Welcome! Please contribute your ideas for what challenges we might aspire to solve, changes in our community that can improve machine learning impact, and examples of machine learning projects that have had tangible impact.
Lacking context for this site? Read the original paper: Machine Learning that Matters (ICML 2012). You can also review the slides.
Beginners Question: scale on optical character recognition
  • I am currently doing Coursera's Machine Learning class. In an exercise on multi-class classification, you tried out both logistic regression and a neural network with a single hidden layer to do recognition of handwritten digits on 20x20 image. You had to "unroll" the image matrix to get a 400 input vector that you would input as features into these algorithms. The neural network performed somewhat better.

    However, I got a sense that the machine learned what a six was, for example, as long as it was drawn about the same size relative to the 20x20 pixel format. Given the same 20x20 pixel space, I could draw a smaller six in the upper-right 1/4th of the space and I dont know if the algorithm would be able to classify it because its input pixel-based features would be distributed in a very different way.

    Do you think the neural network would be able to recognise the smaller six? If not, how may you solve the problem of scale?


To post or add a comment, please sign in or register.


Tip: click the star icon to bookmark (follow) a discussion. You will receive email notifications of subsequent activity.
If search doesn't work, try putting a + in front of your search term.