This project demonstrates how multiclass classification works in a neural network, and how it is different from a regular MLP that only performs binary discrimination. This project is an interactive tutorial. Please open the code and read through the comments as you click on every script to see how it works! Instead of a single output neuron, the output layer of a multiclass classification neural network has one neuron per class. The expected target value therefore has to be a vector of all zeros with the index of the expected class activated to 1. For this there is a "vectorize (number)" function. In order to generate an output vector that can be diffed with the target vector the output layer's response is not activated with sigmoid (or anything else) per neuron, but the whole output vector is instead activated with the softmax function, which answers a probability distribution and sums up all neurons to 1. When using a mutliclass neural network for classifying the output vector is scanned for the index of the greatest probability. That index represents the class.data:image/png;base64,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 project demonstrates how multiclass classification works in a neural network, and how it is different from a regular MLP that only performs binary discrimination. This project is an interactive tutorial. Please open the code and read through the comments as you click on every script to see how it works! Instead of a single output neuron, the output layer of a multiclass classification neural network has one neuron per class. The expected target value therefore has to be a vector of all zeros with the index of the expected class activated to 1. For this there is a "vectorize (number)" function. In order to generate an output vector that can be diffed with the target vector the output layer's response is not activated with sigmoid (or anything else) per neuron, but the whole output vector is instead activated with the softmax function, which answers a probability distribution and sums up all neurons to 1. When using a mutliclass neural network for classifying the output vector is scanned for the index of the greatest probability. That index represents the class. reportAttributeOf reportJSFunctiondraw a list of numbers as vertical lines distributed evenly across the stage.

de:male Balken _ _ ca:dibuixa amb barres _ _ 0.8 0 1posslicewidthstagelengthpen sizesizeleftstage2Draw a picture of the specified model of a neural network where each layer is represented as a vertical line of dots and each weight as a line between 2 neuron-dots. The line width represents the weight's absolute value, negative values can be rendered in another color.

de:male neurales Modell _ _ ca:renderitza el model neuronal _ _ 1 1 rgba(214,49,0,255)flat line endsfalseabsbottomstage100flat line endsReport a copy of the given table in which the numerical data of each column is distributed between 0 and 1 using the column's min and max values (feature scaling).

de:normalisiere Tabelle _ ca:_ normalitzada Report a function (ring) that can be called with a single row (record) in the form of the given data set to normalize it using the sample's min and max values. Use this reporter to create a normalization function from a training set that can be applied to validation or live data.

de:Normalisierung für Tabelle _ ca:normalització per a la taula _ Split a table into 2 sets by randomly assigning its rows to each partition at the given ratio while making sure that every category (as indicated by the tag in the last column) is represented equally in both sets, reports a 2-item list containing the shuffled and split data. Use this block to create training and validation data sets.

de:teile Tabelle _ im Verhältnis _ ca:partició de _ per _ 0.8each groupshuffledeachcolumnsGenerate a new block in the sensing category that either reports the class of a given sample data record (for the the "classifier" option) or whether a given sample data record classifies as the given tag (name) based on an example dataset (for the "predicate" option) in the form of a binary truth table. The generated predicate block offers its estimated accuracy in its help screen / comment and can then be exported and shared. By default training happens all automatically using a neural network with one hidden layer, observing the learning progress and partitioning the dataset internally into a training set and validation set. Optionally you can specify an exact number of epochs (0 = automatic), a partitioning fraction (0 = automatic, 1 = none), and none to 8 hidden layers with arbitrary neurons (0 = no hidden layers). You can abort / shorten the training process manually by positioning the mouse pointer near the stage center and pressing the mouse button down. Running the command again updates any previously generated block, i.e. you can optimize existing blocks by re-training them with different parameters.

