《LMS(最小二乘方法)》由会员分享，可在线阅读，更多相关《LMS(最小二乘方法)（25页珍藏版）》请在金锄头文库上搜索。
1、Machine LearningWhy Machine Learning?nIs computer smart?nDifficulty to get and represent the knowledge.Why Machine Learning?Human learningMachine learningSlowSlow ? Fast?Can not copyCopy easily expansiveCheap? Expansive?Why Machine Learning?nNew kind of capability for computers- Database mining* Med
2、ical records - medical knowledge- Applications we cant program by hand* Autonomous driving* Speech recognitionWhy Machine Learning?nUnderstand human learning and teachingnTime is right- Recent progress in algorithms and theory- Growing flood of online data- Computational power is availableModel Base
3、d LearningnModelingnGet the parameters of model based on the training dataModel Based LearningnExample 1:nData set: nEquation: y=ax2+bx+cnProblem: to find the parameters: a, b, cModel Based LearningnLMS（最小二乘方法） methodModel Based LearningnNew problem: how to update the a, b and c from new data:Model
4、Based LearningnExample 2:nData set: nNormal distribution: nProblem: to find the parameters: Model Based LearningnCases in pattern recognitionnSee: 模式识别,清华大学出版社，2000Model Based LearningnMore general problemnData set: nA mathematical model.nProblem: to find the parameters of model.Model Based Learning
5、nMore examples.nSpeech signal processing.nFace recognitionnSignature verificationnModel Based LearningnProgram: Mark V. Shaneys workn擦洗黑板。观察黑板的干燥过程。水进 入空气。当水进入空气时，水就蒸发了 。把一条湿布系于一个固体或一瓶液体 的一端。看看周围。固体是什么？那些 进入云层的东西是什么？当云层移动时 ，空气就使得云层中的小水滴上下颠簸 。Model Based Learningn有时从云上掉下的水进入了绿色植物的 叶片。绿色植物有相当多的孔，几乎就 象
6、海绵。n为什么我们五个五个一组地计数呢？在 人们学习数许多东西时，往往扳着自己 的手指头来数。首先他们点出恰好与自 己双手的全部手指头一样多的东西。 Model Based LearningnProgram: Mark V. Shaneys worknWhy do we count things in groups of five. When people learned how to count many things, they matched them against their fingers. First they counted out enough things to match
7、 the fingers of both hands. Then they put these things aside in one quart. A giant- size bottle that will hold four quarts is a three- digit number. A digit gets its value from the top of a sidewalk are straight, and Model Based LearningnProgram: Mark V. Shaneys worknMarkov Chain.nSteps:nBuild a wor
8、ds set from the training text.nLearning transformation matrix from the training text.nWrite a new paper.Model Based LearningnModel?一位妇人藏了五只灰色的小猫 在破旧的汽车里 此时那位哀伤的乡农 唤起你痛苦的回忆Model Based LearningnModel?树儿垂下枯萎的枝头 埋在黑色的冰地里 就像三只灰色的母鹅 滑向近旁的雪堤Model Based Learning一位妇人藏了五只灰色 的小猫 在破旧的汽车里 此时那位哀伤的乡农 唤起你痛苦的回忆树儿垂下枯
9、萎的枝头 埋在黑色的冰地里 就像三只灰色的母鹅 滑向近旁的雪堤Model Based Learning科学，scientific American, 1991-1994? 程序创作的散文及韵律诗集， A.K.DewdneyLearn How to Play GameLearn How to Play GamenInitial the board with the same weightnUpdate the weight every step:- find the important positions.(positive and negative )- set the weight of p
10、ositive position bigger.- set the weight of negative position smaller.Inductive LearningOther Theories and MethodsnNeural network- compose new music- control ping-pong ball- 倒立摆nGenetic algorithmnVC dimensionnSome Methods in Pattern RecognitionnnTextbook: Machine learning, Tom Mitchell, McGraw-Hill, 1997
