Semi Supervised
- Induction is reasoning from observed training cases to general rules, which are then applied to the test cases.
- Transduction is reasoning from observed, specific (training) cases to specific (test) cases. 1
现在有这个问题,已知 ABC 的类别,求问号的类别,
inductive learning 就是只根据现有的 ABC,用比如 kNN 距离算法来预测,在来一个新的数据的时候,还是只根据 5 个 ABC 来预测。
transductive learning 直接以某种算法观察出数据的分布,这里呈现三个 cluster,就根据 cluster 判定,不会建立一个预测的模型,如果一个新的数据加进来 就必须重新算一遍整个算法,新加的数据也会导致旧的已预测问号的结果改变。2
ref:
https://zhuanlan.zhihu.com/p/455808338
-
https://www.quora.com/What-is-the-difference-between-inductive-and-transductive-learning ↩
-
如何理解 inductive learning 与 transductive learning? - TechOnly 的回答 - 知乎 https://www.zhihu.com/question/68275921/answer/480709225 ↩