_count_
	Count of learning rate changes so far in the simulation to identify the first change.
double	Wmin_
	Minimal value for synaptic weight.
double	Wmax_
	Maximal value for synaptic weight.
bool	optimize_each_step_
	If true, optimize each step, else once per spike.

Detailed Description

Base class implementing common properties of a weight optimizer model.

The CommonProperties of synapse models supporting weight optimization own an object of this class hierarchy. The values in these objects are used by the synapse-specific optimizer object. Change of the optimizer type is only possible before synapses of the model have been created.

Constructor & Destructor Documentation

WeightOptimizerCommonProperties() [1/2]

nest::WeightOptimizerCommonProperties::WeightOptimizerCommonProperties

(

)

Default constructor.

~WeightOptimizerCommonProperties()

virtual nest::WeightOptimizerCommonProperties::~WeightOptimizerCommonProperties ( )

inlinevirtual

Destructor.

WeightOptimizerCommonProperties() [2/2]

nest::WeightOptimizerCommonProperties::WeightOptimizerCommonProperties

(

const WeightOptimizerCommonProperties &

cp

)

Copy constructor.

Member Function Documentation

clone()

virtual WeightOptimizerCommonProperties * nest::WeightOptimizerCommonProperties::clone ( ) const

pure virtual

Clone constructor.

Implemented in nest::WeightOptimizerCommonPropertiesGradientDescent, and nest::WeightOptimizerCommonPropertiesAdam.

get_name()

virtual std::string nest::WeightOptimizerCommonProperties::get_name ( ) const

pure virtual

Get optimizer name.

Implemented in nest::WeightOptimizerCommonPropertiesGradientDescent, and nest::WeightOptimizerCommonPropertiesAdam.

Referenced by nest::EpropSynapseCommonProperties::get_status(), get_status(), nest::EpropSynapseCommonProperties::set_status(), and nest::EpropSynapseBSSHSLM2020CommonProperties::set_status().

Here is the caller graph for this function:

get_optimizer()

virtual WeightOptimizer * nest::WeightOptimizerCommonProperties::get_optimizer ( ) const

pure virtual

Get optimizer.

Implemented in nest::WeightOptimizerCommonPropertiesGradientDescent, and nest::WeightOptimizerCommonPropertiesAdam.

Referenced by nest::eprop_synapse< targetidentifierT >::check_connection(), and nest::eprop_synapse_bsshslm_2020< targetidentifierT >::check_connection().

Here is the caller graph for this function:

get_status()

void nest::WeightOptimizerCommonProperties::get_status ( Dictionary &  d ) const

virtual

Get parameter dictionary.

Reimplemented in nest::WeightOptimizerCommonPropertiesAdam.

References nest::names::batch_size(), batch_size_, nest::names::eta(), eta_, get_name(), nest::names::optimize_each_step(), optimize_each_step_, nest::names::optimizer(), nest::names::Wmax(), Wmax_, nest::names::Wmin(), and Wmin_.

Referenced by nest::EpropSynapseCommonProperties::get_status(), nest::EpropSynapseBSSHSLM2020CommonProperties::get_status(), and nest::WeightOptimizerCommonPropertiesAdam::get_status().

Here is the call graph for this function:

Here is the caller graph for this function:

get_Wmax()

double nest::WeightOptimizerCommonProperties::get_Wmax ( ) const

inline

Get maximal value for synaptic weight.

References Wmax_.

get_Wmin()

double nest::WeightOptimizerCommonProperties::get_Wmin ( ) const

inline

Get minimal value for synaptic weight.

References Wmin_.

operator=()

WeightOptimizer & nest::WeightOptimizerCommonProperties::operator= ( const WeightOptimizer &  )

delete

Assignment operator.

set_status()

void nest::WeightOptimizerCommonProperties::set_status ( const Dictionary &  d )

virtual

Update parameters in parameter dictionary.

Reimplemented in nest::WeightOptimizerCommonPropertiesAdam.

References nest::names::batch_size(), batch_size_, nest::names::eta(), eta_, eta_change_count_, eta_first_change_, nest::names::optimize_each_step(), optimize_each_step_, nest::names::Wmax(), Wmax_, nest::names::Wmin(), and Wmin_.

Referenced by nest::WeightOptimizerCommonPropertiesAdam::set_status(), nest::EpropSynapseCommonProperties::set_status(), and nest::EpropSynapseBSSHSLM2020CommonProperties::set_status().

Here is the call graph for this function:

Here is the caller graph for this function:

Member Data Documentation

batch_size_

size_t nest::WeightOptimizerCommonProperties::batch_size_

Size of an optimization batch.

Referenced by get_status(), nest::WeightOptimizer::optimized_weight(), and set_status().

eta_

double nest::WeightOptimizerCommonProperties::eta_

Common learning rate for all synapses.

Referenced by get_status(), nest::WeightOptimizer::optimized_weight(), and set_status().

eta_change_count_

long nest::WeightOptimizerCommonProperties::eta_change_count_

Count of learning rate changes so far in the simulation to identify the first change.

Referenced by nest::WeightOptimizer::optimized_weight(), and set_status().

eta_first_change_

double nest::WeightOptimizerCommonProperties::eta_first_change_

First non-default learning rate.

Stores the first non-default learning rate, ensuring correct handling when the global learning rate (eta) is modified before the first optimization step. Once optimization begins, eta_current aligns with eta. Along with eta_change_count and eta_current, this enables simulations with distinct phases, such as training (nonzero learning rate) and validation, testing, or early stopping (zero learning rate).

Referenced by nest::WeightOptimizer::optimized_weight(), and set_status().

optimize_each_step_

bool nest::WeightOptimizerCommonProperties::optimize_each_step_

If true, optimize each step, else once per spike.

Referenced by get_status(), and set_status().

Wmax_

double nest::WeightOptimizerCommonProperties::Wmax_

Maximal value for synaptic weight.

Referenced by get_status(), get_Wmax(), nest::WeightOptimizer::optimized_weight(), and set_status().

Wmin_

double nest::WeightOptimizerCommonProperties::Wmin_

Minimal value for synaptic weight.

Referenced by get_status(), get_Wmin(), nest::WeightOptimizer::optimized_weight(), and set_status().

---

The documentation for this class was generated from the following files:

• weight_optimizer.h
• weight_optimizer.cpp