auryn::CorrelatedPoissonGroup Class Reference

A PoissonGroup with multiple subpopulations that co-modulate their firing rate according to an Ornstein Uhlenbeck process. More...

#include <CorrelatedPoissonGroup.h>

Inheritance diagram for auryn::CorrelatedPoissonGroup:

Collaboration diagram for auryn::CorrelatedPoissonGroup:

Public Member Functions
	CorrelatedPoissonGroup (NeuronID n, AurynDouble rate=5., NeuronID gsize=100, AurynDouble timedelay=50e-3)
virtual	~CorrelatedPoissonGroup ()
virtual void	evolve ()
	Virtual pure evolve function which needs to be implemented by derived classes. More...
void	set_rate (AurynDouble rate)
void	set_amplitude (AurynDouble ampl)
void	set_target_amplitude (AurynDouble ampl)
void	set_tau_amplitude (AurynDouble scale)
void	set_timescale (AurynDouble scale)
void	set_offset (int off)
void	set_threshold (AurynDouble threshold)
void	set_stoptime (AurynDouble stoptime)
AurynDouble	get_rate ()
void	seed ()
Public Member Functions inherited from auryn::SpikingGroup
void	add_state_vector (std::string key, AurynStateVector *state_vector)
	Adds a state vector passed as an argument to the dictinary. More...
void	remove_state_vector (std::string key)
	Removes a state vector passed as an argument to the dictinary. More...
AurynStateVector *	find_state_vector (std::string key)
	Returns existing state vector by name or NULL if it does not exist. More...
AurynStateVector *	get_state_vector (std::string key)
	Creates a new or returns an existing state vector by name. More...
AurynStateVector *	get_existing_state_vector (std::string key)
	Returns an existing state with the supplied name. More...
AurynStateVector *	create_state_vector (std::string key)
	Creates a new state vector and throws an exception if a vector with the same name exists. More...
AurynState *	get_state_variable (std::string key)
	Creates a new group-wide state variable or returns an existing group-wide variable by name then returns a pointer to it. More...
void	randomize_state_vector_gauss (std::string state_vector_name, AurynState mean, AurynState sigma, int seed=12239)
	Randomizes the content of a state vector with Gaussian random numbers. Seeding is MPI save. More...
	SpikingGroup (NeuronID size, NodeDistributionMode mode=AUTO)
	Default constructor. More...
virtual	~SpikingGroup ()
	Default destructor. More...
virtual void	evolve_traces ()
	Evolves traces. More...
void	set_name (std::string s)
	Set connection name. More...
std::string	get_name ()
	Retrieves the groups name. More...
std::string	get_file_name ()
	Extracts the class name of the connection from the file name. More...
std::string	get_log_name ()
	Returns a string which is the combination of file and connection name for logging. More...
void	inc_num_spike_attributes (int x)
	Instructs SpikingGroup to increase the number of spike attributes by x. More...
int	get_num_spike_attributes ()
void	conditional_evolve ()
	Conditional evolve functino which is called by System. More...
unsigned int	get_locked_rank ()
	Returns locked rank for SpikingGroups which are not distributed across all ranks. More...
unsigned int	get_locked_range ()
	Returns locked range of ranks for SpikingGroups which are not distributed across all ranks. More...
SpikeContainer *	get_spikes ()
	Returns pointer to a spike container that contains spikes which arrive in this timestep from all neurons in this group. More...
SpikeContainer *	get_spikes_immediate ()
	Returns pointer to SpikeContainer of spikes generated during the last evolve() step. More...
AttributeContainer *	get_attributes ()
	Returns pointer to Attributecontainer for usage in propagating Connection objects. Same as get_spikes_immediate(), however might be overwritten to contain Spikes that have been delayed. More...
AttributeContainer *	get_attributes_immediate ()
	Returns pointer to Attributecontainer of spikes generated during the last evolve() step. More...
NeuronID	get_size ()
	Returns the size of the group. More...
NeuronID	get_pre_size ()
	Returns the size of the group. More...
NeuronID	calculate_rank_size (int rank=-1)
	Determines rank size and stores it in local variable. More...
NeuronID	get_rank_size ()
	Returns the size on this rank. More...
NeuronID	get_post_size ()
	Returns the size on this rank. More...
void	set_clock_ptr (AurynTime *clock)
bool	evolve_locally ()
	Returns true if the calling instance has units which are integrated on the current rank. More...
NeuronID	get_uid ()
	Get the unique ID of the class. More...
Trace *	get_pre_trace (AurynFloat x)
	Returns a pre trace with time constant x. More...
void	add_pre_trace (Trace *tr)
	Adds trace to pretrace stack of a connection. More...
Trace *	get_post_trace (AurynFloat x)
	Returns a post trace with time constant x. More...
void	add_post_trace (Trace *tr)
	Adds trace to posttrace stack of a connection. More...
void	push_spike (NeuronID spike)
	Pushes a local NeuronID as spike into the axonal SpikeDelay buffer. More...
void	push_attribute (AurynFloat attrib)
	Pushes a spike attribute into the axonal SpikeDelay buffer. More...
void	clear_spikes ()
	Clears all spikes stored in the delays which is useful to reset a network during runtime. More...
Trace *	get_post_state_trace (std::string state_name="mem", AurynFloat tau=10e-3, AurynFloat b=0.0)
	Returns a post trace of a neuronal state variable e.g. the membrane potential with time constant tau. More...
Trace *	get_post_state_trace (AurynStateVector *state, AurynFloat tau=10e-3, AurynFloat b=0.0)
	Returns a post trace of a neuronal state variable specified by pointer. More...
void	set_delay (int d)
	Sets axonal delay for this SpikingGroup. More...
virtual bool	write_to_file (const char *filename)
	Writes current states of SpikingGroup to human-readible textfile if implemented in derived class. More...
virtual bool	load_from_file (const char *filename)
	Reads current states of SpikingGroup to human-readible textfile if implemented in derived class. More...
NeuronID	ranksize ()
	Returns size (num of neurons) on the current rank. More...
NeuronID	global2rank (NeuronID i)
	Converts global NeuronID within the SpikingGroup to the local NeuronID on this rank. More...
NeuronID	rank2global (NeuronID i)
	Converts local NeuronID from the local rank to a global NeuronID. More...
bool	localrank (NeuronID i)
	Checks if the global NeuronID i is integrated on this MPI rank. More...
NeuronID	get_vector_size ()
	Rank size but rounded up to multiples of 4 (or potentially some other and larger number in future versions) for SSE compatibility. More...

