Combining multiple probabilities from a classifier. Propagating probabilities

Question

Combining multiple probabilities from a classifier. Propagating probabilities

AstroAllie

2022年4月17日 03:07

Let's say I have trained a classifier that classifies images of animals into 10 different classes. And let's say that I have 20 different images of a particular animal and because I know the photographer, I know with certainty that all 20 images are of the same animal. So I use my classifier to make a prediction on what animal it is and get 20 predictions one for each image. The model predicts all the images to be a dog with varying probabilities. image 1: 80% dog image 2: 90% dog image 3: 75% dog and so on.

What is the probability that the animal in question is a dog? Let's say they predict cat with smaller probabilities, 5%, 2%, 4% ... What is the probability it is a cat?

I've tried a few different approaches, applying Bayes Theorem but I keep getting numbers that add up to be more than one. Could it really be just the average?

Topic bayesian probability classifier statistics

Category Data Science

yakhyo · Accepted Answer · 2021年2月19日 02:15

If you feed the model with 20 images for testing then your output should look like array of [20x10]. For each row represents the probabilities of all classes(in your case 10). Let's see the example below, and 1-index is dog with 0.9 probability, so your model classified it correctly. If your model classifies 19 times dog-correct and 1 times cat-wrong then your model's test accuracy will be calculated as correct_preds / (correct_preds+wrong_preds)

[[0.0, 0.9, 0.0, 0.0, 0.0, 0.1, 0.0, 0.0, 0.0, 0.0],
 [0.0, 0.8, 0.0, 0.0, 0.0, 0.1, 0.0, 0.1, 0.0, 0.0],
 ..,
 [0.0, 0.75, 0.0, 0.0, 0.0, 0.1, 0.0, 0.1, 0.05, 0.0]]

Combining multiple probabilities from a classifier. Propagating probabilities

About