neuroflow demos

MNIST Classifier

Classify handwritten digits from the MNIST dataset using a neural network. Note, the model starts with pre trained weights for the purpose of this demo.

Training Set

Step:
Loss (CrossEntropy):
Accuracy:
Learning Rate:
Batch Size:
Alpha:
Lambda (L2 Reg):

Loss Curve

Test Set

Step correctly classified of images. Accuracy of . Incorrect predictions summarized below:

Model Architecture