MATLAB Keywords

Extract

[...] The learning rate schedule can be controlled with the options. A common problem with using gradients is that the loss function may change rapidly with some parameters but slowly with others. Very large gradient values can make your parameter values change wildly. A simple way to control this is to put a maximum size on the gradient values. 1st input to TrainingOptions : ALGORITHM 'sgdm' - stochastic gradient descent with momentum use a different learning rate for each parameter RMSProp algorithm keeps a history of the size of the gradient, and uses this to scale the learning rate for each parameter. [...]

[...] To compare the feature values of different classes, use the "Group" option. data = categorical(data,catNames) : change the name of the categories with the variable string catNames T = rmmissing(T) : remove rows with missing values from T stdevData = groupsummary(data,"Label","std") : standard deviation (std) of the data in data, grouped by data.Label table.variable = : remove the variable from the table table.Properties.VariableNames : access the variable names in a table joinedData = innerjoin(tableA,tableB) : joins two tables, and includes only observations whose key variable values appear in both tables. [...]

[...] E : Eigenvalues of the matrix D : distance / dissimilarity vector pareto(vector) : create a Pareto chart, which visualizes relative magnitudes of a vector in descending order. k-mean clustering : d'abord afficher le graphe pour identifier le nombre de clusters idx = kmeans(X,k, "Distance","correlation"/ « sqeuclidean »/ « cityblock »/ « cosine »/ « hamming », "Start",[coordonnées centre_cluster_1; coordonnées centre_cluster_2 ; . ] / « cluster », "Replicates",m) idx : cluster indices (column) X : data K : number of clusters M : repeats the clustering m times and returns the clusters with lowest sumd. [...]

[...] coefficient of indicates a perfect negative linear correlation coefficient of indicates a perfect positive linear correlation coefficient of 0 indicates no linear correlation T = table(x,y,z,'VariableNames',["X","Y","Z"]) : create a table boxplot(x,c) : plot where the boxes represent the distribution of the values of x for each of the classes in c. loss(model,testdata) : Loss is a fairer measure of misclassification that incorporates the probability of each class (based on the distribution in the data). parallelcoords(data,"Group",classes) : parallel coordinates plot shows the value of the features (or "coordinates") for each observation as a line. [...]

[...] The output is a variable containing the fitted model. predClass = predict(model,newdata) : The inputs are the trained model and new observations. The output is a categorical array of the predicted class for each observation in newdata. confusionchart(ytrue,ypred) confusionchart(ytrue, ypred, "RowSummary","row-normalized") : same as former but with rates datastore(« juin2022*.txt ») : tous les fichiers texte commençant par juin2022 read(file) : import data from the 1st file (1st time), then if we call it again i twill import data from the 2nd file . [...]