Icon

Autoencoder MNIST MidPoint Focus

Exploring the latent space of an autoencoder for dimensional reduction

Keras Autoencoder Architecture Keras Autoencoder Architecture Data Preprocessing Training the Autoencoder Final Performance Training the Autoencoder Extracting midpoint Keras Autoencoder Architecture Final Performance 2D Midpoint 3D Midpoint Extracting midpoint Keras Dense Architecture 3D Midpoint DM Training the network Final Performance Extracting midpoint Extracting midpoint Final Performance Training the network Keras Dense Architecture 2D Midpoint DM DBSCAN CHECK Use Case DescriptionThe dataset is the well known MNIST dataset from Yan LeCun. 70000 grey levels(255) images sized 28x28Contains images of handwritten digits 0-9 and are classified with a 0-9 int classification column.We present two different approaches of using a neural network for dimension reductions, both of them with two branches for 3D and 2D representation. Plotly Chart StudioPrepare data to be imported in Plotly Keras Autoencoder Architecture Keras Dense Architecture DBSCAN CHECK DBSCAN CHECK DBSCAN CHECK Training the Autoencoder Keras Convolutional Architecture 3D Midpoint Extracting midpoint Final Performance Data preprocessing Write network modelTrain with Loss function=MSEOptimizer=AdamApply networkApply network forthe encoderEpsilonMinPtsDistance functionfor DBSCAN. EulideanClusteringNode 854Fetching data from OpenMLwith SciKitLearnNode 863Number of rowsNode 87275% for trainingMin-max normalization10 % for validationWritenormalizationmodelApply network forthe encoderWrite network modelApply networkNode 901One Hot vectorsClass to StringNode 905Apply networkTrain with Loss function=MSEOptimizer=AdamWrite network modelApply network forthe encoderNode 912Node 913Node 914Node 915Node 916Node 917Node 919Node 920Node 921Write network modelApply network forthe encoderNode 924Apply networkNode 926Train with Loss function=MSEOptimizer=AdamNode 933Node 936Node 952Node 955Node 956Node 957Node 958Node 959Node 961Node 962Node 963Node 965Node 966Node 968Colors fordifferentdigitsNode 970Node 971Node 972Node 974Node 975Node 976Node 977Node 978Node 979Node 980Node 981Node 982Node 983Node 984Class duplicatedto intNode 1020Node 1021MinPtsClusteringDistance functionfor DBSCAN. EulideanNode 1026EpsilonNode 1029Distance functionfor DBSCAN. EulideanClusteringNode 1032Node 1035Node 1036MinPtsNode 1038EpsilonClusteringNode 1041Distance functionfor DBSCAN. EuclideanNode 1045Node 1046EpsilonNode 1048MinPtsNode 1050Node 1051Node 1055Write network modelApply network forthe encoderNode 1067Node 1068Node 1069Node 1070Train with Loss function=MSEOptimizer=AdamNode 1075Node 1077Node 1078Node 1079Node 1080Node 1082Node 1083Node 1086Node 1087Node 1088Node 1089Node 1090Node 1091Node 1092Node 1093Node 1094Node 1095Node 1096Node 1097Keras NetworkLearner Keras NetworkWriter Keras NetworkLearner Keras NetworkExecutor Keras NetworkExecutor Double Input Integer Input Numeric Distances DBSCAN Merge Variables Python Source Scatter Plot(Plotly) Integer Input 3D ScatterPlot (Plotly) Keras In-5-Out Partitioning Normalizer Partitioning Model Writer Keras NetworkExecutor Keras NetworkWriter Keras NetworkExecutor MSE CalculationPython & Images Color Appender One to Many Number To String Column Resorter Keras In-5-Out(Indep.Act) Keras NetworkExecutor Keras NetworkLearner Keras NetworkWriter Keras NetworkExecutor Color Appender 3D ScatterPlot (Plotly) Many to One String Manipulation Scorer Scorer (JavaScript) Keras In-5-Out(Indep.Act) String Manipulation Scorer (JavaScript) Color Appender Keras NetworkWriter Keras NetworkExecutor Scorer Keras NetworkExecutor Many to One Keras NetworkLearner Column Filter Extract Color Python Script Color Modelto K.table Table Reader Sorter Sorter Sorter Sorter Math Formula(Multi Column) CSV Writer Partitioning Color Appender Column Filter Visualization Sorter Color Manager Color Appender Python View Python View Column Filter Column Filter Column Filter Python View Column Filter Scatter Plot(Plotly) Python View Keras NetworkReader Keras NetworkReader Keras NetworkReader Keras NetworkReader String Manipulation AssigningCluster/Digit Column Filter Merge Variables Integer Input DBSCAN Numeric Distances Color Appender Double Input AssigningCluster/Digit Sorter Numeric Distances DBSCAN Color Appender Visualization AssigningCluster/Digit Column Filter Merge Variables Integer Input Sorter Double Input DBSCAN Color Appender Numeric Distances Visualization Sorter Column Filter Double Input Merge Variables Integer Input Create CollectionColumn Data Row to Image Keras NetworkReader Keras NetworkWriter Keras NetworkExecutor Keras In-5-Out(Indep.Act) Many to One String Manipulation Scorer Scorer (JavaScript) AssigningCluster/Digit Keras In-5-Out Keras In-7-Out Keras NetworkLearner k-Means AssigningCluster/Digit Color Appender Sorter Column Filter k-Means AssigningCluster/Digit Color Appender Sorter Visualization Column Filter Python Script Math Formula(Multi Column) Table Reader CSV Writer Partitioning Column Rename String Manipulation Table Writer Table Editor Add Empty Rows Table Writer MSE CalculationPython & Images Keras Autoencoder Architecture Keras Autoencoder Architecture Data Preprocessing Training the Autoencoder Final Performance Training the Autoencoder Extracting midpoint Keras Autoencoder Architecture Final Performance 2D Midpoint 3D Midpoint Extracting midpoint Keras Dense Architecture 3D Midpoint DM Training the network Final Performance Extracting midpoint Extracting midpoint Final Performance Training the network Keras Dense Architecture 2D Midpoint DM DBSCAN CHECK Use Case DescriptionThe dataset is the well known MNIST dataset from Yan LeCun. 