Designing Machine Learning Systems with Python by 2016

Designing Machine Learning Systems with Python by 2016

Author:2016
Language: eng
Format: epub, mobi
Publisher: Packt Publishing


The ordered list approach

As we add literals to a conjunctive rule, we aim to increase the homogeneity of each subsequent set of instances covered by the rule. This is similar to constructing a path in the hypothesis space as we did for our logical trees in the last section. A key difference with the rule approach is that we are only interested in the purity of one of the children, the one where the added literal is true. With tree-based models, we use the weighted average of both children to find the purity of both branches of a binary split. Here, we are still interested in calculating the purity of subsequent rules; however, we only follow one side of each split. We can still use the same methods for finding purity, but we no longer need to average over all children. As opposed to the divide and conquer strategy of decision trees, rule-based learning is often described as separate and conquer.

Let's briefly consider an example using our conifer categorization problem from the previous section.



Download



Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.