A08: Extra credit: Manatee Shmanatee

You may choose to complete this assignment for extra credit. If you get a good grade on this assignment, it will replace the grade for your lowest assignment (e.g., one you didn’t turn in).


  1. Find 50 “absent” and 10 “present” files from /bigdata/data/manatees/splits and place them in /bigdata/data/manatees/present and /bigdata/data/manatees/absent on delenn. Nathan Wolek’s notes in the _notes folder will help find the times when manatees are present. I can see who created which files based on the file username (seen with ls -l); make sure you find present/absent files that no one else has already found. Copy equal splits of your files into the train/ and test/ folders also. Make sure you keep channels together (don’t put ch6.split123 in test and ch5.split123 in train). Consider also using Sonic Visualiser on the MIXPRE files in the folders named after dates.
  2. Build a convolutional neural net with binary outputs to predict present/absent. Include an appropriate evaluation on the test data. Try to get the highest possible accuracy.

Your submission will be graded on completeness and insight, i.e., finding a good model.

CSCI 431 material by Joshua Eckroth is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License. Source code for this website available at GitHub.