de:generiere _ für _ in _ _ $_predicate$_predicate=$_predicate $_classifier=$_classifier$_tag§_dynamicMenu$_auto 0.8 $_autoclassifierclasspredicate01

de:klone _ _ ca:clon _ _ $_field $_thing11

de:Objekt _ ca:objecte _ $_field $_thing11

last1logitsobjecttopologyset learning ratealphapredictsampleclassifylength1outputsamplefitnlasterrordeltailength1layerabssampletrainsetvalidatelengthsetget modelweightsset modellengthlast1logitsmodelshuffleget topology2dimensionsweights111dimensionsweightsversion1_Layer_inputsthingweightsthingsetupinoutsolvesamplelearnerrorreshuffledeltathingadjust weightslearning rate0.1activationsigmoidCreate and report a new neural network with the specified topology representing the number of inputs, arbitrary hidden layers, and output(s).

de:neues neuronales Netzwerk _ ca:nova xarxa neuronal _ 5 1

de:generiere Prädikat für _ in _ _ $_tag§_dynamicMenu$_auto 0.8 $_autovar name_AI: var getterai1var_declare(scope, name)global1initsampleblocks222213323423get model1normblocks22labelide_translate(text)is _ ?def1txt_transform(name, txt)selectclassifycommentpredict whether the data sample classifies as , estimated accuracy: 210001099.9%.oldcustom?type3labelblockspredicate01

$_auto 0.8 $_autotopology1length23auto303length.25lengthcolumnsaiepochs0donefalselastlengthlogscale110rendererTurtle spriteflat linesflat line endsflat line endstruereadout1last100010%flat line endsaccuracyvalidatereadout100010%Train a single epoch or validate a neural model on a truth-table dataset (a list of number-vectors with the expected classification in the last column). For "train" this reports the accumulated activation error over the epoch. For "validate" this reports the overall classification accuracy of the dataset. For "confusion matrix" this reports an aggregated binary error table in the form: TP, FP FN, TN

de:_ Netzwerk _ mit _ ca:_ xarxa _ amb _ $_traintrain=$_train validate=$_validate ~ confusion matrix=$_confusion matrixshuffledPredict and report the classification of a given data sample - a list of numbers representing a single record in a dataset - using the specified neural network instance. The result is a list of numbers representing the neural network's output layer. The "classify" selector reports the absolute classification, "predict" reports the unrounded output of the forward pass, letting you determine the neural network's confidence.

de:_ _ mit Netzwerk _ ca:_ _ amb la xarxa _ $_classifyclassify=$_classify predict=$_predictQuery the model or the learning rate of a given neural network. For "model" this reports a list of weight-matrices representing the neural network's hidden and output layers. Models can be exported and shared among projects. For "learning rate" this reports a single number representing the neural network's eagerness to adjust its weights when learning.

de:_ von Netzwerk _ ca:_ de la xarxa _ $_get modelget model=$_get model get learning rate=$_get learning rate get topology=$_get topologyAssign a pre-trained model to the given neural network or change its learning rate.

de:_ von Netzwerk _ auf _ ca:_ de la xarxa _ a _ $_set modelset model=$_set model set learning rate=$_set learning rate

de:blockifiziere _

de:generiere Prädikat für _ in _ _ $_tag§_dynamicMenu$_auto 0.8 $_auto1initsampleblocks222213323423get model1normblocks22label ide_translate(text)of _deftxt_transform(name, txt)selectclassifycommentpredict the data sample's , estimated accuracy: 210001099.9%.oldcustom?type2labelblocksreporterclassCreate and report a new sprite that can serve as a layer in a neural network of sprites. It responds to 3 events / methods: 1) setup Initializes the layer with a list of 2 numbers representing the number of inputs and the number of desired output neurons. 2) predict Reports the result of a forward pass of a single sample / record with the precision of the activation function, i.e. not rounded for classification. If you wish to use this answer to classify the record you must also sum and round the list of results. 3) learn Adjusts the layer's weights depending on the given delta vector and reports a new delta vector to be backpropagated to the previous layer, if any. You can either use a single sprite as a SLP (perceptron), or clone it several time to create additional hidden layers for a deep neural network by using a forward PIPE for prediction and a backward PIPE for learning.

