This trains a model based on Evidential Deep Learning to Quantify Classification Uncertainty on MNIST dataset.
\n": "\u3053\u308c\u306b\u3088\u308a\u3001\u30a8\u30d3\u30c7\u30f3\u30b7\u30e3\u30eb\u30c7\u30a3\u30fc\u30d7\u30e9\u30fc\u30cb\u30f3\u30b0\u306b\u57fa\u3065\u304f\u30e2\u30c7\u30eb\u3092\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u3057\u3066\u3001MNIST\u30c7\u30fc\u30bf\u30bb\u30c3\u30c8\u306e\u5206\u985e\u306e\u4e0d\u78ba\u5b9f\u6027\u3092\u5b9a\u91cf\u5316\u3057\u307e\u3059\u3002
\n", "We use _^_0_^_ configurations.
\n": "_^_0_^_\u69cb\u6210\u3092\u4f7f\u7528\u3057\u307e\u3059\u3002
\n", "\n": "\n", "
'loss_func': 'max_likelihood_loss', 'loss_func': 'cross_entropy_bayes_risk',
\n": "'loss_func': 'max_likelihood_loss', 'loss_func': 'cross_entropy_bayes_risk',
\n", "\n": "\u30af\u30ed\u30b9\u30a8\u30f3\u30c8\u30ed\u30d4\u30fc\u30d9\u30a4\u30ba\u30ea\u30b9\u30af
\n", "\n": "KL \u30c0\u30a4\u30d0\u30fc\u30b8\u30a7\u30f3\u30b9\u6b63\u5247\u5316
\n", "\n": "\u6700\u5927\u78ba\u7387\u640d\u5931
\n", "\n": "\u4e8c\u4e57\u8aa4\u5dee\u30d9\u30a4\u30ba\u30ea\u30b9\u30af
\n", "Stats module for tracking
\n": "\u8ffd\u8de1\u7528\u7d71\u8a08\u30e2\u30b8\u30e5\u30fc\u30eb
\n", "_^_0_^_ max-pooling
\n": "_^_0_^_\u30de\u30c3\u30af\u30b9\u30d7\u30fc\u30ea\u30f3\u30b0
\n", "Apply dropout
\n": "\u30c9\u30ed\u30c3\u30d7\u30a2\u30a6\u30c8\u3092\u9069\u7528
\n", "Apply final layer and return
\n": "\u6700\u7d42\u30ec\u30a4\u30e4\u30fc\u3092\u9069\u7528\u3057\u3066\u623b\u308b
\n", "Apply first convolution and max pooling. The result has shape _^_0_^_
\n": "\u6700\u521d\u306e\u30b3\u30f3\u30dc\u30ea\u30e5\u30fc\u30b7\u30e7\u30f3\u3068\u30de\u30c3\u30af\u30b9\u30d7\u30fc\u30ea\u30f3\u30b0\u3092\u9069\u7528\u3057\u307e\u3059\u3002\u7d50\u679c\u306b\u306f\u5f62\u304c\u3042\u308a\u307e\u3059 _^_0_^_
\n", "Apply hidden layer
\n": "\u96a0\u3057\u30ec\u30a4\u30e4\u30fc\u3092\u9069\u7528
\n", "Apply second convolution and max pooling. The result has shape _^_0_^_
\n": "2 \u56de\u76ee\u306e\u30b3\u30f3\u30dc\u30ea\u30e5\u30fc\u30b7\u30e7\u30f3\u3068\u6700\u5927\u30d7\u30fc\u30ea\u30f3\u30b0\u3092\u9069\u7528\u3057\u307e\u3059\u3002\u7d50\u679c\u306b\u306f\u5f62\u304c\u3042\u308a\u307e\u3059 _^_0_^_
\n", "Calculate KL Divergence regularization loss
\n": "KL \u30c0\u30a4\u30d0\u30fc\u30b8\u30a7\u30f3\u30b9\u6b63\u5247\u5316\u640d\u5931\u306e\u8a08\u7b97
\n", "Calculate gradients
\n": "\u52fe\u914d\u306e\u8a08\u7b97
\n", "Calculate loss
\n": "\u640d\u5931\u306e\u8a08\u7b97
\n", "Clear the gradients
\n": "\u30b0\u30e9\u30c7\u30fc\u30b7\u30e7\u30f3\u3092\u30af\u30ea\u30a2
\n", "Create a relative piecewise schedule
\n": "\u76f8\u5bfe\u7684\u306a\u533a\u5206\u7684\u30b9\u30b1\u30b8\u30e5\u30fc\u30eb\u306e\u4f5c\u6210
\n", "Create configurations
\n": "\u69cb\u6210\u306e\u4f5c\u6210
\n", "Create experiment
\n": "\u5b9f\u9a13\u3092\u4f5c\u6210
\n", "Dropout
\n": "\u30c9\u30ed\u30c3\u30d7\u30a2\u30a6\u30c8
\n", "Dropout for the hidden layer
\n": "\u96a0\u3057\u30ec\u30a4\u30e4\u30fc\u306e\u30c9\u30ed\u30c3\u30d7\u30a2\u30a6\u30c8
\n", "Final fully connected layer to output evidence for _^_0_^_ classes. The ReLU or Softplus activation is applied to this outside the model to get the non-negative evidence
