2b:null 2c:["$","title",null,{"children":"Kaggle Titanic Data Exploration | Jose Jimenez | Observable"}] 31:T49f,multi_layer_perceptron = { if (false){ runMLP; const mlp = new ml.MLP({ 'input' : training_data.inputs, 'label' : training_data.outputs, 'n_ins' : training_data.inputs[0].length, 'n_outs' : training_data.outputs[0].length, 'hidden_layer_sizes': [training_data.inputs[0].length / 8 * 7,training_data.inputs[0].length / 6 * 3,training_data.inputs[0].length / 8 * 6] }); mlp.set('log level',1); // 0 : nothing, 1 : info, 2 : warning. mlp.train({ 'lr' : 0.6, 'epochs' : 100 }); const predictions = mlp.predict(testing_data.inputs); const diffs = predictions.map( (p,i) => testing_data.outputs[i][0] - p[0]); const averageDiff = diffs.reduce( (a,d) => a+d) / diffs.length; const normalized_predictions = predictions.map( (p) => Math.round(p[0])); const num_right_answers = normalized_predictions.filter( (p,i) => (testing_data.outputs[i][0] - p) == 0).length / testing_data.outputs.length; return ({RightAnswers: num_right_answers, guess: predictions, actual: testing_data.outputs, diffs: diffs, averageDiff: averageDiff}); } else { return 'edit the false at the top of cell\'s code to true to run this it averages ~62% correct'; } }32:T447,multi_layer_perceptron_fewer_layers = { if (false){ runMLP; const mlp = new ml.MLP({ 'input' : training_data.inputs, 'label' : training_data.outputs, 'n_ins' : training_data.inputs[0].length, 'n_outs' : training_data.outputs[0].length, 'hidden_layer_sizes': hidden_layers }); mlp.set('log level',1); // 0 : nothing, 1 : info, 2 : warning. mlp.train({ 'lr' : 0.2, 'epochs' : 3000 }); const predictions = mlp.predict(testing_data.inputs); const diffs = predictions.map( (p,i) => testing_data.outputs[i][0] - p[0]); const averageDiff = diffs.reduce( (a,d) => a+d) / diffs.length; const normalized_predictions = predictions.map( (p) => Math.round(p[0])); const num_right_answers = normalized_predictions.filter( (p,i) => (testing_data.outputs[i][0] - p) == 0).length / testing_data.outputs.length; return ({RightAnswers: num_right_answers, guess: predictions, actual: testing_data.outputs, diffs: diffs, averageDiff: averageDiff}); } else { return 'edit the false at the top of cell\'s code to true to run this it averages ~82% correct'; } }33:T799,Normalizer_normalize = function (normalizer_data_store) { normalizer_data_store.datasetMeta = (normalizer_data_store.datasetMeta === null) ? Normalizer_analyzeMetaData(normalizer_data_store) : normalizer_data_store.datasetMeta; var binaryInput = []; var binaryOutput = []; for (var i in normalizer_data_store.dataset) { var row = normalizer_data_store.dataset[i]; var index = 0; var inputBits = []; var outputBits = []; for (var prop in row) { var bitsArr = void 0; var value = row[prop]; var meta = normalizer_data_store.datasetMeta[prop]; switch (meta.type) { case 'number': bitsArr = [Normalizer_numToBit(meta.min, meta.max, value)]; break; case 'boolean': bitsArr = [Normalizer_boolToBit(value)]; break; case 'string': bitsArr = Normalizer_strToBitsArr(meta.distinctValues, value); break; case 'array': bitsArr = Normalizer_arrToBitsArr(meta.distinctValues, value); break; default: break; } if (normalizer_data_store.outputProperties.indexOf(prop) > -1) { outputBits = outputBits.concat(bitsArr); } else { inputBits = inputBits.concat(bitsArr); } index++; } if (inputBits.length > 0) { normalizer_data_store.binaryInput.push(inputBits); } if (outputBits.length > 0) { normalizer_data_store.binaryOutput.push(outputBits); } } } 34:T641,Normalizer_analyzeMetaData = function (normalize_data_store) { var firstRow = normalize_data_store.dataset[0]; var distinctProps = Normalizer_distinctProps(firstRow); var distinctTypes = Normalizer_distinctTypes(firstRow); var metadata = {}; var bitDataset = []; for (var _i = 0, distinctProps_1 = distinctProps; _i < distinctProps_1.length; _i++) { var prop = distinctProps_1[_i]; var type = distinctTypes[prop]; metadata[prop] = { type: type, min: null, max: null, distinctValues: null, }; switch (type) { case 'number': var minMax = Normalizer_getMinMax(prop, normalize_data_store.dataset); metadata[prop].min = minMax[0]; metadata[prop].max = minMax[1]; break; case 'boolean': metadata[prop].min = 0; metadata[prop].max = 1; break; case 'string': var distinctStrVals = Normalizer_getDistinctVals(prop, normalize_data_store.dataset); metadata[prop].distinctValues = distinctStrVals; break; case 'array': var distinctArrVals = Normalizer_getDistinctArrayVals(prop, normalize_data_store.dataset); metadata[prop].distinctValues = distinctArrVals; break; } } return metadata; } 35:[] 2f:{"id":"40b20e6c348fa9de","slug":"kaggle-titanic-data-exploration","title":"Kaggle Titanic Data Exploration","thumbnail":null,"access_level":"public","create_time":"$D2018-01-31T17:18:07.432Z","update_time":"$D2018-03-21T17:01:00.770Z","view_time":null,"first_public_time":"$D2018-01-31T18:15:13.639Z","fork":false,"stars":2,"starred":false,"owner":{"type":"team","id":"11eff763c21b0530","login":"jjimenez","name":"Jose Jimenez","avatar_url":"https://avatars-next.observableusercontent.com/5d28ac3054c420a200dcd8a0bd3367e2263d2f168c4aa5eaa0c89640622d6088","tier":"starter_2024"},"version":969,"publish_version":969,"role":null,"authored":false,"authors":[{"type":"individual","id":"214adfe1a708eb31","login":"jjimenez","name":"Jose Jimenez","avatar_url":"https://avatars-next.observableusercontent.com/5d28ac3054c420a200dcd8a0bd3367e2263d2f168c4aa5eaa0c89640622d6088","tier":null}],"members":[{"type":"individual","id":"214adfe1a708eb31","login":"jjimenez","name":"Jose Jimenez","avatar_url":"https://avatars-next.observableusercontent.com/5d28ac3054c420a200dcd8a0bd3367e2263d2f168c4aa5eaa0c89640622d6088","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"subscribed":null,"forks":4,"collection":null,"type":"notebook","body":{"title":"Kaggle Titanic Data Exploration","cells":[{"id":0,"value":"md`# Kaggle Titanic Data Exploration`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":327,"value":"md`### Introduction\n\nI'm working on the [kaggle training competition](https://www.kaggle.com/c/titanic) and learning more about the Observable platform. Here I'm also using [Vega-Lite](https://vega.github.io/vega-lite/) for the first time. That is a lot of new things all at once. Did I mention I'm not great at markdown either?\n\nCheck out the [Observable training](https://beta.observablehq.com/collection/introduction) first if you aren't familiar with Observable. You can edit this notebook in place to try your hand at changing the visualizations. Your changes will be reverted once you refresh the page. If you want to play with this for a while, fork the notebook and you will have your own copy to update as you please.\n\nThere are two major sections to this notebook. First you will see the Visualizations section. \nIf you are wondering where the data comes into the notebook, look for the second section below called \"Data and Requires.\"`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":63,"value":"md`## Visualizations`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":65,"value":"md`### Age distribution of passengers\n\nHere is a very simple visualization of the distribution of the ages of the passengers in the training dataset. Remember to click on the area to the left of the cell to see the code for it. Click it again to hide the code.\nFor example, change the line\n\n\n x: {field: \"Age\", type: variables['Age'].type}\n\nto be \n\n x: {field: \"Fare\", type: variables['Fare'].type}\n\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":12,"value":"vegalite({\n data: {values: data},\n mark: \"tick\",\n encoding: {\n x: {field: \"Age\", type: variables['Age'].type}\n }\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":67,"value":"md`### Passenger count by Selected Prime Variable\n\nHere I build out a select box using tagged template literals and html options based on an array in the Data Section. \n\nThen I use that select to determine what data to plot using vega lite.\n\nExplore the data by changing the selected variable to plot and examining how that variable impacts survival. For example, note that after selecting the varialble \"Sex\", \"female\" has a higher likelyhood of survival.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":224,"value":"viewof prime = html``","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":373,"value":"md`In the code here I need to add a couple of calculated columns. Note that this is just an object, so I do it in the data section once so I can reuse it.