de:neues Perzeptron Objekt var_declare(scope, name)globallearning ratelearning rate

data:image/png;base64,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This project demonstrates how multiclass classification works in a neural network, and how it is different from a regular MLP that only performs binary discrimination. This project is an interactive tutorial. Please open the code and read through the comments as you click on every script to see how it works! Instead of a single output neuron, the output layer of a multiclass classification neural network has one neuron per class. The expected target value therefore has to be a vector of all zeros with the index of the expected class activated to 1. For this there is a "vectorize (number)" function. In order to generate an output vector that can be diffed with the target vector the output layer's response is not activated with sigmoid (or anything else) per neuron, but the whole output vector is instead activated with the softmax function, which answers a probability distribution and sums up all neurons to 1. When using a mutliclass neural network for classifying the output vector is scanned for the index of the greatest probability. That index represents the class.0.8split the data seit in two parts, one for training the neural network, the other one to validate it latertagslastcolumnsthe last column contains the tags for the species (i.e. "versicolor", "setosa" or "virginica"classessorteduniquesfeaturescolumns14columnsonly the columns containing the features get normalized (feature-scaling)normalizethe normalization function gets stored so it can be applied to "real" (validation) records in the futuredata14lastcolumnscolumnscolumnsafter normalization the last column is again added to the data set, the former tags are replaced by integers that represent the class, i.e. the position of the tag in the list of classesTurtle spriteaccuracy1length100010%0.1sepal_length,sepal_width,petal_length,petal_width,species5.1,3.5,1.4,0.2,setosa4.9,3.0,1.4,0.2,setosa4.7,3.2,1.3,0.2,setosa4.6,3.1,1.5,0.2,setosa5.0,3.6,1.4,0.2,setosa5.4,3.9,1.7,0.4,setosa4.6,3.4,1.4,0.3,setosa5.0,3.4,1.5,0.2,setosa4.4,2.9,1.4,0.2,setosa4.9,3.1,1.5,0.1,setosa5.4,3.7,1.5,0.2,setosa4.8,3.4,1.6,0.2,setosa4.8,3.0,1.4,0.1,setosa4.3,3.0,1.1,0.1,setosa5.8,4.0,1.2,0.2,setosa5.7,4.4,1.5,0.4,setosa5.4,3.9,1.3,0.4,setosa5.1,3.5,1.4,0.3,setosa5.7,3.8,1.7,0.3,setosa5.1,3.8,1.5,0.3,setosa5.4,3.4,1.7,0.2,setosa5.1,3.7,1.5,0.4,setosa4.6,3.6,1.0,0.2,setosa5.1,3.3,1.7,0.5,setosa4.8,3.4,1.9,0.2,setosa5.0,3.0,1.6,0.2,setosa5.0,3.4,1.6,0.4,setosa5.2,3.5,1.5,0.2,setosa5.2,3.4,1.4,0.2,setosa4.7,3.2,1.6,0.2,setosa4.8,3.1,1.6,0.2,setosa5.4,3.4,1.5,0.4,setosa5.2,4.1,1.5,0.1,setosa5.5,4.2,1.4,0.2,setosa4.9,3.1,1.5,0.1,setosa5.0,3.2,1.2,0.2,setosa5.5,3.5,1.3,0.2,setosa4.9,3.1,1.5,0.1,setosa4.4,3.0,1.3,0.2,setosa5.1