I am curious about the "under the hood" mechanics of the model to learn what parts of the photo is making the algorithm select "like" rather than "nope." It seems like the algorithm picks up on photo qualities like how centered or blurry is the photo subject is, but I'm curious of how much algorithm accounts for personality attributes when assigning a confidence percentage (i.e. the photo's background is outdoors or the photo subject is smiling). I envisioned the formula being something like 0.5(outdoors_dummy_variable) + 0.46(smiling_dummy_variable) = 0.96 for a confidence percentage of 96%. Do you have any insight on how I could find this in the Tinder Deep Learning model that you created?
I hope you're having a good December and thanks in advance!
Hello!
I am curious about the "under the hood" mechanics of the model to learn what parts of the photo is making the algorithm select "like" rather than "nope." It seems like the algorithm picks up on photo qualities like how centered or blurry is the photo subject is, but I'm curious of how much algorithm accounts for personality attributes when assigning a confidence percentage (i.e. the photo's background is outdoors or the photo subject is smiling). I envisioned the formula being something like 0.5(outdoors_dummy_variable) + 0.46(smiling_dummy_variable) = 0.96 for a confidence percentage of 96%. Do you have any insight on how I could find this in the Tinder Deep Learning model that you created?
I hope you're having a good December and thanks in advance!