FrostMonarch的博客

    Any machine learning can be thought as 3 parts . (1) Experience  ，E means the machine learn from what experience (In chinese means 需要怎样的学习过程)(2)P,P means the performance measure. (In chinese means...

Anomaly detectionThis algorithm can help us to realize that whether some data sets are abnormal. We should follow the steps:(1)Choose the features that you think may relate to anomalous examples(2)(3)

Model choosingAfter,we get a model. Sometimes we will wonder how we can optimize it.In order to do so, we can divide our data set into 3 parts, First the traning set(60%),Second the cross validation ...

Cost function for neuron modelBack propagationMore intuition about back propWe assume that we have just one unit in the output layer. and we define cost(i) followed:the delta term(error term)  is :Bec...

Fmincg  functionIt's a OCTAVE function. It can be helped to calculate the gradient decent.Example:options = optimset('GradObj','on','MaxIter',number_of_iterations);theta = fmincg((@t)cost_function_nam...

Unsupervised learning(1)In unsupervised learning, data have no labels which means they don't have yi {i=1~m} (2)Data has its own structure and a algorithm which can find clusters are called clustering...

Artificial Neuron Model   (1) (2)We should learn some denotion first to understand the above second model.For the activation units ,we have the below expression :Finally , we can get the conclusion th...

In the PPT, andrew tells us that we can choose the new feature which means we can choose x x^2 x^3 as our feature.(For example , the hypohesis = x0 +theta1 *x1 +theta2*x2 ... now we can change it into...

In order to make the gradient decent faster ,we can use this formular . After updating the xi ,we could use gradient decent faster. NOTICE : If we want to predict some other values , the input use the...

Logistic functionWhen we are solving classifcation problem,we seldom use linear regression method.We should define a new function called logistic function to be our  hypothesis.and we should notice th...

Vectorization    Vectorization is a useful method to calculate. It may make the calculation much more faster in Octave/Matlab.So how to make our calculation vectorize.    (1)When we see a add operator...

Over Fitting ProblemOver fitting means that the hypothesis can get good predictions on the traning set but can not get good predictions outside the tranining set. The figure looks like the below secon...

Multi-class classificationWhen facing the classification problem, the output are multiple values , we call this multi-class classfication problem.In order to solve the problem, we use the algorithm 'o...

In this part , I am going to talk about how to minimize the cost function .In Andrew's class , He talked about a method called gradient decent .  is the cost function ,of course ,there can be more var...

When facing the large scale data sets, it's necessary to compute more efficiently. Before we need to use a more computational method, we'd better sanity check first.There are several improving gradien...