SEIS 764 Artificial Intelligence 1
Graduate Program in Software
SEIS 764: Artificial Intelligence
Assignment #4 (100 points)
Due Date: 5:45pm, March 3rd
The dataset on the Blackboard (CellDNA.csv) contains various measurements (i.e. size, center, etc) from thousands of bacterium under microscope. The non-zero values in the last column are the target responses that indicate the bacterium (rows) that are interesting enough for further study. The 0s in the last column indicate the bacterium (rows) are NOT interesting candidates for further study. Convert this target dependent variable (last column) to binary values of either 0s or 1s for your two-class classification.
Write programs using either Python, MatLab, to perform the following tasks:
1. Train a neural network using **ALL** the records in this dataset to perform a two-class classification analysis.
2. Extract the hidden representations (Z-Code) of all the input records from the last layer (the layer right before the Softmax layer) of the trained neural network.
3. If the dimension of the Z-Code is greater than two dimensions, perform a t-sne compression to reduce the dimension to 2.
4. Use t-sne compression results to reduce the dimensionality of the raw data from 13 to 2.
5. Create two group plots: the one on the left contains points you obtained in step 4, and the one on the right contains points you obtained in step 3. Please make sure that you use two colors to represent the classes of the data points in each plot. Please see our slides for the example plots.
6. Put your plots in a WORD document.
7. Save your code as “a4.py”, “a4.ipynb”, or “a4.m”.
Submission Guideline:
1. Please include the WORD document your created in answering the above questions. Please include your name on the top of your WORD document.
2. Please print your program (matlab or python) as PDF and include the PDF in your submission.
3. Please also include your program in the formats like .mat/.mlx/.py/.inpyb in your submission.
4. Prepare EVERYTHING mentioned in the submission guideline and submit them on Canvas no later than the due date. Please do NOT zip your files.
© Dr. Chih Lai University of St. Thomas