\n": "_^_0_^_\u30af\u30e9\u30b9\u306e\u30a8\u30d3\u30c7\u30f3\u30b9\u3092\u51fa\u529b\u3059\u308b\u305f\u3081\u306e\u6700\u5f8c\u306e\u5b8c\u5168\u63a5\u7d9a\u30ec\u30a4\u30e4\u30fc\u3002\u3053\u308c\u306bReLU\u307e\u305f\u306fSoftplus\u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3\u3092\u30e2\u30c7\u30eb\u5916\u3067\u9069\u7528\u3059\u308b\u3068\u3001\u975e\u9670\u6027\u30a8\u30d3\u30c7\u30f3\u30b9\u304c\u5f97\u3089\u308c\u307e\u3059
\u3002\n", "First _^_0_^_ convolution layer
\n": "_^_0_^_\u6700\u521d\u306e\u7573\u307f\u8fbc\u307f\u5c64
\n", "First fully-connected layer that maps to _^_0_^_ features
\n": "\u30d5\u30a3\u30fc\u30c1\u30e3\u306b\u30de\u30c3\u30d4\u30f3\u30b0\u3055\u308c\u308b\u6700\u521d\u306e\u5b8c\u5168\u63a5\u7d9a\u30ec\u30a4\u30e4\u30fc _^_0_^_
\n", "Flatten the tensor to shape _^_0_^_
\n": "\u30c6\u30f3\u30bd\u30eb\u3092\u5e73\u3089\u306b\u3057\u3066\u5f62\u3092\u6574\u3048\u308b _^_0_^_
\n", "Get evidences _^_0_^_
\n": "\u8a3c\u62e0\u3092\u53d6\u5f97 _^_0_^_
\n", "Get model outputs
\n": "\u30e2\u30c7\u30eb\u51fa\u529b\u3092\u53d6\u5f97
\n", "KL Divergence loss coefficient _^_0_^_
\n": "KL \u767a\u6563\u640d\u5931\u4fc2\u6570 _^_0_^_
\n", "KL Divergence regularization coefficient schedule
\n": "KL \u30c0\u30a4\u30d0\u30fc\u30b8\u30a7\u30f3\u30b9\u6b63\u5247\u5316\u4fc2\u6570\u30b9\u30b1\u30b8\u30e5\u30fc\u30eb
\n", "Load configurations
\n": "\u69cb\u6210\u3092\u30ed\u30fc\u30c9
\n", "Module to convert the model output to non-zero evidences
\n": "\u30e2\u30c7\u30eb\u51fa\u529b\u3092\u30bc\u30ed\u4ee5\u5916\u306e\u30a8\u30d3\u30c7\u30f3\u30b9\u306b\u5909\u63db\u3059\u308b\u30e2\u30b8\u30e5\u30fc\u30eb
\n", "Move data to the device
\n": "\u30c7\u30fc\u30bf\u3092\u30c7\u30d0\u30a4\u30b9\u306b\u79fb\u52d5
\n", "One-hot coded targets
\n": "\u30ef\u30f3\u30db\u30c3\u30c8\u30b3\u30fc\u30c7\u30a3\u30f3\u30b0\u30bf\u30fc\u30b2\u30c3\u30c8
\n", "ReLU activation
\n": "ReLU \u30a2\u30af\u30c6\u30a3\u30d9\u30fc\u30b7\u30e7\u30f3
\n", "ReLU to calculate evidence
\n": "\u30a8\u30d3\u30c7\u30f3\u30b9\u306e\u8a08\u7b97\u306b\u306f ReLU
\n", "Save the tracked metrics
\n": "\u8ffd\u8de1\u3057\u305f\u30e1\u30c8\u30ea\u30af\u30b9\u3092\u4fdd\u5b58\u3059\u308b
\n", "Second _^_0_^_ convolution layer
\n": "2 _^_0_^_ \u756a\u76ee\u306e\u7573\u307f\u8fbc\u307f\u5c64
\n", "Set tracker configurations
\n": "\u30c8\u30e9\u30c3\u30ab\u30fc\u69cb\u6210\u3092\u8a2d\u5b9a
\n", "Softplus to calculate evidence
\n": "\u8a3c\u62e0\u8a08\u7b97\u7528\u30bd\u30d5\u30c8\u30d7\u30e9\u30b9
\n", "Start the experiment and run the training loop
\n": "\u5b9f\u9a13\u3092\u958b\u59cb\u3057\u3001\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30eb\u30fc\u30d7\u3092\u5b9f\u884c\u3057\u307e\u3059
\n", "Take optimizer step
\n": "\u6700\u9069\u5316\u306e\u4e00\u6b69\u3092\u8e0f\u307f\u51fa\u3059
\n", "Total loss
\n": "\u7dcf\u640d\u5931
\n", "Track statistics
\n": "\u30c8\u30e9\u30c3\u30af\u7d71\u8a08
\n", "Train the model
\n": "\u30e2\u30c7\u30eb\u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0
\n", "Training/Evaluation mode
\n": "\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0/\u8a55\u4fa1\u30e2\u30fc\u30c9
\n", "Update global step (number of samples processed) when in training mode
\n": "\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u30e2\u30fc\u30c9\u6642\u306b\u30b0\u30ed\u30fc\u30d0\u30eb\u30b9\u30c6\u30c3\u30d7 (\u51e6\u7406\u3055\u308c\u305f\u30b5\u30f3\u30d7\u30eb\u6570) \u3092\u66f4\u65b0
\n", "