It turns out that I wasn't doing my math wrong yesterday, I was just not adding correctly. That somehow sounds worse. In any case, I have a working two-class logistic regression program, which is kind of nice. The multiclass and outlier tolerant versions are the next logical extensions, which shouldn't be that much harder.
 |
| "You know what? We're done now. I'm a fucking bear, and I have some fucking dignity. Also: your face." |
 |
| More buildings need more buildings on top. |
- Looks like the Red Cross needs to read a certain blog post.
- More things should be done via binary tree.
- This is a link to wikipedia's logistic regression page. It is shit at explaining what's going on, so here's a picture that explains what it does far better. The data is the two-class data, and the model is the best fitting probability curve that assigns new data to the two classes.
No comments:
Post a Comment