Public Attributes
AurynDouble	mean
Public Attributes inherited from auryn::SpikingGroup
SpikeDelay *	delay
bool	active
	Toggles group active. More...
std::map< std::string, AurynStateVector * >	state_vectors
std::map< std::string, AurynState >	state_variables

Protected Attributes
AurynDouble	lambda
AurynTime	tstop
NeuronID	groupsize
NeuronID	remainersize
NeuronID	ngroups
AurynDouble	timescale
int	offset
AurynTime	delay
AurynDouble	o
AurynDouble *	delay_o
AurynDouble	amplitude
AurynDouble	tau_amplitude
AurynDouble	target_amplitude
AurynDouble	thr
NeuronID *	x
Protected Attributes inherited from auryn::SpikingGroup
std::vector< Trace * >	pretraces
	Pretraces. More...
std::vector< Trace * >	posttraces
	Posttraces. More...
std::vector< Trace * >	post_state_traces
	Post state traces. More...
std::vector< AurynFloat >	post_state_traces_spike_biases
std::vector< AurynStateVector * >	post_state_traces_states
std::string	group_name
	Identifying name for object. More...
NeuronID	size
	Stores the size of the group. More...
NeuronID	rank_size
	Stores the size of the group on this rank. More...
SpikeContainer *	spikes
	SpikeContainers to store spikes produced during one step of evolve. More...
AttributeContainer *	attribs

Additional Inherited Members
Protected Member Functions inherited from auryn::SpikingGroup
virtual void	load_input_line (NeuronID i, const char *buf)
virtual std::string	get_output_line (NeuronID i)
virtual void	virtual_serialize (boost::archive::binary_oarchive &ar, const unsigned int version)
	Implementatinon of serialize function for writing. More...
virtual void	virtual_serialize (boost::archive::binary_iarchive &ar, const unsigned int version)
	Implementatinon of serialize function for reading. More...
void	free ()
	Frees potentially allocated memory. More...
Static Protected Attributes inherited from auryn::SpikingGroup
static AurynTime *	clock_ptr = NULL

Detailed Description

A PoissonGroup with multiple subpopulations that co-modulate their firing rate according to an Ornstein Uhlenbeck process.

The group can be used to provide a stimulus for Hebbian learning in networks. When defined a given number of groups of given size are defined, as well as a correlation time.

Constructor & Destructor Documentation

CorrelatedPoissonGroup()

CorrelatedPoissonGroup::CorrelatedPoissonGroup	(	NeuronID	n,
		AurynDouble	rate = 5.,
		NeuronID	gsize = 100,
		AurynDouble	timedelay = 50e-3
	)

Default constructor.

Parameters

n	the size of the group which will be devided by the size of one subgroup to give the number of groups.
rate	the mean firing rate of all cells.
gsize	the size of one subgroup.
timedelay	delay between groups.

                                        : SpikingGroup( n ) 
{
        init(rate, gsize,  timedelay);
}

~CorrelatedPoissonGroup()

CorrelatedPoissonGroup::~CorrelatedPoissonGroup ( )

virtual

{
        if ( evolve_locally() ) {
                delete dist;
                delete shared_noise;
                delete rank_noise;
                delete delay_o;
        }
}

Here is the call graph for this function:

Member Function Documentation

evolve()

void CorrelatedPoissonGroup::evolve ( )

virtual

Virtual pure evolve function which needs to be implemented by derived classes.

The evolve function is called during simulations in every timestep by the System class. It updates the internal state of the spiking group and pushes spikes which are generated in this timestep to the axonal output delay (SpikeDelay).

Implements auryn::SpikingGroup.

