H. Potlapalli et Rc. Luo, PROJECTION LEARNING FOR SELF-ORGANIZING NEURAL NETWORKS, IEEE transactions on industrial electronics, 43(4), 1996, pp. 485-491
A new learning scheme, called projection learning (PL), for self-organ
izing neural networks is presented, By iteratively subtracting out the
projection of the ''winning'' neuron onto the null space of the input
vector, the neuron is made more similar to the input, By subtracting
the projection onto the null space as opposed to making the weight vec
tor directly aligned to the input, we attempt to reduce the bias of th
e weight vectors, This reduced bias will improve the generalizing abil
ities of the network, Such a feature is important in problems where th
e in-class variance is very high, such as, traffic sign recognition pr
oblems, Comparisons of PL with standard Kohonen learning indicate that
projection learning is faster, Projection learning is implemented on
a new self organizing neural network model called the reconfigurable n
eural network (RNN). The RNN is designed to incorporate new patterns o
nline without retraining the network, The RNN is used to recognize tra
ffic signs for a mobile robot navigation system.