Part 1:welcome
Examples:
Database Mining : Large datasets from growth of automation/web.
E.g. ,Web click data, medical records, biology, engineeringApplications can’t program by hand
E.g. ,Autonomous helicopter, handwriting recognition, most of Natural Language Processing (NLP), Computer Vision.Self-customizing programs
E.g. ,Amazon, Netflix(一家在线影片租赁提供商) product recommendationsUnderstanding human learning (brain, real AI).
Part 2:What’s Machine Learning
1.By Arthur Samuel :
Machine Learning : Field of study that gives computers the ability to learn without being explicitly programmed.
- Example: checkers(国际象棋) learning program
2.By Tom Mitchell :
Well-posed Learning Problem : A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.
- Example: Email anti-spam Classify Program
3. Machine Learning Algorithms:
- Supervised learning 有监督学习
- Unsupervised learning 无监督学习
- Others: Reinforcement learning(增强学习)、Recommender systems 推荐系统
Part 3: Supervised Learning
- the “right answers” of * data set* were given.(数据集已经给出了分类,为每一个数据打了标签)
- Regression(回归问题): Predict continuous valued output
E.g. : Classification(分类问题)
- Breast cancer prediction:malignant or benign(only one feature: Tumor size)
- Two feature: Age and Tumor size,
- ….
- infinite feature : Support Vector Machine
Part 4:Unsupervised Learning
Doesn’t have any labels or that all has the same label or really no labels.(只有数据,没有标签)
- Clustering(聚类问题)
- 新闻分类、基因分类
- 处理器集群聚类、社会关系分析
- Market segmentation、Astronomical data analysis
Coktail party problem
将混合的两种声音分离开来,用C++、Java等实现起来非常复杂,但是用Octave只需一行代码:[W,s,v] = svd((repmat(sum(x.*x,1),size(x,1),1).*x)*x');