Solution 2 – prediction stacking
We started this chapter by stating that NSAI is not constrained by design, development rules, or principles. It is simply the marriage of symbolic learning and NNs. What does this mean for us? First, we can still leverage the power of NSAI without using complex algorithms or spending too much time figuring out the best way to extract the knowledge base. NSAI is highly creative. Following, we will go through the process of implementing a much simpler NSAI system using the same dataset.
In our previous example, we focused on representing knowledge as axioms. We feed this representation to the NN to map the relationships between the various dimensions to learn knowledge. Another way to extract knowledge in the form of symbolic statements would be to use decision trees (DTs). DTs use logical rules to make decisions and map the training data in a tree-like structure. Every node in the tree represents some logical condition, and the subsequent nodes...
