Before the workshop

• Watch lectures 5 to 8 and read the corresponding chapters in the course literature
• Prepare questions (if you have any) on the contents in lectures 5 to 8

Software

It is recommended that you use a suitable Docker image for the workshop assignments, see Docker page.

Aim of the workshop

The aim of this workshop is to discuss the contents in lectures 5 to 8, and do practical assignments on what you have learned in the lectures.

Assignments

A1: Spiral dataset in the Weka tool

• Classify the Spiral dataset in Weka using the Linear (functions/Liblinear), Neural Network (MultiLayerPerceptron) and Support Vector Machines (functions/LibSVM) algorithms
• The Spiral dataset can be downloaded from the Datasets page
• Why does the Linear classifier has much lower accuracy than Neural Network and Support Vector Machine?
• Note that Weka automatically tries to determine the size of the hidden layer. It often happens that it uses too few hidden units to be able to accurately learn the concept. Try to change the hiddenLayers field from a to 72.

A2: Spiral dataset in Scikit

• Classify the Spiral dataset in Scikit using a Neural Network algorithm
• You need to write code for loading csv dataset files (use the Pandas library)

A3: Diabetes dataset in Scikit

• Classify the Diabetes datasets in Scikit using the Neural Network and Xgboost algorithms
• You need to write code for loading csv dataset files (use the Pandas library)

A4: Diabetes dataset in Keras

• Write code for loading, training and evaluating the Diabetes dataset using a neural network classifier in Keras
• See instructions at the Keras page

A5: MNIST dataset in Keras

• Write code for loading, training and evaluating the MNIST dataset using a Linear, Neural Network and a ConvNet classifier in Keras
• See instructions at the Keras page

A6: Banknote dataset in Web ML Experimenter

• Download the Banknote dataset from the Datasets page
• Upload the dataset in the Web ML Experimenter
• Try classifying the Banknite dataset using different classifiers. Test with different hyperparameter settings.
• Which classifier had the highest accuracy?

A7: Diabetes dataset in R

• Classify the diabetes dataset in R using Neural Networks, SVM and RandomForest
• Split the dataset into 80% training and 20% testing, and evaluate accuracy on the test dataset
• Which classifier had the highest accuracy?

A8: Pre-trained models in Keras

• Use the pre-trained models VGG16 and VGG19 in Keras to classify images
• See instructions at the Keras page
• Test on other images than the examples. Are they classified correctly?