Question 1
Download the SGEMM GPU kernel performance dataset from the below link. https://archive.ics.uci.edu/ml/datasets/SGEMM+GPU+...
Understand the dataset by performing exploratory analysis. Prepare the target parameter by taking the average of the THREE (3) runs with long performance times. Design a linear regression model to estimate the target using only THREE (3) attributes from the dataset. Discuss your results, relevant performance metrics and the impact of normalizing the dataset.
Question 2
Load the wine dataset from sklearn package. Perform exploratory data analysis and Design a simple TWO (2) layer neural network for the classification. Compare the performance with the Naïve Bayes algorithm. Train the neural network such that it has better or same performance as that of the Naïve Bayes algorithm.
Question 3
Download the MAGIC gamma telescope data 2004 dataset available in Kaggle (https://www.kaggle.com/abhinand05/magic-gamma-telescope-dataset). Prepare the dataset and perform exploratory data analysis. Set-up a random forest algorithm for identifying whether the pattern was caused by gamma signal or not. Propose optimal values for the depth and number of trees in the random forest. Assess and compare the performance of optimized random forest with the Naïve Bayes algorithm. Discuss the performance metrics and the computational complexity.
Question 4
Use the Fashion MNIST dataset from the keras package. Perform exploratory data analysis. Show a random set of FIVE (5) images from each class in the dataset with their corresponding class names. Prepare the dataset by normalizing the pixel values to be between 0 and 1. Design a CNN with TWO (2) convolutional layers and FOUR (4) dense layers (including the final output layer). Employ ‘ReLU’ activation and ‘MaxPooling’. Keep 15% of the train dataset for validation. Rate the performance of the algorithm and provide necessary plots. Pick a random image from the test dataset, pass it to the algorithm and compare the algorithm output with the actual class label.
Question 5
Questions
- Use data until 31 Dec 2020 for training and the remaining data for testing. You must select the stock such that the data is available from 1 Jan 2020 to 3 Jan
- Use previous 30 days of stock information to predict the next day stock price. Use the data in ‘High’ column to predict the price, i.e., the next day high price of the stock. Design a LSTM network to do the predictions. You are required to use LSTM with a cell state of at least 60 dimension and do at least 50 epochs of training. Rate the performance of the LSTM classifier and provide necessary plots.