Neural network Bayesian classifier, training, and validation statistics

Cluster (C)Training setValidation set
MembersCorrect classification (%)MembersCorrect classification (%)
13791.893590.05
22491.673293.75
31681.251181.82
4850.001369.23
519100.0014100.00
611100.001097.00
Total11589.5611587.94

A total correct classification of 89.6% was obtained during the training of the probabilistic neural network classifier (PNNC) with 115 subjects selected by random split of the total sample and 87.9% during validation with the remaining 115 subjects. Input factors: z-scored P300 latency (P300l), P300 amplitude (P300a), wideband power (zWBP), delta power (zDP), z-scored fast beta power (26–29 Hz sub-band, regional) (zFBP), maturational asymmetry (zMA), P50 ratio (zP50r). photic response at 20 Hz (zHR) coded bisynchronous spike waves (BSSW) and centrotemporal spikes (CTS), The spacing parameter was 0.198438 (optimized by jackknifing during training)