Scaling down Deep Learning

Scaling down Deep Learning

Scaling down Deep Learning

“Constructing the MNIST-1D dataset. As with the original MNIST dataset, the task is to learn to classify the digits 0-9. Unlike the MNIST dataset, which consists of 28×28 images, each of these examples is a one-dimensional sequence of points. To generate an example, we begin with 10 digit templates and then randomly pad, translate, add noise, and transform them as shown above…”

Source: greydanus.github.io/2020/12/01/scaling-down/

Code: https://github.com/greydanus/mnist1d