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":19,"value":"vegalite({\n \"width\": width - 50,\n \"autosize\": {\n \"type\": \"fit\",\n \"contains\": \"padding\"\n },\n title: `Count of Passengers by ${variables[prime].label}`,\n data: {values: data_with_title},\n mark: \"bar\",\n transform: transform_settings, \n encoding: {\n x: {field: prime, type: variables[prime].type, bin: variables[prime].bin},\n y: {aggregate: \"count\", field: \"*\", type: \"quantitative\"},\n color: { field: 'Survived', type: 'nominal'}\n }\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":309,"value":"md`### Confidence Interval Lower Bound for Selected Prime Variable\n\nHere is the [confidence interval](https://en.wikipedia.org/wiki/Confidence_interval), or, more precisely from the Vega-Lite Documentation, \"The lower boundary of the bootstrapped 95% confidence interval of the mean field value\" for the Selected Prime Variable. Again, if you have selected \"Sex\", note that the confidence interval is close to 0.7. Try other variables to see their confidence interval`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":307,"value":"vegalite({\"width\": width - 50,\n \"autosize\": {\n \"type\": \"fit\",\n \"contains\": \"padding\"\n },\n title: `95% confidence interval for ${variables[prime].label}`,\n data: {values: data_with_title},\n mark: \"bar\",\n transform: transform_settings, \n encoding: {\n x: {field: prime, type: variables[prime].type, bin: variables[prime].bin},\n y: {aggregate: \"ci0\", field: \"Survived\", type: \"quantitative\"}\n }\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":397,"value":"md`If you selected \"honorific\" as the variable, it looks like the confidence level is pretty high for some honorifics such as \"Mlle\", \"Ms\", and \"Sir\". These might look like great variables to use for predicting survivors. Taking a look at the chart above this one, it seems there were very few individuals with those honorifics. Using these may turn out to be a case of \"[overfitting](https://en.wikipedia.org/wiki/Overfitting).\"`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":69,"value":"md`### Survivability by Prime and Comparison Variable\n\nOften, when trying to determine what variables have significance in a dataset, it is important to review combinations of variables together. This chart compares the variable selected against the comparison variable.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":409,"value":"viewof variable = html``","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":128,"value":"viewof comparison = html``","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":26,"value":"vegalite({\n \"autosize\": {\n \"type\": \"fit\",\n \"contains\": \"padding\"\n },\n \"width\": width - 50,\n \"height\": 300,\n title: `Survivability by ${variables[variable].label} and ${variables[comparison].label}`,\n \"data\": {values: data_with_title},\n \"mark\": \"circle\", \n transform: transform_settings, \n \"encoding\": {\n \"x\": {\"field\": variable, \"type\": \"Ordinal\", bin: variables[variable].bin},\n \"y\": {\"field\": comparison, \"type\": \"Ordinal\", bin: variables[comparison].bin},\n \"color\": {\"field\": \"Survived\", \"type\": \"nominal\" },\n \"size\": { \"aggregate\": \"count\", \"field\": \"*\", \"type\": \"quantitative\"}\n }\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":325,"value":"md`### Age vs Several Variables\n\nI originally wanted to use the [scatter plot matrix example here](https://vega.github.io/vega-lite/docs/repeat.html#example-scatterplot-matrix-splom), but I couldn't figure out how to use the variables object and the defined bin, and type I have defined in the data section. So I manually created something along the lines of the scatter plot matrix using the repeat column for the variables I wanted.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":319,"value":"vegalite(first_row)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":321,"value":"vegalite(second_row)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":323,"value":"vegalite(third_row)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":638,"value":"md`## Machine Learning`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":767,"value":"viewof runMLP = button({value: \"Rerun MLP Guess\", description: \"Rerun the Multi Layer Perceptron guess below\"})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":640,"value":"$31","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":886,"value":"hidden_layers = [Math.floor(training_data.inputs[0].length / 8 * 5),Math.floor(training_data.inputs[0].length / 8 * 2) ]","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":834,"value":"$32","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":929,"value":"support_vector_machine = { \nreturn 'this is broken';\n runMLP;\n \n var x = training_data.inputs;\n\nvar y = training_data.outputs.map((t) => t[0]);\nvar svm = new ml.SVM({\n x : x,\n y : y\n});\n \nsvm.train({\n C : 1.1, // default : 1.0. C in SVM.