The training features, time, errors, accuracy, and input attribute weights of the prototype system are analyzed.
The ar-chitecture uses the color and texture features as input attributes to discriminate the objects in a scene.
Simultaneous use of Vp, Vs and Vp/Vs as input variables to the network in training the network give more accurate predictions than when say, Vp or Vs is used individually as input attribute in the inversion process.
Next, we compute individual clusterings based on each input attribute.
The selected features are used as the input attributes to the classifier system.
Figure 3 Rand index values for each input attribute at various cut heights.
Hence, we start by clustering on each individual attribute independently, before combining the input attributes as described below.
For each input attribute, excluding images, we calculated both the Jaccard and Cosine distances between all pairs of websites creating pairwise distance matrices for each input attribute and distance measure.
After the individual clusterings were calculated for each input attribute, websites could be sorted to identify manual clustering candidates which were placed in the exact same clusters for each individual input attribute's automated clustering.
Distance matrices were created in parallel for each input attribute by 'mapping' website input attributes into pairwise matches, and then simultaneously 'reducing' pairwise matches into distances using the appropriate distance metric.
