机器学习基石-03-2-learning with different Data Labels
2017-10-22 10:32
351 查看
1.supervised learning监督学习:每一个xn都有对应的yn
2.unsupervised learning无监督学习,没有yn
3.Semi-supervised半监督学习: Coin Recognition with Someyn
4.Reinforcement Learning强化学习
当你很难定义yn="坐下"的时候,可以找另一个yn="撒尿是不对的行为”进行惩罚punish;
此时狗狗听指令坐下了,但是yn="坐下"还是很难定义,再找一个新的yn="坐下是好的行为"进行奖励reward。
reinforcement learning:
1.不是在原有的输出yn=“sit”做评价,而是在新的yn上面进行“punish惩罚”或者“reward奖励”;
2.不是大批量输入的而是sequentially一次一次地逐次输入的。
总结
2.unsupervised learning无监督学习,没有yn
3.Semi-supervised半监督学习: Coin Recognition with Someyn
4.Reinforcement Learning强化学习
当你很难定义yn="坐下"的时候,可以找另一个yn="撒尿是不对的行为”进行惩罚punish;
此时狗狗听指令坐下了,但是yn="坐下"还是很难定义,再找一个新的yn="坐下是好的行为"进行奖励reward。
reinforcement learning:
1.不是在原有的输出yn=“sit”做评价,而是在新的yn上面进行“punish惩罚”或者“reward奖励”;
2.不是大批量输入的而是sequentially一次一次地逐次输入的。
总结
相关文章推荐
- 机器学习基石-03-4-learning with different Input Space
- 机器学习基石-03-1-learning with different Output Space
- 机器学习基石 3.2 Learning with Different Data Label
- 机器学习基石 3-2 Learning with different data label
- 机器学习基石 3.1 Learning with Different Output Space
- 机器学习基石 3-3 Learning with different protocol
- 机器学习基石 3-1 Learning with different output space
- 机器学习基石 3.3 Learning with Different Protocol
- 机器学习基石 3-4 Learning with different input space
- 机器学习基石 3.4 Learning with Different Input Space
- Lecture3-1Learning with different data label
- [D3] Create Labels from Non-numeric Data with Ordinal Scales in D3 v4
- 林轩田-机器学习基石 课堂笔记(四)Feasibility of Learning
- 机器学习基石笔记 Lecture 1: The Learning Problem
- 【Perceptron Learning Algorithm】林轩田机器学习基石
- Learning Data Mining with Python-第一章-affinity analysis
- 機器學習基石(Machine Learning Foundations) 机器学习基石 作业三 课后习题解答
- TensorFlow Machine Learning with Financial Data on Google Cloud Platform
- AlphaGo:用机器学习技术古老的围棋游戏掌握AlphaGo: Mastering the ancient game of Go with Machine Learning
- 机器学习基石 - Types of Learning