|
|
|
|
|
|
|
|
|
|
|
• |
Maximal margin
hyperplane
|
|
|
• |
Highly effective
for text classification
|
|
|
|
– |
Can we do even
better with active learning?
|
|
|
• |
Assumption:
|
|
|
|
– |
Actual labeling
of the entire pool is linearly separable
|
|
|
– |
Remark:
assumption can be relaxed
|
|
|
• |
We will give
querying functions for SVMs
|
|