ai learning algorithms

the point that solves the likelihood function to the parameter derivative of zero: $$\phi=\frac{1}{m}\sum_{i=1}^{m}1\{y^{(i)}=1\} \ \mu_0= \frac{\sum_{i=1}^{m}1\{ y^{(i)}=0\}x^{(i)}} {\sum_{i=1}^m1\{y^{(i)}=0\}} \ \mu_1= \frac{\sum_{i=1}^{m}1\{y^{(i)}=1\}x^{(i)}} {\sum_{i=1}^ m1\{y^{(i)}=1\}} \ \ \sigma = \frac{1}{m}\sum_{i=1}^m (x^{(i)}-\mu_{y^{(i)}) (x^{(i)}-\mu_{y^{(i)}) ^t$$ get the estimated value of the parameter $ (\ PHI,\MU_0,\MU_1,\SIGMA) $, that is to get the distribution function $p

Search Algorithms for algorithm learning: static search table (1) sequential table search, algorithm static Introduction: The following operations are generally performed on a query table: 1. query whether a "specific" data element is in the query table; 2. query the attributes of a "specific" data element; 3. Insert a data element into the search table; 4. delete a data element from the search table. The

, example to get the classification result of Class 1The same input is transferred to different nodes and the results are different because the respective nodes have different weights and biasThis is forward propagation.10. MarkovVideoMarkov Chains is made up of state and transitionsChestnuts, according to the phrase ' The quick brown fox jumps over the lazy dog ', to get Markov chainStep, set each word to a state, and then calculate the probability of transitions between statesThis is the proba

. U3S3 , because using the idea above can explain all the other Δ I If there is a problem with this idea, please let us know and appreciate it. The path to the non-normalized probability maximum value is:ψ2 (1) = 1Δ2 (2) in the same vein.I=3 is the same (just for Δ3(1) in the u5s5, I think it should be u3s3, first not to say S3 the corresponding is y3=1 in the original question, and there is no S5 function).Part III termination :This step is simple, in δ3 (L) δ3 (1) = 4.3 maximum, so y3 1 is the

') plt.ylabel (' Ratio_sugar ') plt.title (' LDA ') plt.show () W=calulate_w () plot (W)The results are as follows: The corresponding W value is:[ -6.62487509e-04, -9.36728168e-01]Because of the relationship between data distribution, LDA's effect is not obvious. So I changed the number of samples of several label=0, rerun the program to get the result as follows:The result is obvious, the corresponding W value is:[-0.60311161,-0.67601433]Transferred from: http://cache.baiducontent.com/c?m= 9d7

In general, the information explosion has created an overloaded message. There are two main types of solutions: Search and recommendation. The search is the active purposeful, the intention is clear, the recommendation is non-active, the intention is not clear.The most classic aspect of recommendation is the collaborative filtering recommendation. My blog Here are two articles, a partial theory, an article about the ALS actual combat."Collaborative filtering of CF ALS and implementation on spar

Reprint Please specify:View animation learning algorithms and data structures (not updated periodically) ()Quick Sort (QuickSort)Animated Demo:Java code:public class QuickSort {private int array[];p rivate int length;public void sort (int[] inputarr) {if (Inputarr = = NULL | | i Nputarr.length = = 0) {return;} This.array = Inputarr;length = Inputarr.length;quicksort (0, length-1);} private void QuickSort (

Nine algorithms for machine learning---regressionTransferred from: http://blog.csdn.net/xiaohai1232/article/details/59551240Regression analysis is to quantify the size of the dependent variable affected by the independent variable, to establish a linear regression equation or a nonlinear regression equation, so as to predict the dependent variable, or the interpretation of the dependent variable.The regress

. Or after the derivation of the formula can not be interpreted, or the number of unknown parameters is greater than the number of equations. At this point, the iterative algorithm is used to find the optimal solution step-after-step. In particular, if the optimization function is a convex function, then there is a global optimal solution, if the function is non-convex, then there will be many local optimal solutions, so the importance of convex optimization is self-evident. People always wan