70000 grey levels(255) images sized 28x28Contains images of handwritten digits 0-9 and are classified with a 0-9 int classification column.We present two different approaches of using a neural network for dimension reductions, both of them with two branches for 3D and 2D representation. Plotly Chart StudioPrepare data to be imported in Plotly Keras Autoencoder Architecture Keras Dense Architecture DBSCAN CHECK DBSCAN CHECK DBSCAN CHECK Training the Autoencoder Keras Convolutional Architecture 3D Midpoint Extracting midpoint Final Performance Data preprocessing Write network modelTrain with Loss function=MSEOptimizer=AdamApply networkApply network forthe encoderEpsilonMinPtsDistance functionfor DBSCAN. EulideanClusteringNode 854Fetching data from OpenMLwith SciKitLearnNode 863Number of rowsNode 87275% for trainingMin-max normalization10 % for validationWritenormalizationmodelApply network forthe encoderWrite network modelApply networkNode 901One Hot vectorsClass to StringNode 905Apply networkTrain with Loss function=MSEOptimizer=AdamWrite network modelApply network forthe encoderNode 912Node 913Node 914Node 915Node 916Node 917Node 919Node 920Node 921Write network modelApply network forthe encoderNode 924Apply networkNode 926Train with Loss function=MSEOptimizer=AdamNode 933Node 936Node 952Node 955Node 956Node 957Node 958Node 959Node 961Node 962Node 963Node 965Node 966Node 968Colors fordifferentdigitsNode 970Node 971Node 972Node 974Node 975Node 976Node 977Node 978Node 979Node 980Node 981Node 982Node 983Node 984Class duplicatedto intNode 1020Node 1021MinPtsClusteringDistance functionfor DBSCAN. EulideanNode 1026EpsilonNode 1029Distance functionfor DBSCAN. EulideanClusteringNode 1032Node 1035Node 1036MinPtsNode 1038EpsilonClusteringNode 1041Distance functionfor DBSCAN. EuclideanNode 1045Node 1046EpsilonNode 1048MinPtsNode 1050Node 1051Node 1055Write network modelApply network forthe encoderNode 1067Node 1068Node 1069Node 1070Train with Loss function=MSEOptimizer=AdamNode 1075Node 1077Node 1078Node 1079Node 1080Node 1082Node 1083Node 1086Node 1087Node 1088Node 1089Node 1090Node 1091Node 1092Node 1093Node 1094Node 1095Node 1096Node 1097Keras NetworkLearner Keras NetworkWriter Keras NetworkLearner Keras NetworkExecutor Keras NetworkExecutor Double Input Integer Input Numeric Distances DBSCAN Merge Variables Python Source Scatter Plot(Plotly) Integer Input 3D ScatterPlot (Plotly) Keras In-5-Out Partitioning Normalizer Partitioning Model Writer Keras NetworkExecutor Keras NetworkWriter Keras NetworkExecutor MSE CalculationPython & Images Color Appender One to Many Number To String Column Resorter Keras In-5-Out(Indep.Act) Keras NetworkExecutor Keras NetworkLearner Keras NetworkWriter Keras NetworkExecutor Color Appender 3D ScatterPlot (Plotly) Many to One String Manipulation Scorer Scorer (JavaScript) Keras In-5-Out(Indep.Act) String Manipulation Scorer (JavaScript) Color Appender Keras NetworkWriter Keras NetworkExecutor Scorer Keras NetworkExecutor Many to One Keras NetworkLearner Column Filter Extract Color Python Script Color Modelto K.table Table Reader Sorter Sorter Sorter Sorter Math Formula(Multi Column) CSV Writer Partitioning Color Appender Column Filter Visualization Sorter Color Manager Color Appender Python View Python View Column Filter Column Filter Column Filter Python View Column Filter Scatter Plot(Plotly) Python View Keras NetworkReader Keras NetworkReader Keras NetworkReader Keras NetworkReader String Manipulation AssigningCluster/Digit Column Filter Merge Variables Integer Input DBSCAN Numeric Distances Color Appender Double Input AssigningCluster/Digit Sorter Numeric Distances DBSCAN Color Appender Visualization AssigningCluster/Digit Column Filter Merge Variables Integer Input Sorter Double Input DBSCAN Color Appender Numeric Distances Visualization Sorter Column Filter Double Input Merge Variables Integer Input Create CollectionColumn Data Row to Image Keras NetworkReader Keras NetworkWriter Keras NetworkExecutor Keras In-5-Out(Indep.Act) Many to One String Manipulation Scorer Scorer (JavaScript) AssigningCluster/Digit Keras In-5-Out Keras In-7-Out Keras NetworkLearner k-Means AssigningCluster/Digit Color Appender Sorter Column Filter k-Means AssigningCluster/Digit Color Appender Sorter Visualization Column Filter Python Script Math Formula(Multi Column) Table Reader CSV Writer Partitioning Column Rename String Manipulation Table Writer Table Editor Add Empty Rows Table Writer MSE CalculationPython & Images

Nodes

Extensions

Links