Example of machine learning:how a viewer would rate a movie?
you get the movie vector that resulted from that process.
the essence of machine learning:A pattern exist(the essense of it is you should collect data,ande you will get the answer)
Wa cannot pin it down mathematically
We have data on it
Components of learning
metaphor:credit approval
applicant information
formalization:input:x
output:y
target function:f
data:(x,y)
hypothesis g:you're hypothesis set is the set of all possible hypothesis.
solution components:the hypothesis set
the learning algorithm
A simple model:perception
a simple hypothesis set x=(x,x,x,x,x,)
approve credit if >threshold
deny credit if <threshold
this linear formula h can be written as H(x)=sign(( )-threshold)
let's assume that the data you are working with are linearly separable(it's not a problem).And realiaze what defines H is your choice of Wi and the threshold.we do that because we are going to introduce an artificial coordinate.
random weights?now we have the hypothesis set,the hypothesis set tells you the resources you can work with.now we need the algorithm that is going to look at the data.the training data that you are going to useand navigates through the space of the hypothesis,to bring the one that is going to output as the final hypothesis.
one iteration of the PLA:
inner product,
Types of learning:basic premise of learning :use a set of observations to concover an underlying process.such as statistics.the underlying process is a probability distribution,and the observations are samples generated by that distribution.
Puzzle
Supervised learning:引导性学习
Unsupervised learning:never tell you the target at all.
Reinforcement learning :give input and some output,drade for the output
Machine learning :use least hypothesis get the target function
Sampling bias:
Value functions:
Minimize the error function:error function and error measure
Choice of the hypothesis(model selection)