NAME
weight_pool -- create and adapt connections between two neuron pools
PROTOTYPE
unitptr weight_pool( int iSrcNeurons, int iDestNeurons, float *pfScale, float (*pWeightProb)(float), float (*pWeightValue)(float), float (*pWeightRndGen)(void), float (*pFF)(float), float (*pGG)(float), float (*pWeightChange)(), char *pcOpt, unitptr uHost)
ARGUMENTS
- int iSrcNeurons
- nr of neurons in source population
- int iDestNeurons
- nr of neurons in destination population
- float *pfScale
- scaling factors: pfScale[0] multiplies probability function pWeightProb, pfScale[1..3] for connections along x,y and z-direction. Positive values select euclidean, negative values select max distance along the direction
- float (*pWeightProb)(float)
- distance dependent probability of presence of weight. NULL selects 1-0 threshold, yielding a deterministic network topology.
- float (*pWeightValue)(float)
- distance dependent factor to scale random weight value provided by random generator function for weights [see below]
- float (*pWeightRndGen)(void)
- random generator for weight values. Result will be multiplied by distance function pWeightValue. NULL selects gaussian distribution with variance of 1
- float (*pFF)(float)
- function for transforming postsynaptic neuron activity into `postsynaptic learning signal'' (see ADAPTATION AND LEARNING RULES). Specifying pF=FUN\_VOID makes extra input field to specify learning signal via input vector.
- float (*pGG)(float)
- function for transforming presynaptic neuron activity into `presynaptic learning signal'' (see ADAPTATION AND LEARNING RULES). Specifying pG=FUN\_VOID makes extra input (or output in case of pH=FUN_BP) field to specify learning (or retrieve backpropagated error) signal.
- float (*pWeightChange)()
- function with 1 or 3 float arguments [depending on pcOpt] used in learning rule (see ADAPTATION AND LEARNING RULES).
- char *pcOpt
- options
- unitptr uHost
- host unit
FILE
/amnt/loge/users/nistaff02/nistaff/rhaschke/nst7/man/../o.linux//../foldersrc/nst_neuron.c