\n tol : 1e-5, // default : 1e-4. Higher tolerance --> Higher precision\n max_passes : 20, // default : 20. Higher max_passes --> Higher precision\n alpha_tol : 1e-5, // default : 1e-5. Higher alpha_tolerance --> Higher precision\n kernel : { type: \"polynomial\", c: 1, d: 5}\n});\nreturn testing_data.inputs.map((d) => svm.predict(d));\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":74,"value":"md`## More Soon!`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":59,"value":"md`## Data and Requires`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":647,"value":"import {button} from \"@jashkenas/inputs\"","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":417,"value":"md`A function to niavely get the honorific from the name of a passenger. \"the\" ends up being one of them.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":205,"value":"function getTitle(name) {\n var parts = name.split(', ');\n var second = parts[1];\n var title = second.split(' ')[0];\nreturn title.replace(/\\./, '');\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":420,"value":"md`A reusable object to setup the for the mock scatter plot matrix`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":311,"value":"scatter_plot_setup = ({\n \"data\": {values: data_with_title},\n spec: {\n \"mark\": \"circle\", \n transform: transform_settings, \n \"encoding\": {\n \"x\": {\"field\": 'Age', \"type\": \"Ordinal\", bin: variables['Age'].bin},\n \"y\": {\"field\": {repeat: 'column'}, \"type\": \"nominal\", bin: false},\n \"color\": {\"field\": \"Survived\", \"type\": \"nominal\" },\n \"size\": { \"aggregate\": \"count\", \"field\": \"*\", \"type\": \"quantitative\"}\n }\n }\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":422,"value":"md`Calculated columns for the data. I should probably do the same with the data`\n","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":347,"value":"transform_settings = [\n {calculate: \"datum.Cabin.length > 0 ? substring(datum.Cabin,0,1) : '?'\", as: 'CabinPrefix'},\n {calculate: \"floor(parseFloat(datum.Fare))\", as:'BaseFare'}\n ]","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":425,"value":"md`These next few variables identify the columns for the mock scatterplot matrix`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":313,"value":"first_row = (function () { var new_val = Object.assign({},scatter_plot_setup); new_val.repeat = {column: ['Pclass', 'Sex', 'SibSp']}; return new_val;})()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":315,"value":"second_row = (function () { var new_val = Object.assign({},scatter_plot_setup); new_val.repeat = {column: ['Parch', 'CabinPrefix']}; return new_val;})()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":317,"value":"third_row = (function () { var new_val = Object.assign({},scatter_plot_setup); new_val.repeat = {column: ['Embarked', 'Title']}; return new_val;})()","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":429,"value":"md`A list of the variables to use`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":215,"value":"variables = ({\n Pclass: {field: 'Pclass', bin: false, type: 'ordinal', label: 'Passenger Class'},\n Age: { field: 'Age', bin: true, type: 'quantitative', label: 'Age'},\n Sex: { field: 'Sex', bin: false, type: 'nominal', label: 'Sex'},\n SibSp: { field: 'SibSp', bin: false, type: 'quantitative', label: 'Count of Siblings / Spouses On Board'},\n Parch: { field: 'Parch', bin: false, type: 'quantitative', label: 'Count of Parents / Children On Board'},\n CabinPrefix: { field: 'CabinPrefix', bin: false, type: 'nominal', label: 'Prefix of Cabin'},\n BaseFare: { field: 'BaseFare', bin: true, type: 'quantitative', label: 'Rounded Fare'},\n Embarked: { field: 'Embarked', bin: false, type: 'nominal', label: 'Point of Embarkation'},\n Title: { field: 'Title', bin: false, type: 'nominal', label: 'Honorific'},\n Fare: { field: 'Fare', bin: true, type: 'quantitative', label: 'Fare Paid'},\n})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":431,"value":"md`The options to use in the drop downs are generated clumsily here. Is there a better way?