udemy machine learning course review

(that is, Xi in {1,..., | v|} Value in | V| is the vocabulary of the lexicon), n-word messages will be represented by a vector of length n, and the length of the vectors for different articles will probably not be the same.In the multiple event model, we assume that this is the case with the message: first determine whether this is a spam message through P (Y), and then independently determine each word by multiple distributions P (x|y). The probability of the final generation of the entire mes

Andrew ng Machine Learning course 17 (2)Disclaimer: Reference Please specify source http://blog.csdn.net/lg1259156776/Description: This paper mainly introduces the use of value iteration and policy iteration two kinds of iterative algorithms to solve MDP problem, also introduced in practical application how to accumulate "experience" to update the transfer probab

can be processed.Cons: Easy to fit.How to avoid overfitting:(1) dimensionality reduction, can use PCA algorithm to reduce the dimension of the sample, so that the number of theta of the model is reduced, the number of times will be reduced, to avoid overfitting;(2) regularization, the design of regular items regularization term.The regularization function is to prevent some properties before the coefficient weight is too large, there has been a fitting.Note that the way to resolve overfitting i

is more than one, the Newton method iterates over the rule:Newton's method usually has a faster convergence rate than the batch gradient, and it takes a much smaller number of iterations to get close to the minimum value. However, when the parameters of the model are many (n), the computational cost of the Hessian matrix will be large, resulting in a slower convergence rate, but when the number of arguments is not long, the Newton method is usually much faster than the gradient descent.Summariz

Translator Note : This article is translated from the Stanford cs231n Course Note convnet notes, which is authorized by the curriculum teacher Andrej Karpathy. This tutorial is completed by Duke and monkey translators, Kun kun and Li Yiying for proofreading and revision.The original text is as follows Content list: structure Overview A variety of layers used to build a convolution neural networkThe dimension setting regularity of the arrangement law l

Model (how to simulate)---strategy (risk function)-algorithm (optimization method)First section:Basic concepts and classifications of machine learningSection II:Linear regression, least squaresBatch gradient descent (BGD) and random gradient descent (SGD)Section III:Over-fitting, under-fittingNon-parametric learning algorithm: Local weighted regressionThe probability angle interprets the linear regression.

Extremely light of a semester finally passed, summer vacation intends to learn the big step down this machine learning techniques.The first lesson is the introduction of SVM, although I have learned it before, but I heard a feeling is very rewarding. The blogger sums up a ballpark figure, and the specifics areTo listen: http://www.cnblogs.com/bourneli/p/4198839.htmlThe blogger sums it up in detail: http://w

I've been talking about why machines can learn, and starting with this lesson are some basic machine learning algorithms, i.e. how machines learn.This lesson is about linear regression, starting with the minimization of Ein, introducing the Hat Matrix to understand the geometric meaning. Finally, the linear regression and binary classification are compared, and the reason why linear regression can be used t

reduced after removing the label, (2) using the data of the reduced dimension to train the model, (3) for the new data points, the PCA reduced dimension to obtain the dimensionality reduction data, and the model to obtain the predicted value. Note : You should only use the training set data for PCA dimensionality reduction get Map $x^{(i)}\rightarrow z^{(i)}$, and then apply the mapping (PCA-selected principal matrix $u_reduce$) to the validation set and test set do not use PCA to block ove

: One-to-multiple ) Sometimes the problem is not as simple as determining whether a patient's tumor is malignant or benign. For example, determining whether the weather is sunny, cloudy, raining, Or snowing is necessary. We can use a line to separate binary classification. What about multiclass classification? There is a simple method, that is, to separate only one category at a time. There are several categories to construct several decision edge, that is, severalH (x): In th

ADD1 () DROP1 () 9. Regression Diagnostics Does the sample conform to the normal distribution? Normality test: function shapiro.test (X$X1) The distribution of normality Learning set/Is there outliers? How to find Outliers is the linear model reasonable? Maybe the relationship between nature is more complicated. Whether the error satisfies the independence, equal variance (the error is no

classifier will be severely affected, as shown in:To solve the above two problems, we adjust the optimization problem to:Note: When ξ>1, it is possible to allow the classification to be wrong, and then we add the ξ as a penalty to the target function.Using Lagrange duality again, we get the duality problem as:Surprisingly, after adding the L1 regularization item, only a αi≤c is added to the like limit in the dual problem. Note that the b* calculation needs to be changed (see Platt's paper)KKT d

