michael_nielsen

New notes on using ML to generalize from small amounts of training data: attacking MNIST with just 10 training examples for each digit. 93.81% accuracy: http://cognitivemedium.com/rmnist_anneal_ensemble …

If you need a benchmark, this type of stuff is relatively easy to do with Gaussian mixture models. Use every example as a mean & modify the covariances to increase the accuracy.

I ran a half-dozen different baselines in the post linked at the top. None got much above 75%. I haven't done GMM, but I'd be surprised if they were much better.

If I remember correctly I did get around 98.4% accuracy on the test set (without any preprocessing) with GMMs (20-25 Gaussians) using all the training data. SVM with rbf kernel was something like 98.6%. So it wasn't overfitting much.

Did you read the post?

Yes thats why I said use each digit as a mean and choose a covariance. No training ver easy to form baseline classifier. What I meant was, as far as I remember, GMMs didn't perform that bad on MNIST.