Congratulations on finding another Hummus Challenge. This time it’s a faceoff between you and a machine learning model. Can you correctly classify more images than the machine?
Machine learning models are designed to be used with large sets of data. Similarly, humans do better when they gather more information and see more examples. So how well can each perform when they are given very little data?
To put this to the test, we’ll give the machine learning model 60 pieces of training data, but we’ll give you only 6. Humans are great at recognizing patterns, so to make this a fair fight we’ve given the data similar names and you’ll only get 3 seconds per question. Can you beat the machine even though it has 10 times the data to learn from?
As you can see below, we’ve randomly generated images that follow certain patterns. All images with the same pattern belong to the same class. In our data the names of the classes are Hummly, Hummel, and Hummster. In machine learning we consider the images the inputs, and the names, or labels, the outputs.
Our training data provides both the inputs and outputs, which is called supervised learning. Also, since all images belong to one (and only one) class, this is a multi-class classification problem.
Learn from the data, take the quiz, and beat the bot to get your next clue!