,3.4,1.5,0.2,setosa5.0,3.5,1.3,0.3,setosa4.5,2.3,1.3,0.3,setosa4.4,3.2,1.3,0.2,setosa5.0,3.5,1.6,0.6,setosa5.1,3.8,1.9,0.4,setosa4.8,3.0,1.4,0.3,setosa5.1,3.8,1.6,0.2,setosa4.6,3.2,1.4,0.2,setosa5.3,3.7,1.5,0.2,setosa5.0,3.3,1.4,0.2,setosa7.0,3.2,4.7,1.4,versicolor6.4,3.2,4.5,1.5,versicolor6.9,3.1,4.9,1.5,versicolor5.5,2.3,4.0,1.3,versicolor6.5,2.8,4.6,1.5,versicolor5.7,2.8,4.5,1.3,versicolor6.3,3.3,4.7,1.6,versicolor4.9,2.4,3.3,1.0,versicolor6.6,2.9,4.6,1.3,versicolor5.2,2.7,3.9,1.4,versicolor5.0,2.0,3.5,1.0,versicolor5.9,3.0,4.2,1.5,versicolor6.0,2.2,4.0,1.0,versicolor6.1,2.9,4.7,1.4,versicolor5.6,2.9,3.6,1.3,versicolor6.7,3.1,4.4,1.4,versicolor5.6,3.0,4.5,1.5,versicolor5.8,2.7,4.1,1.0,versicolor6.2,2.2,4.5,1.5,versicolor5.6,2.5,3.9,1.1,versicolor5.9,3.2,4.8,1.8,versicolor6.1,2.8,4.0,1.3,versicolor6.3,2.5,4.9,1.5,versicolor6.1,2.8,4.7,1.2,versicolor6.4,2.9,4.3,1.3,versicolor6.6,3.0,4.4,1.4,versicolor6.8,2.8,4.8,1.4,versicolor6.7,3.0,5.0,1.7,versicolor6.0,2.9,4.5,1.5,versicolor5.7,2.6,3.5,1.0,versicolor5.5,2.4,3.8,1.1,versicolor5.5,2.4,3.7,1.0,versicolor5.8,2.7,3.9,1.2,versicolor6.0,2.7,5.1,1.6,versicolor5.4,3.0,4.5,1.5,versicolor6.0,3.4,4.5,1.6,versicolor6.7,3.1,4.7,1.5,versicolor6.3,2.3,4.4,1.3,versicolor5.6,3.0,4.1,1.3,versicolor5.5,2.5,4.0,1.3,versicolor5.5,2.6,4.4,1.2,versicolor6.1,3.0,4.6,1.4,versicolor5.8,2.6,4.0,1.2,versicolor5.0,2.3,3.3,1.0,versicolor5.6,2.7,4.2,1.3,versicolor5.7,3.0,4.2,1.2,versicolor5.7,2.9,4.2,1.3,versicolor6.2,2.9,4.3,1.3,versicolor5.1,2.5,3.0,1.1,versicolor5.7,2.8,4.1,1.3,versicolor6.3,3.3,6.0,2.5,virginica5.8,2.7,5.1,1.9,virginica7.1,3.0,5.9,2.1,virginica6.3,2.9,5.6,1.8,virginica6.5,3.0,5.8,2.2,virginica7.6,3.0,6.6,2.1,virginica4.9,2.5,4.5,1.7,virginica7.3,2.9,6.3,1.8,virginica6.7,2.5,5.8,1.8,virginica7.2,3.6,6.1,2.5,virginica6.5,3.2,5.1,2.0,virginica6.4,2.7,5.3,1.9,virginica6.8,3.0,5.5,2.1,virginica5.7,2.5,5.0,2.0,virginica5.8,2.8,5.1,2.4,virginica6.4,3.2,5.3,2.3,virginica6.5,3.0,5.5,1.8,virginica7.7,3.8,6.7,2.2,virginica7.7,2.6,6.9,2.3,virginica6.0,2.2,5.0,1.5,virginica6.9,3.2,5.7,2.3,virginica5.6,2.8,4.9,2.0,virginica7.7,2.8,6.7,2.0,virginica6.3,2.7,4.9,1.8,virginica6.7,3.3,5.7,2.1,virginica7.2,3.2,6.0,1.8,virginica6.2,2.8,4.8,1.8,virginica6.1,3.0,4.9,1.8,virginica6.4,2.8,5.6,2.1,virginica7.2,3.0,5.8,1.6,virginica7.4,2.8,6.1,1.9,virginica7.9,3.8,6.4,2.0,virginica6.4,2.8,5.6,2.2,virginica6.3,2.8,5.1,1.5,virginica6.1,2.6,5.6,1.4,virginica7.7,3.0,6.1,2.3,virginica6.3,3.4,5.6,2.4,virginica6.4,3.1,5.5,1.8,virginica6.0,3.0,4.8,1.8,virginica6.9,3.1,5.4,2.1,virginica6.7,3.1,5.6,2.4,virginica6.9,3.1,5.1,2.3,virginica5.8,2.7,5.1,1.9,virginica6.8,3.2,5.9,2.3,virginica6.7,3.3,5.7,2.5,virginica6.7,3.0,5.2,2.3,virginica6.3,2.5,5.0,1.9,virginica6.5,3.0,5.2,2.0,virginica6.2,3.4,5.4,2.3,virginica5.9,3.0,5.1,1.8,virginicasetosa,versicolor,virginica6.5,2.8,4.6,1.57.7,3.8,6.7,2.26.7,3.1,4.4,1.45