{
        // check if the group has timed out
        if ( tstop && auryn::sys->get_clock() > tstop ) return;

        // move amplitude
        amplitude += (target_amplitude-amplitude)*auryn_timestep/tau_amplitude;

        // integrate Ornstein Uhlenbeck process
        o += ( mean - o )*auryn_timestep/timescale;

        // noise increment
        o += 2.0*((AurynDouble)(*shared_noise)()-0.5)*sqrt(auryn_timestep/timescale)*amplitude;


        // if ( sys->get_clock() % 10000  == 0 ) {
        //      std::cout << sys->mpi_rank() << " " << o << std::endl;
        // }

        int len = delay*ngroups;
        delay_o[auryn::sys->get_clock()%len] = std::max(thr,o*lambda);

        for ( unsigned int g = 0 ; g < ngroups ; ++g ) {
                AurynDouble grouprate = delay_o[(auryn::sys->get_clock()-(g+offset)*delay)%len];
                AurynDouble r = (*rank_noise)()/(auryn_timestep*grouprate); // think before tempering with this! 
                // I already broke the corde here once!
                x[g] = (NeuronID)(r); 
                while ( x[g] < groupsize ) {
                        push_spike ( g*groupsize + x[g] );
                        r = (*rank_noise)()/(auryn_timestep*grouprate);
                        x[g] += (NeuronID)(r); 
                }
        }
}

Here is the call graph for this function:

get_rate()

AurynDouble CorrelatedPoissonGroup::get_rate

(

)

{
        return lambda;
}

seed()

void CorrelatedPoissonGroup::seed

(

)

{
                shared_noise_gen.seed(sys->get_synced_seed()); 
                rank_noise_gen.seed(sys->get_seed()); 
}

Here is the call graph for this function:

set_amplitude()

void CorrelatedPoissonGroup::set_amplitude

(

AurynDouble

ampl

)

{
        amplitude = amp;
        auryn::logger->parameter("amplitude",amplitude);
}

Here is the call graph for this function:

set_offset()

void CorrelatedPoissonGroup::set_offset

(

int

off

)

{
        offset = off;
        auryn::logger->parameter("offset",offset);
}

Here is the call graph for this function:

set_rate()

void CorrelatedPoissonGroup::set_rate

(

AurynDouble

rate

)

{
        lambda = rate;
}

set_stoptime()

void CorrelatedPoissonGroup::set_stoptime

(

AurynDouble

stoptime

)

{
        tstop = stoptime*auryn_timestep;
        auryn::logger->parameter("stoptime",stoptime);
}

Here is the call graph for this function:

set_target_amplitude()

void CorrelatedPoissonGroup::set_target_amplitude

(

AurynDouble

ampl

)

{
        target_amplitude = amp;
        auryn::logger->parameter("target_amplitude",target_amplitude);
}

Here is the call graph for this function:

set_tau_amplitude()

void CorrelatedPoissonGroup::set_tau_amplitude

(

AurynDouble

scale

)

{
        tau_amplitude = tau;
        auryn::logger->parameter("tau_amplitude",tau_amplitude);
}

Here is the call graph for this function:

set_threshold()

void CorrelatedPoissonGroup::set_threshold

(

AurynDouble

threshold

)

{
        thr = std::max(1e-6,threshold);
}

set_timescale()

void CorrelatedPoissonGroup::set_timescale

(

AurynDouble

scale

)

{
        timescale = scale;
        auryn::logger->parameter("timescale",timescale);
}

Here is the call graph for this function:

Member Data Documentation

amplitude

AurynDouble auryn::CorrelatedPoissonGroup::amplitude

protected

Stores parameters for moving amplitude for slow changes to correlations structure

delay

AurynTime auryn::CorrelatedPoissonGroup::delay

protected

delay_o

AurynDouble* auryn::CorrelatedPoissonGroup::delay_o

protected

groupsize

NeuronID auryn::CorrelatedPoissonGroup::groupsize

protected

lambda

AurynDouble auryn::CorrelatedPoissonGroup::lambda

protected

mean

AurynDouble auryn::CorrelatedPoissonGroup::mean

ngroups

NeuronID auryn::CorrelatedPoissonGroup::ngroups

protected

o

AurynDouble auryn::CorrelatedPoissonGroup::o

protected

Stores Ornstein Uhlenbeck state

offset

int auryn::CorrelatedPoissonGroup::offset

protected

remainersize

NeuronID auryn::CorrelatedPoissonGroup::remainersize

protected

target_amplitude

AurynDouble auryn::CorrelatedPoissonGroup::target_amplitude

protected

tau_amplitude

AurynDouble auryn::CorrelatedPoissonGroup::tau_amplitude

protected

thr

AurynDouble auryn::CorrelatedPoissonGroup::thr

protected

timescale

AurynDouble auryn::CorrelatedPoissonGroup::timescale

protected

tstop

AurynTime auryn::CorrelatedPoissonGroup::tstop

protected

x

NeuronID* auryn::CorrelatedPoissonGroup::x

protected

Stores the hopping state

---

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

• /home/zenke/auryn/src/auryn/CorrelatedPoissonGroup.h
• /home/zenke/auryn/src/auryn/CorrelatedPoissonGroup.cpp