Parkinson’s Telemonitoring

This assignment contains 3 problems
Problem 1 – MNIST Data Set (10 points)
Use the MNIST dataset to classify the type of apparel using an artificial nueral network.
You may use the library of your choosing
We went over this is class, so this is an easy one to get you started.
Grading criteria: Your results on the test data should exceed 0.80 accuracy and you have some explaination about the model.
Problem 2 – Poker Hand Classification (30 points)
Use the Pokerhand dataset at
Pay attention to class distribution.Do you need to do anything to balance the data?
How does it affect ANNs if the training data is ordered?
I mentioned early stopping in class but didn’t go into detail. Implement early stopping in your modelThis article provides an example of early stopping with Keras and Tensorflow
Provide a visualization of training and cross validation loss at each epoch
Provide a confusion matrix and F1 score for the test data
The output of the model will be probabilities of each class. I like
to use np.argmax to get the most probable class, but you may use other
Problem 3 – Parkinson’s Telemonitoring (30 points)
In class, we focused on artificial nueral networks for classification purposes
Use the data set at goal is to predict the ‘motor_UPDRS’
Remove the ‘total_UPDRS’ column. The goal of the data set author is
to predict both and ‘total_UPDRS’ but I don’t want to have two predicted
variables. I also don’t want to use it as a predictor because ‘total_UPDRS’ will be too colinear with ‘motor_UPDRS’
In this example, we want to use a regression ANN
Use a 80/20 train test split for your model
I acknowledge that you can solve this without an ANN, but please use a regression ANNMake sure you look at fields that are numbers but are not truly ordinal. How would you handle them?
This article provides an overview of regression ANN using Tensorflow.

Place this order or similar order and get exceptional paper written by our team of experts at an affordable price

Leave a Reply