#MachineLearning – Breakthrough with sklearn. Train with serial dependent sample batches. 98.5% MNIST 1000.

#MachineLearning – Made a Breakthrough with neural_network.MLPClassifier() You predict sample classes then train on another set with the same series of classes so it becomes dependent on the initial batch. Got very good results. 98.5% MNIST 1000 test with the hidden layersize which I iterated to get, (208, 284, 189). So the method can be used to train larger models. No convolution just fully connected layers. Wow