ENG335: Design a linear regression model to estimate the target using only FOUR (4) attributes from the dataset: Machine Learning Assignment, SUSS, Singapore

Question 1

Download the SGEMM GPU kernel performance dataset from the below link.

Construct the target parameter by taking the average of the TWO (2) runs with long performance times. Design a linear regression model to estimate the target using only FOUR (4) attributes from the dataset. Discuss your results and estimate the relevant metrics and values.

Question 2

Load the wine dataset from the SK learn package. Perform exploratory data analysis and set up a KNN classifier. Propose an appropriate value for K. Show the relevant performance metrics. Assess whether scaling the data improves the performance.

Question 3

Download the MAGIC gamma telescope data 2004 dataset available in Kaggle

Understand the dataset and perform exploratory data analysis and set up a decision tree for identifying whether the pattern was caused by gamma signal or not. Get the tree depth, performance metrics, and the number of leaves in the tree before and after optimization for the tree depth. You are required to use the ‘entropy’ criterion for the decision tree and also show the optimized decision tree.

Question 4

Use the CIFAR10 dataset from the Keras package. Perform exploratory data analysis. Show a random set of SIX (6) images from each class in the dataset with their corresponding class names. Research on histogram equalization for color images. Keep 20% of the training dataset for model validation.

Prepare the dataset by performing histogram equalization and keeping the pixel values to be between 0 and 1. Adopt LeNet-5 architecture for the CNN retaining the parameters used for the convolutional layers.

For the first TWO (2) dense layers after the fully connected layer, keep the output to 200 and 100, respectively. Use dropout layers if required. Rate the performance of the algorithm and provide necessary plots. Pick a random image of a horse from the test dataset, pass it to the algorithm and compare the algorithm output with the actual class label.

Question 5

Use the IMDB movie review sentiment classification dataset from the Keras package. While loading the dataset, set the num_words parameter to 30000. Perform exploratory data analysis. Show a sample review as text. Get the maximum, minimum, and average length of the review in the dataset. Employ LSTM for the review classification.

Use the nearest multiple of 10 greater than the average review length as the maximum length for truncating or padding the sequence used for training and testing. Set the number of units in LSTM to be 128, batch size to 32, and the number of training epochs greater than 10. Use 20% of the training dataset for validation. Rate the performance of the LSTM classifier and provide necessary plots.