]) *double (Dy[i])#Sqx = double (Dx[i]) **2Sumxy= VDOT (Dx,dy)#returns the point multiplication of two vectors multiplySQX = SUM (Power (dx,2))#Square of the vector: (x-meanx) ^2#calculate slope and interceptA = sumxy/SQXB= meany-a*MeanxPrintA, b#Draw a graphicPlotscatter (XMAT,YMAT,A,B,PLT)7.1.4 Normal Equation Group methodCode implementation of 7.1.5 normal equation set#data Matrix, category labelsXarr,yarr = Loaddataset ("Regdataset.txt")#Importing Data Filesm= Len (Xarr)#generate x-coordinat

value;If it becomes smaller, the new puzzle will replace the original;If it becomes larger, the probability of replacing the old one with the new one depends on the current temperature value, where the temperature will begin to slow down at a relatively high value, which is why the algorithm is more receptive to relatively poor performance in the early stages of execution, so that we can effectively avoid the possibility of falling into the local minimum, when the temperature reaches 0, The alg

Perception Machine: This is the simplest machine learning algorithm, but there are a few points to note. The first is the selection of the loss function, and in order to minimize the loss function, the gradient descent method used in the iterative process, finally obtains the optimal w,bThe visual interpretation is to adjust the value of the w,b, so that the separation of the super-plane to the wrong classification point to move to reduce the distance

value of 3.For example: Np.random.randint (3, 6, size=[2,3]) returns data with a dimension of 2x3. The value range is [3,6].(4). Random_integers (low[, high, size]), similar to the above randint, the difference between the range of values is closed interval [low, high].(5). Random_sample ([size]), returns the random floating-point number in the half-open interval [0.0, 1.0]. If it is another interval [a, b), it can be converted (b-a) * Random_sample ([size]) + AFor example: (5-2) *np.random.ran

minimizing the degree of impurity at each step, the cart can handle the outliers and be able to handle the vacancy values. The termination condition of the tree partition: 1, the node achieves the complete purity; 2, the depth of the tree reaches the depth of the user3, the number of samples in the node belongs to the user specified number;Pruning method of tree is a pruning method of cost complexity;See details: http://blog.csdn.net/tianguokaka/article/details/9018933 Copyright NOTICE: This ar

than or equal to I for(inti =1; I 1; i++) {C[i] = C[i] + c[i-1]; }//c[i] is the position of this element in the sorted array for(inti = count-1; I >=0; i--) {b[c[a[i]]-1] = A[i];//This is c[a[i]]-1 because C + + array starts from 0 and pseudocode starts with 1C[a[i]]--;//Consider two elements equal, and the next element equal to the current element is placed in front //And since this is a backward-forward traversal, the relative position of the equal number h

Matching of random and scattered learning algorithms and re-describing strings Body String Matching is an old topic, and we are also keen to learn and explore it, and we often use it. For example, when we use Vim to open a text file and search for a string in this file, we only need to input/string in the baseline mode. For example, on the Linux terminal, We need to print all the C files in the current di

, the message is the probability of classification C, when the word appears more time, will come to the problem of accuracy, you can dissolve the problem into a joint probability, that is, the probability of each word to find P (c| Wi), and then take out the probability of the largest topn to solve, such as n=10,n=15, and so on, the joint probability formula is as follows: p=p1*p2*p3*....pn/(p1*p2*p3*....pn+ (1-P1) * (1-P2) * (1-P3) ... * (1-PN)), where P1-PN is our chosen topn probability.

attention to the fact that it is possible to encounter more than one classification probability in the actual operation or the probability of each classification is 0, at this time it is generally random to select a classification as the result. But sometimes it should be treated with care, such as using Bayesian to identify spam, if the probability is the same, even if the two probability difference is not large, it should be treated as non-spam, because the failure to identify the impact of s

1. Linear modelSimple form, easy to model, good explanatory2. Logistic regressionNo prior assumptions about the data distribution;Approximate probability prediction can be obtained.Many numerical optimization algorithms can be directly used to calculate the optimal solution for the convex function of arbitrary order of the rate function.3. Linear discriminant Analysis (LDA)When two kinds of data are the same as prior, Gaussian distribution and covaria

Recommend a Visual Web site "Visual Algo": Url= ' https://visualgo.net/en/sorting 'This website gives the principles and processes of various sorting algorithms, which are visualized through dynamic forms. The related pseudo-code are also given, as well as the specific steps to execute to code."Bubble Sort"You need to repeatedly visit the sequence of columns that need to be sorted. The size of the adjacent two items is compared during the visit, and i