.8,2.7,4.1,1.06.3,2.5,5.0,1.96.3,3.3,4.7,1.66.5,3.0,5.8,2.25.4,3.4,1.7,0.26.3,2.3,4.4,1.36.1,3.0,4.6,1.47.2,3.0,5.8,1.66.7,3.1,4.7,1.55.1,3.4,1.5,0.26.2,2.8,4.8,1.85.1,3.8,1.6,0.26.0,2.7,5.1,1.65.4,3.9,1.3,0.44.6,3.2,1.4,0.25.0,3.4,1.6,0.45.0,3.3,1.4,0.25.1,3.8,1.5,0.36.5,3.0,5.2,2.05.8,2.6,4.0,1.26.3,2.8,5.1,1.56.3,3.4,5.6,2.44.9,3.1,1.5,0.15.6,3.0,4.1,1.36.2,3.4,5.4,2.34.8,3.0,1.4,0.34.9,3.1,1.5,0.16.8,2.8,4.8,1.45.4,3.4,1.5,0.45.0,3.5,1.6,0.66.7,3.3,5.7,2.15.0,3.0,1.6,0.25.6,2.8,4.9,2.06.4,2.8,5.6,2.26.7,3.3,5.7,2.54.8,3.1,1.6,0.25.0,2.3,3.3,1.05.2,3.5,1.5,0.24.3,3.0,1.1,0.16.4,2.7,5.3,1.95.8,4.0,1.2,0.25.7,2.5,5.0,2.05.5,4.2,1.4,0.26.5,3.0,5.5,1.87.4,2.8,6.1,1.96.5,3.2,5.1,2.04.4,2.9,1.4,0.25.2,3.4,1.4,0.25.2,4.1,1.5,0.16.8,3.0,5.5,2.15.9,3.0,4.2,1.56.9,3.1,4.9,1.56.7,3.0,5.0,1.76.8,3.2,5.9,2.36.3,3.3,6.0,2.56.3,2.7,4.9,1.87.3,2.9,6.3,1.86.1,2.8,4.0,1.37.1,3.0,5.9,2.16.3,2.9,5.6,1.86.9,3.1,5.4,2.15.1,3.3,1.7,0.55.0,3.4,1.5,0.26.3,2.5,4.9,1.56.6,3.0,4.4,1.47.2,3.6,6.1,2.55.5,2.4,3.8,1.15.4,3.7,1.5,0.25.4,3.0,4.5,1.56.1,3.0,4.9,1.85.2,2.7,3.9,1.46.6,2.9,4.6,1.34.4,3.2,1.3,0.25.7,2.8,4.5,1.37.7,3.0,6.1,2.34.7,3.2,1.3,0.24.6,3.1,1.5,0.24.4,3.0,1.3,0.24.6,3.4,1.4,0.36.1,2.8,4.7,1.26.0,2.2,5.0,1.54.8,3.0,1.4,0.14.9,2.4,3.3,1.05.5,2.3,4.0,1.35.6,2.5,3.9,1.14.7,3.2,1.6,0.25.8,2.7,5.1,1.96.4,3.1,5.5,1.85.7,4.4,1.5,0.45.7,2.6,3.5,1.05.4,3.9,1.7,0.46.9,3.1,5.1,2.35.6,2.7,4.2,1.35.8,2.7,5.1,1.96.2,2.9,4.3,1.34.8,3.4,1.9,0.25.5,2.4,3.7,1.05.7,3.0,4.2,1.26.0,2.9,4.5,1.56.1,2.9,4.7,1.44.5,2.3,1.3,0.37.6,3.0,6.6,2.15.6,3.0,4.5,1.56.7,2.5,5.8,1.85.5,2.5,4.0,1.35.1,3.5,1.4,0.35.3,3.7,1.5,0.25.6,2.9,3.6,1.37.2,3.2,6.0,1.84.9,2.5,4.5,1.76.4,3.2,4.5,1.55.0,3.5,1.3,0.36.4,3.2,5.3,2.35.0,2.0,3.5,1.07.7,2.6,6.9,2.35.1,3.5,1.4,0.26.0,2.2,4.0,1.0versicolor,virginica,versicolor,versicolor,virginica,versicolor,virginica,setosa,versicolor,versicolor,virginica,versicolor,setosa,virginica,setosa,versicolor,setosa,setosa,setosa,setosa,setosa,virginica,versicolor,virginica,virginica,setosa,versicolor,virginica,setosa,setosa,versicolor,setosa,setosa,virginica,setosa,virginica,virginica,virginica,setosa,versicolor,setosa,setosa,virginica,setosa,virginica,setosa,virginica,virginica,virginica,setosa,setosa,setosa,virginica,versicolor,versicolor,versicolor,virginica,virginica,virginica,virginica,versicolor,virginica,virginica,virginica,setosa,setosa,versicolor,versicolor,virginica,versicolor,setosa,versicolor,virginica,versicolor,versicolor,setosa,versicolor,virginica,setosa,setosa,setosa,setosa,versicolor,virginica,setosa,versicolor,versicolor,versicolor,setosa,virginica,virginica,setosa,versicolor,setosa,virginica,versicolor,virginica,versicolor,setosa,versicolor,versicolor,versicolor,versicolor,setosa,virginica,versicolor,virginica,versicolor,setosa,setosa,versicolor,virginica,virginica,versicolor,setosa,virginica,versicolor,virginica,setosa,versicolor4.321.10.13.40000000000000042.40000000000000045.8000000000000012.40.6470588235294118,0.3333333333333332,0.6034482758620688,0.5833333333333334,21,0.7499999999999998,0.9655172413793102,0.8750000000000001,30.7058823529411765,0.4583333333333333,0.5689655172413793,0.5416666666666666,20.4411764705882352,0.2916666666666667,0.5172413793103446,0.375,20.588235294117647,0.20833333333333331,0.6724137931034482,0.75,30.588235294117647,0.5416666666666665,0.6206896551724137,0.625,20.6470588235294118,0.41666666666666663,0.8103448275862066,0.8750000000000001,30.323529411764706,0.5833333333333333,0.10344827586206894,0.04166666666666667,10.588235294117647,0.1249999999999999,0.5689655172413793,0.5,20.5294117647058822,0.41666666666666663,0.6034482758620688,0.5416666666666666,20.8529411764705882,0.41666666666666663,0.8103448275862066,0.625,30.7058823529411765,0.4583333333333333,0.6206896551724137,0.5833333333333334,20.23529411764705874,0.5833333333333333,0.06896551724137928,0.04166666666666667,10.5588235294117647,0.3333333333333332,0.6379310344827585,0.7083333333333334,30.23529411764705874,0.7499999999999998,0.08620689655172413,0.04166666666666667,10.5,0.2916666666666667,0.6896551724137929,0.625,20.323529411764706,0.7916666666666665,0.03448275862068964,0.12500000000000003,10.088235294117647,0.5,0.05172413793103445,0.04166666666666667,10.2058823529411765,0.5833333333333333,0.08620689655172413,0.12500000000000003,10.2058823529411765,0.5416666666666665,0.05172413793103445,0.04166666666666667,10.23529411764705874,0.7499999999999998,0.06896551724137928,0.08333333333333333,10.6470588235294118,0.41666666666666663,0.7068965517241378,0.7916666666666666,30.4411764705882352,0.25,0.49999999999999994,0.4583333333333333,20.588235294117647,0.3333333333333332,0.6896551724137929,0.5833333333333334,30.588235294117647,0.5833333333333333,0.7758620689655171,0.9583333333333333,30.17647058823529427,0.4583333333333333,0.06896551724137928,0,10.3823529411764705,0.41666666666666663,0.5172413793103446,0.5,20.5588235294117647,0.5833333333333333,0.7413793103448276,0.9166666666666666,30.14705882352941174,0.41666666666666663,0.05172413793103445,0.08333333333333333,10.17647058823529427,0.4583333333333333,0.06896551724137928,0,10.7352941176470588,0.3333333333333332,0.6379310344827585,0.5416666666666666,20.323529411764706,0.5833333333333333,0.06896551724137928,0.12500000000000003,10.2058823529411765,0.6249999999999999,0.08620689655172413,0.20833333333333334,10.7058823529411765,0.5416666666666665,0.7931034482758619,0.8333333333333334,30.2058823529411765,0.41666666666666663,0.08620689655172413,0.04166666666666667,10.3823529411764705,0.3333333333333332,0.6551724137931034,0.7916666666666666,30.6176470588235295,0.3333333333333332,0.7758620689655171,0.8750000000000001,30.7058823529411765,0.5416666666666665,0.7931034482758619,1,30.14705882352941174,0.4583333333