sPyNNaker neural_modelling 7.2.2
|
Izhekevich neuron type. More...
#include "neuron_model.h"
Go to the source code of this file.
Data Structures | |
struct | neuron_params_t |
definition of neuron parameters More... | |
struct | neuron_t |
definition for LIF neuron state More... | |
Functions | |
static void | neuron_model_initialise (neuron_t *state, neuron_params_t *params, uint32_t n_steps_per_timestep) |
static void | neuron_model_save_state (neuron_t *state, neuron_params_t *params) |
static void | rk2_kernel_midpoint (REAL h, neuron_t *neuron, REAL input_this_timestep) |
Midpoint is best balance between speed and accuracy so far. | |
static state_t | neuron_model_state_update (uint16_t num_excitatory_inputs, const input_t *exc_input, uint16_t num_inhibitory_inputs, const input_t *inh_input, input_t external_bias, REAL current_offset, neuron_t *restrict neuron) |
primary function called in timer loop after synaptic updates | |
static void | neuron_model_has_spiked (neuron_t *restrict neuron) |
Indicates that the neuron has spiked. | |
static state_t | neuron_model_get_membrane_voltage (const neuron_t *neuron) |
get the neuron membrane voltage for a given neuron parameter set | |
static void | neuron_model_print_state_variables (const neuron_t *neuron) |
static void | neuron_model_print_parameters (const neuron_t *neuron) |
Variables | |
static const REAL | SIMPLE_TQ_OFFSET = REAL_CONST(1.85) |
For linear membrane voltages, 1.5 is the correct value. However with actual membrane voltage behaviour and tested over an wide range of use cases 1.85 gives slightly better spike timings. | |
static const REAL | MAGIC_MULTIPLIER = REAL_CONST(0.040008544921875) |
The original model uses 0.04, but this (1 ULP larger?) gives better numeric stability. | |
Izhekevich neuron type.
Definition in file neuron_model_izh_impl.h.
struct neuron_params_t |
definition of neuron parameters
The state parameters of an Izhekevich model neuron.
definition for LIF neuron parameters
Definition at line 47 of file neuron_impl_stoc_exp.h.
Data Fields | ||
---|---|---|
UREAL | tau_ms | The tau value of the neuron. |
UREAL | time_step |
The timestep of the neuron being used. The time step in milliseconds. current timestep in ms |
REAL | bias | The bias value. |
uint32_t | refract_init | The initial refractory timer. |
mars_kiss64_seed_t | random_seed | Random seed to use. |
REAL | v_init | The initial membrane voltage. |
REAL | v_reset | The reset membrane voltage after a spike. |
UREAL | tau | The tau value of the neuron, multiplied by 2^v to get probability. |
UREAL | tau_refract |
The refractory period of the neuron in milliseconds. The refractory period of the neuron, in ms. |
REAL | alpha | The alpha value of the neuron prob = (2^(-2^(alpha x voltage))) |
REAL | V_init | membrane voltage [mV] |
REAL | c_m | membrane capacitance [nF] |
REAL | tau_m | membrane decay time constant |
REAL | I_offset | offset current [nA] |
REAL | V_reset | post-spike reset membrane voltage [mV] |
REAL | T_refract_ms | refractory time of neuron [ms] |
int32_t | refract_timer_init | initial refractory timer value (saved) |
REAL | time_step |
The time step in milliseconds. current timestep in ms |
REAL | A | |
REAL | B | |
REAL | C | |
REAL | D | |
REAL | V | |
REAL | U | |
REAL | next_h | next value of this_h (saved) |
REAL | V_rest | membrane resting voltage [mV] |
struct neuron_t |
definition for LIF neuron state
The state variables of an Izhekevich model neuron.
Definition at line 53 of file neuron_model_if_trunc.h.
Data Fields | ||
---|---|---|
REAL | V_membrane | membrane voltage [mV] |
REAL | R_membrane | membrane resistance [MOhm] |
REAL | I_offset | offset current [nA] |
int32_t | refract_timer | countdown to end of next refractory period [timesteps] |
REAL | V_reset | post-spike reset membrane voltage [mV] |
int32_t | T_refract | refractory time of neuron [timesteps] |
REAL | A | |
REAL | B | |
REAL | C | |
REAL | D | |
REAL | V | |
REAL | U | |
REAL | this_h | current timestep |
REAL | reset_h | timestep to reset to when not just spiked |
REAL | V_rest | membrane resting voltage [mV] |
REAL | exp_TC |
'fixed' computation parameter - time constant multiplier for closed-form solution exp(-(machine time step in ms)/(R * C)) [.] |
|
inlinestatic |
Definition at line 68 of file neuron_model_izh_impl.h.
|
inlinestatic |
Definition at line 81 of file neuron_model_izh_impl.h.
|
inlinestatic |
Midpoint is best balance between speed and accuracy so far.
From ODE solver comparison work, paper shows that Trapezoid version gives better accuracy at small speed cost
[in] | h | threshold |
[in,out] | neuron | The model being updated |
[in] | input_this_timestep | the input |
Definition at line 129 of file neuron_model_izh_impl.h.
|
inlinestatic |
primary function called in timer loop after synaptic updates
[in] | num_excitatory_inputs | Number of excitatory receptor types. |
[in] | exc_input | Pointer to array of inputs per receptor type received this timer tick that produce a positive reaction within the neuron in terms of stimulation. |
[in] | num_inhibitory_inputs | Number of inhibitory receptor types. |
[in] | inh_input | Pointer to array of inputs per receptor type received this timer tick that produce a negative reaction within the neuron in terms of stimulation. |
[in] | external_bias | This is the intrinsic plasticity which could be used for ac, noisy input etc etc. (general purpose input) |
[in,out] | neuron | the pointer to a neuron parameter struct which contains all the parameters for a specific neuron |
Definition at line 166 of file neuron_model_izh_impl.h.
|
inlinestatic |
Indicates that the neuron has spiked.
[in,out] | neuron | pointer to a neuron parameter struct which contains all the parameters for a specific neuron |
Definition at line 193 of file neuron_model_izh_impl.h.
get the neuron membrane voltage for a given neuron parameter set
[in] | neuron | a pointer to a neuron parameter struct which contains all the parameters for a specific neuron |
Definition at line 209 of file neuron_model_izh_impl.h.
|
inlinestatic |
Definition at line 213 of file neuron_model_izh_impl.h.
|
inlinestatic |
Definition at line 219 of file neuron_model_izh_impl.h.
|
static |
For linear membrane voltages, 1.5 is the correct value. However with actual membrane voltage behaviour and tested over an wide range of use cases 1.85 gives slightly better spike timings.
Definition at line 91 of file neuron_model_izh_impl.h.
|
static |
The original model uses 0.04, but this (1 ULP larger?) gives better numeric stability.
Thanks to Mantas Mikaitis for this!
Definition at line 119 of file neuron_model_izh_impl.h.