`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":222,"value":"options_from_variables = Object.keys(variables).map(x => ``).join('')","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":436,"value":"md`Here we load the data from a public gist I created from the original Kaggle data file. I rely on the reactiveness of observable and don't worry about how long it takes to load the data. When it loads, the data will display in the appropriate cells`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":6,"value":"data = d3.csv(\"https://cdn.rawgit.com/jjimenez/ea74f864fd34b6e177b416a325daa232/raw/76a64a7f899c7f001313fa0b864975564da80e58/KaggleTitanicTrain.csv\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":621,"value":"data_count = data.length","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":615,"value":"md`this data will be used as training data for machine learning`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":623,"value":"training_data = ({inputs: ml_data.inputs.slice(0,data_count / 2), outputs: ml_data.outputs.slice(0, data_count / 2) })","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":625,"value":"testing_data = ({inputs: ml_data.inputs.slice(data_count / 2), outputs: ml_data.outputs.slice(data_count / 2) })","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":450,"value":"ml_data = { \n const select_data = data.map((d) => ({ Survived: (d.Survived == 1 ? true : false), Pclass: d.Pclass, Title: getTitle(d.Name), Sex: d.Sex, Age: d.Age == \"\" ? -1.0 : parseFloat(d.Age), SibSp: parseInt(d.SibSp), Parch: parseInt(d.Parch), Fare: parseFloat(d.Fare), CabinPrefix: d.Cabin.length > 0 ? d.Cabin.substring(0,1) : '?', Embarked: d.Embarked }))\n\n const ml_data_normalizer = Normalizer(normalizer_store,select_data);\n Normalizer_setOutputProperties(normalizer_store, 'Survived');\n Normalizer_normalize(normalizer_store);\n return {\n metadata: Normalizer_getDatasetMetaData(normalizer_store),\n inputs: Normalizer_getBinaryInputDataset(normalizer_store),\n outputs: Normalizer_getBinaryOutputDataset(normalizer_store)\n }\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":438,"value":"md`I need the data to have at least one additional field, so I compute a new data set here. I'm trying not to mutate the original data set.`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":207,"value":"data_with_title = data.map(function(d) { d.Title = getTitle(d.Name); return d })","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":440,"value":"md`Require D3. \n\nAgain, I don't worry about waiting because Observable is awesome`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":10,"value":"d3 = require(\"https://d3js.org/d3.v5.min.js\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":445,"value":"md`Require the Vega-Lite library`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":4,"value":"vegalite = require(\"@observablehq/vega-lite@0.0\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":459,"value":"function isArray(input) {\n return (Object.prototype.toString.call(input) === '[object Array]') ? true : false;\n}","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":604,"value":"md`## Normalizer`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":474,"value":"md`I couldn't use unpkg, or wzrd.in, or any other tricks to load [neural-data-normalizer](https://www.npmjs.com/package/neural-data-normalizer), so I copied the code here and refactored it to work with the notebook.` ","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":485,"value":"normalizer_store = ({})","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":498,"value":"function Normalizer(normalizer_data_store, data) {\n if (data === void 0) { data = []; }\n normalizer_data_store.dataset = [];\n normalizer_data_store.datasetMeta = null;\n normalizer_data_store.binaryInput = [];\n normalizer_data_store.binaryOutput = [];\n normalizer_data_store.outputProperties = [];\n normalizer_data_store.dataset = data;\n if (normalizer_data_store.dataset.length <= 0) {\n throw new Error(\"Normalizer input data shouldn't be empty\");\n }\n if (Object.keys(normalizer_data_store.dataset[0]).length <= 0) {\n throw new Error(\"Normalizer input data rows has to contain some properties (only 1st row is checked)\");\n }\n }","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":503,"value":"Normalizer_getOutputLength = (normalizer_data_store) => normalizer_data_store.outputProperties.length\n ","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":508,"value":" Normalizer_getOutputProperties = (normalizer_data_store) => normalizer_data_store.outputProperties","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":511,"value":"Normalizer_getInputLength = (normalizer_data_store) => normalizer_data_store.binaryInput[0].length","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":515,"value":"Normalizer_getBinaryInputDataset = (normalizer_data_store) => normalizer_data_store.binaryInput","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":519,"value":"Normalizer_getBinaryOutputDataset = (normalizer_data_store) => normalizer_data_store.binaryOutput","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":522,"value":"Normalizer_getDatasetMetaData = (normalizer_data_store) => normalizer_data_store.datasetMeta","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":525,"value":"Normalizer_setDatasetMetaData = (normalizer_data_store, metadata) => normalizer_data_store.datasetMeta = metadata","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":528,"value":"Normalizer_convertOutput = (normalizer_data_store) => normalizer_data_store.datasetMeta","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":531,"value":"$33","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":540,"value":"$34","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":545,"value":"Normalizer_setOutputProperties = (normalize_data_store, props) => normalize_data_store.outputProperties = props","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":548,"value":"Normalizer_getMinMax = function (prop, data) {\n var min = null;\n var max = null;\n for (var i in data) {\n var val = data[i][prop];\n if (min === null || val < min) {\n min = val;\n }\n if (max === null || val > max) {\n max = val;\n }\n }\n return [min, max];\n }","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":552,"value":"Normalizer_getDistinctVals = function (property, data) {\n var count = 0;\n var distinctValues = [];\n for (var _i = 0, data_1 = data; _i < data_1.length; _i++) {\n var row = data_1[_i];\n var val = row[property];\n if (distinctValues.indexOf(val) === -1) {\n distinctValues.push(val);\n }\n }\n return distinctValues;\n }","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":555,"value":"Normalizer_getDistinctArrayVals = function (property, data) {\n var count = 0;\n var distinctValues = [];\n for (var _i = 0, data_2 = data; _i < data_2.length; _i++) {\n var row = data_2[_i];\n var arrVal = row[property];\n for (var _a = 0, arrVal_1 = arrVal; _a < arrVal_1.length; _a++) {\n var val = arrVal_1[_a];\n if (distinctValues.indexOf(val) === -1) {\n distinctValues.push(val);\n }\n }\n }\n return distinctValues;\n }","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":558,"value":"\n Normalizer_numToBit = function (min, max, value) {\n var num = (value - min) / (max - min);\n return Number((num).toFixed(6));\n }","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":561,"value":" Normalizer_boolToBit = (val) => +val","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":564,"value":" Normalizer_strToBitsArr = function (distinctValues, val) {\n var bitArr = new Array(distinctValues.length);\n bitArr.fill(0);\n for (var i in distinctValues) {\n if (val === distinctValues[i]) {\n bitArr[i] = 1;\n }\n }\n return bitArr;\n }","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":567,"value":" Normalizer_arrToBitsArr = function (distinctValues, vals) {\n var bitArr = new Array(distinctValues.length);\n bitArr.fill(0);\n for (var j in vals) {\n var val = vals[j];\n var idx = distinctValues.indexOf(val);\n bitArr[idx] = 1;\n }\n return bitArr;\n }","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":570,"value":" Normalizer_distinctProps = (row) => Object.keys(row)","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":577,"value":"Normalizer_distinctTypes = function (row) {\n var distinctTypes = {};\n for (var prop in row) {\n var value = row[prop];\n if (typeof value === 'object' && isArray(value)) {\n distinctTypes[prop] = 'array';\n }\n else if (typeof value === 'object') {\n distinctTypes[prop] = 'object';\n }\n else {\n distinctTypes[prop] = typeof (value);\n }\n }\n return distinctTypes;\n }","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":468,"value":"Normalizer_getRow1stValue = (row) => row[Object.keys(row)[0]]\n","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":608,"value":"md`## Synaptic\n\nsee [the github repo](https://github.com/cazala/synaptic)\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":610,"value":"synaptic = require(\"https://cdnjs.cloudflare.com/ajax/libs/synaptic/1.1.4/synaptic.js\")","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"},{"id":618,"value":"ml = require('https://wzrd.in/standalone/machine_learning@0.1.1')","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined"}],"files":"$W35","stdlib":"1","titling":"auto"}} 30:"$2f:body"