Decision tree is to select the most information gain properties, classification.The core part is to use information gain to judge the classification performance of attributes. The information gain is calculated as follows:Information entropy:Multiple categories are allowed.Calculates the information gain for all attributes, choosing the largest root node as the decision tree. Then, the sample branches, continuing to determine the remaining properties of the information gain.Information gain has

This is what we have learned (except decision tree)Here is a typical decision tree algorithm, with four places to choose from:Then introduced a cart algorithm: By decision Stump divided into two categories, the criterion for measuring subtree is that the data are divided into two categories, the purity of these two types of data (purifying).The following is a measure of purity:Finally, when to stop:Decision tree may be overfitting, reducing the number of Ein and leaves (indicating the complexity

dimension.Finally, we propose a method for solving overfitting, including data cleaning/pruning, data hinting, regularization (regularization), confirmation (validation), andTo drive for example to illustrate the role of these methods, the latter two methods are also the contents of the following two lessons.Data cleaning/pruning is to correct or delete the wrong sample points, processing is simple, but usually such sample points are not easy to find.Data hinting generate more sample numbers by

This section is about regularization, in the optimization of the use of regularization, in class when the teacher a word, not too much explanation. After listening to this class,To understand the difference between a good university and a pheasant university. In short, this is a very rewarding lesson.First of all, we introduce the reason for regularization, simply say that the complex model with a simple model to express, as to how to say, there is a series of deduction hypothesis, very creative

columnsHttps://zhuanlan.zhihu.com/p/20878530?refer=intelligentunit Stanford cs231n Course Assignment # 1 Introduction-Smart Unit-Know the columnhttps://zhuanlan.zhihu.com/p/21441838 Stanford cs231n Course Assignment # 2 Introduction-Smart Unit-Know the columnhttps://zhuanlan.zhihu.com/p/21941485 Stanford cs231n Course Assignment # 3 Introduction-Smart Unit-Know

The idea of boosting is to integrate learning and combine many weak classifiers to form a strong classifier.First enter the original training sample, get a weak classifier, you can know its correct rate and error rate. Calculate the weight of the weak classifier as follows:Then increase the weight of the error classification sample, let the following classifier focus them, adjust the weight of the sample:If the original classification is correct:If th

. DrawingT=[0:0.01:0.98]Y1=sin (2*pi*t)Plot (t,y1) % drawingOnY2=cos (2*pi*t)Plot (T,y2, ' R ')Xlabel (' time ')Ylabel (' value ')Legend (' Sin ', ' cos ') % legendTitle (' My Plot ')Print-dpng ' myplot.png ' % saved as picture fileClose % Closes the current diagramFigure (1) % Create a diagramCLF % Empty chart Current ContentsSubplot (1,2,2) % graph cut to 1*2 grid, draw 2nd gridAxis ([0.5 1-1 1]) % axis changed to x belongs to [0.5,1],y belonging to [ -1,1]Imagesc (The Magic ()), Colorbar,colo

general, multiplication does not satisfy the exchange law: $ \ Matrix {A} \ times \ matrix {B} \ not = \ matrix {B} \ times \ matrix {A} $Special Matrix $ \ Matrix {I }=\ matrix {I _ {n \ times N }}=\ begin {bmatrix} 1 0 \ cdots 0 0 \ Cr0 1 \ cdots 0 0 \ Cr \ vdots \ vdots \ Cr0 0 \ cdots 1 0 \ Cr0 0 \ cdots 0 1 \ Cr \ end {bmatrix} $ For any matrix $ \ matrix {A} $: $ \ Matrix {A} \ times \ matrix {I }=\ matrix {I} \ times \ matrix {A }=\ matrix {A} $Inverse Matrix and inverte