A Comparative Study And Integration With Generative AI Tools

Question 1: Variations of the Two-Spiral Task

Perform an experimental study on the following variations of the two-spiral task:

1. ANN: Start with the original dataset of Lang and Wit- brock (1988) with 194 training points (see Figure on the

right). How fast and how well can you solve this task using a feed-forward NN? The (x, y)-coordinates of the points in the dataset will be supplied in Canvas

1. ANN: Generate a variation of the 2-spiral Then solve the associated classification task using ANNs and discuss your approach and solution in comparison to a). You may consider, e.g., to generate three spirals or four spirals, use expanding and rotating spirals (Chalup and Wiklendt, 2007), or use code such as available

1. ANN vs. SVM: Run experiments that allow to compare how well ANNs and SVMs can solve the two classification tasks above, and discuss the

For each subquestion try out different architectures, hyperparameters, and methods. Compare and discuss their performance (e.g., speed, generalisation, convergence). The performance of the solutions can be evaluated by visual inspection of a generalisation test applied to all pixels of a section of the (x, y)-plane (that for the 2-spiral data should result in two intertwined spiral shaped regions). You may also think about alternative performance measures, e.g., considering false positives and false negatives, confusion matrices, learning curves etc.

Question 2: BUPA Liver Disorders Dataset

The BUPA Liver Disorders Data Set was collected by BUPA Medical Research Ltd. in the 1980s for research on liver health. It has since been donated to the UCI Machine Learning Repository and is available from:

The dataset contains 345 samples of five explanatory variables and one target variable. The ex- planatory variables represent various blood markers and come as integers, but can be regarded as continuous features. The target variable is the self-reported number of alcoholic drinks con- sumed per day. There is also a seventh variable distributed with the dataset, which represents the train/test split chosen in the original research. This variable has often been mistakenly interpreted as the target class (McDermott and Forsyth, 2016) and can be disregarded here.

1. Prepare the data for a binary classification task by dichotomising the original target variable into two classes using a cut-off value (e.g., 3 drinks/day, such that the two classes are relatively balanced). You will also normalise the explanatory variables, which may help with model convergence later
2. Train different classifiers and report the maximum accuracy that you can achieve in a 10-fold metric cross-validation experiment. For solving this you can train an SVM and/or a Neural Network, or some combination. Your experiments should also investigate what impact normalisation or some alternative forms of preprocessing of the explanatory variables has on the

Question 3: Using LLMs or generative AI tools

In this part of the assignment we investigate how questions 1 and 2 above can be addressed when using the help of additional tools like ChatGPT or similar.

1. State some research questions or hypotheses and conduct some experiments using one or more of the available tools like
2. Discuss outcomes and

This part of the assignment is more open-ended and it would probably use about two pages of your report.

Are you struggling to keep up with the demands of your academic journey? Don't worry, we've got your back! Exam Question Bank is your trusted partner in achieving academic excellence for all kind of technical and non-technical subjects.

Our comprehensive range of academic services is designed to cater to students at every level. Whether you're a high school student, a college undergraduate, or pursuing advanced studies, we have the expertise and resources to support you.

To connect with expert and ask your query click here Exam Question Bank