QuickLearn- A Novel Supervised On-line Learning Algorithm for Pattern Recognition

This paper introduces a new kind of supervised classifier so-called QuickLearn (QL) which is utilized for training as well as for recall. It has two stages. Hn the first stage, mapping level, input data are represented to the multi input-single output mapping function (MF) with fixed weights during the training phase. So, this kind of map only is a kind of simple shifting and scaling of the input data before representing to the second stage. We can select any kind of mathematical function for this map while its complexity depends on input data complexity. By representing input data to the first stage, MF gives us a scalar value. After shifting and scaling that value to the range [0,T], we can round it to have an integer value (y). The second stage, matching level, only includes an array with T cells called Affine Lootk-up Table (ALT). The training phase for QL includes only one step, no learning cycles. In this single step, the integer value y is used as a reference address to call and upload the label for corresponding input samples in N cells of ALT (copying label from the cell [y-N/2] till the cell [y+ N/2-1]) (data leakages to N-1 neighbor cells). Elapsed time for training a fresh QL classifier is typically only few milliseconds. In the testing phase, we need only to recall and introduce value of the cell with the index y as the final output (final winner class). The QL is evaluated and compared with existing supervised neural networks on a variety of some well known pattern classification problems