Narayanan, S., Aspinwall, C. et al
It’s 2021. You’ve been locked down with randos from Craigslist for almost a year. The dishes pile has finally taken over your entire kitchen. Despite months of their denials, you are still pretty sure that at least one roommate hasn’t done a single dish since March 2020. Here’s how to prove it.
Random Forest Models: Gather 14 of your closet roommates for a classic whodunit approach to a slovely sink. Let the wild accusations fly and boom! Whomever the most people randomly accuse is the culprit. Pick up a sponge Debbie!
Naive Bayes Classifier: Maybe you naively assumed your roommate wasn’t raised in a barn, but hopefully this classifier can tell you that, given the absurd number of dishes covered in “hot-Cheeto flavored cheesy pasta,” they’re all probably hers.
Gaussian Mixture Models: Implementation is easy and vengeance swift with this handy technique! Pick a plate you know is your roommate’s and work your way out. On the off chance you left plates in the sink group dishes next to yours in your own ‘cluster.’ The closer to the center of a pile an item is the greater chance it belongs to that cluster. WOW look at that, the enormous frying pan your roommate left blocking the drain is the only centroid and these are all theirs!
Principal Component Analysis: Oh come on! These plates are covered in tuna and gelatinous, two-week old, peanut-butter oatmeal — something only your roommate who eats “nutty fish oats” would enjoy. To prove it, gather data on what flavor profiles your roommate typically likes and then characterize the dish residue similarly. If you aren’t going to rinse your plate it can and will be used against you.
Partial Least Squares Regression (PLS): The most simplistic method, try it out by regressing to third grade and kindly asking your roommate to do their fair share “pretty pls.” Note the limitations: rarely effective in more complicated situations and overlooks the root cause of your roommate’s selfishness.
Fuzzy Clustering: Are you looking down at the growing carpet of mold in your sink and wondering where one dish begins and the other ends? You may be interested in using a Fuzzy Clustering analysis tool to prove your roommate’s “soaking dishes” are not your problem.
Density Based Clustering: Is your slob abroad? With dirty dishes everywhere but the sink? Then this is the technique for you. Find one of your roommates dirty dishes by the TV, take a step, find the fork they left on the couch, then the glass on the table, and keep going until you’re out of hands and worked into a frenzy. If it’s within arm reach you can blame it on them!
Mean Shift Clustering: Get a few housemates on your side to provoke the village layabout into action. Hit them with a good-roommate/bad-roommate routine and watch how the dramatic tone shifts inspire them to rinse a few plates. Sure you shouldn’t have to stoop to being the mean one in the house, but are you here to make friends or to see the bottom of the sink? That’s what I thought.
Blind Signal Separation: You’ve had it up to here and by here you mean the foot of water in the clogged sink. Find out what your roommate did now by using BSS to fish around in the murky waters and pull out the clog of diced tomatoes, orzo, and… hair? Careful! Even algorithms on the cutting edge of signal separation can’t see the rusty knife hiding in the deep.
‘Kay Nearest Neighbor! Use proximity to prove your roommate’s guilt. Yeah right that isn’t your dish, you’re standing right next to it!