333333,0.08620689655172413,0.04166666666666667,10.2058823529411765,0.1249999999999999,0.37931034482758613,0.375,20.26470588235294124,0.6249999999999999,0.06896551724137928,0.04166666666666667,10,0.41666666666666663,0,0,10.6176470588235295,0.2916666666666667,0.7241379310344825,0.75,30.4411764705882352,0.8333333333333333,0.017241379310344803,0.04166666666666667,10.411764705882353,0.20833333333333331,0.6724137931034482,0.7916666666666666,30.35294117647058826,0.9166666666666666,0.05172413793103445,0.04166666666666667,10.6470588235294118,0.41666666666666663,0.7586206896551724,0.7083333333333334,30.911764705882353,0.3333333333333332,0.8620689655172413,0.75,30.6470588235294118,0.5,0.6896551724137929,0.7916666666666666,30.029411764705882505,0.3749999999999999,0.05172413793103445,0.04166666666666667,10.26470588235294124,0.5833333333333333,0.05172413793103445,0.04166666666666667,10.26470588235294124,0.8749999999999998,0.06896551724137928,0,10.7352941176470588,0.41666666666666663,0.7586206896551724,0.8333333333333334,30.47058823529411775,0.41666666666666663,0.5344827586206896,0.5833333333333334,20.7647058823529412,0.4583333333333333,0.6551724137931034,0.5833333333333334,20.7058823529411765,0.41666666666666663,0.6724137931034482,0.6666666666666666,20.7352941176470588,0.5,0.8275862068965517,0.9166666666666666,30.588235294117647,0.5416666666666665,0.8448275862068965,1,30.588235294117647,0.2916666666666667,0.6551724137931034,0.7083333333333334,30.8823529411764705,0.3749999999999999,0.8965517241379308,0.7083333333333334,30.5294117647058822,0.3333333333333332,0.49999999999999994,0.5,20.8235294117647057,0.41666666666666663,0.8275862068965517,0.8333333333333334,30.588235294117647,0.3749999999999999,0.7758620689655171,0.7083333333333334,30.7647058823529412,0.4583333333333333,0.7413793103448276,0.8333333333333334,30.23529411764705874,0.5416666666666665,0.10344827586206894,0.16666666666666669,10.2058823529411765,0.5833333333333333,0.06896551724137928,0.04166666666666667,10.588235294117647,0.20833333333333331,0.6551724137931034,0.5833333333333334,20.676470588235294,0.41666666666666663,0.5689655172413793,0.5416666666666666,20.8529411764705882,0.6666666666666666,0.8620689655172413,1,30.35294117647058826,0.1666666666666666,0.4655172413793102,0.4166666666666667,20.323529411764706,0.7083333333333333,0.06896551724137928,0.04166666666666667,10.323529411764706,0.41666666666666663,0.5862068965517241,0.5833333333333334,20.5294117647058822,0.41666666666666663,0.6551724137931034,0.7083333333333334,30.26470588235294124,0.2916666666666667,0.4827586206896551,0.5416666666666666,20.676470588235294,0.3749999999999999,0.6034482758620688,0.5,20.029411764705882505,0.5,0.03448275862068964,0.04166666666666667,10.411764705882353,0.3333333333333332,0.5862068965517241,0.5,21,0.41666666666666663,0.8620689655172413,0.9166666666666666,30.11764705882352951,0.5