Loading...
Please wait, while we are loading the content...
Similar Documents
Constructive induction for improved learning of boolean functions
| Content Provider | Semantic Scholar |
|---|---|
| Author | Drake, Peter |
| Copyright Year | 1995 |
| Abstract | approved: Prasad Tadepalli Supervised learning programs, such as decision tree learners and neural networks, often must learn Boolean functions. The concept being learned may not easily be expressed in terms of the atomic features given. Constructive induction automatically produces higher level features (combinations of the atomic features), which can improve learning performance. The FLIP algorithm, introduced in this thesis, uses an information gain metric to search for useful conjunctions of atomic features. Given these conjunctions, a decision tree learner is shown to produce trees which are both smaller and more accurate, when learning both random CNF functions and functions from game-playing domains. Furthermore, evidence is provided that FLIP constructs even better features when it has access to training sets for additional functions related to the function being learned. Redacted for privacy |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://ir.library.oregonstate.edu/downloads/5138jj19f?locale=en |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
