Databases Reference
In-Depth Information
1. For neurons in the base layer, their bias entry takes the form
{left, right}, where left and right represent the row number of left-
adjacent and right-adjacent neurons, respectively.
2. For neurons in the middle layer, their bias entry takes the form
{leftIndex, lowerIndex, rightIndex}, where leftIndex, lowerIndex,
and rightIndex represent indices obtained from its left, lower (within
the same column), and right neurons, respectively.
3. The bias entry structure of the top layer neuron is in the form
{lowerIndex}, which is the index obtained from its lower layer neuron
(within the same column).
4.2 Complexity and Scalability of Hierarchical DPR
Scheme
4.2.1 Complexity Estimation
The following discussion focuses on the complexity analysis of the HGN
pattern recognition scheme. We will focus on the bias array capacity analysis
and Big-O estimation of the HGN network. A similar analysis was carried out
on a GN network in Section 3.3.
For Big-O estimation, the HGN strictly follows the adjacency compari-
son approach of the GN recognition procedure. The difference between the
HGN and GN implementations is their execution process. The HGN applies
multiple-stage execution (based on the hierarchical structure), and GN im-
plements single-stage execution. Therefore, the Big-O estimation of the com-
plexity of the HGN is O (n).
In the storage capacity analysis, we consider the bias array capacity of each
neuron within the HGN composition. A detailed analysis of the HGN storage
capacity has been discussed in [3]. Though we do not intend to repeat the
explanation in this topic, a summary of the complexity estimation will be
presented.
In this analysis, the size of the bias array is observed as different patterns are
stored. The number of possible pattern combinations increases exponentially
with increasing pattern size. The impact of the pattern size on the bias array
storage is an important factor in any bias array complexity analysis. The
analysis is conducted by segregating the bias arrays according to the layers of
a particular HGN network. The following equations show the bias array size
estimation for binary patterns. This bias array size is determined using the
number of bias entries recorded for each neuron.