,0.03448275862068964,0.04166666666666667,10.088235294117647,0.4583333333333333,0.06896551724137928,0.04166666666666667,10.029411764705882505,0.41666666666666663,0.03448275862068964,0.04166666666666667,10.088235294117647,0.5833333333333333,0.05172413793103445,0.08333333333333333,10.5294117647058822,0.3333333333333332,0.6206896551724137,0.4583333333333333,20.5,0.0833333333333334,0.6724137931034482,0.5833333333333334,30.14705882352941174,0.41666666666666663,0.05172413793103445,0,10.17647058823529427,0.1666666666666666,0.37931034482758613,0.375,20.35294117647058826,0.1249999999999999,0.49999999999999994,0.5,20.3823529411764705,0.20833333333333331,0.4827586206896551,0.4166666666666667,20.11764705882352951,0.5,0.08620689655172413,0.04166666666666667,10.4411764705882352,0.2916666666666667,0.6896551724137929,0.75,30.6176470588235295,0.4583333333333333,0.7586206896551724,0.7083333333333334,30.411764705882353,1,0.06896551724137928,0.12500000000000003,10.411764705882353,0.25,0.4137931034482758,0.375,20.323529411764706,0.7916666666666665,0.10344827586206894,0.12500000000000003,10.7647058823529412,0.4583333333333333,0.6896551724137929,0.9166666666666666,30.3823529411764705,0.2916666666666667,0.5344827586206896,0.5,20.4411764705882352,0.2916666666666667,0.6896551724137929,0.75,30.5588235294117647,0.3749999999999999,0.5517241379310344,0.5,20.14705882352941174,0.5833333333333333,0.13793103448275856,0.04166666666666667,10.35294117647058826,0.1666666666666666,0.44827586206896547,0.375,20.411764705882353,0.41666666666666663,0.5344827586206896,0.4583333333333333,20.5,0.3749999999999999,0.5862068965517241,0.5833333333333334,20.5294117647058822,0.3749999999999999,0.6206896551724137,0.5416666666666666,20.058823529411764754,0.1249999999999999,0.03448275862068964,0.08333333333333333,10.9705882352941175,0.41666666666666663,0.9482758620689654,0.8333333333333334,30.3823529411764705,0.41666666666666663,0.5862068965517241,0.5833333333333334,20.7058823529411765,0.20833333333333331,0.8103448275862066,0.7083333333333334,30.35294117647058826,0.20833333333333331,0.49999999999999994,0.5,20.23529411764705874,0.6249999999999999,0.05172413793103445,0.08333333333333333,10.2941176470588235,0.7083333333333333,0.06896551724137928,0.04166666666666667,10.3823529411764705,0.3749999999999999,0.43103448275862066,0.5,20.8529411764705882,0.5,0.8448275862068965,0.7083333333333334,30.17647058823529427,0.20833333333333331,0.5862068965517241,0.6666666666666666,30.6176470588235295,0.5,0.5862068965517241,0.5833333333333334,20.2058823529411765,0.6249999999999999,0.03448275862068964,0.08333333333333333,10.6176470588235295,0.5,0.7241379310344825,0.9166666666666666,30.2058823529411765,0,0.4137931034482758,0.375,21,0.25,1,0.9166666666666666,30.23529411764705874,0.6249999999999999,0.05172413793103445,0.04166666666666667,10.5,0.0833333333333334,0.49999999999999994,0.375,2