This trains a model based on Evidential Deep Learning to Quantify Classification Uncertainty on MNIST dataset.
\n": "\u8fd9\u5c06\u8bad\u7ec3\u4e00\u4e2a\u57fa\u4e8e\u8bc1\u636e\u6df1\u5ea6\u5b66\u4e60\u7684\u6a21\u578b\uff0c\u4ee5\u91cf\u5316MNIST\u6570\u636e\u96c6\u4e0a\u7684\u5206\u7c7b\u4e0d\u786e\u5b9a\u6027\u3002
\n", "We use _^_0_^_ configurations.
\n": "\u6211\u4eec\u4f7f\u7528_^_0_^_\u914d\u7f6e\u3002
\n", "\n": "\n", "
'loss_func': 'max_likelihood_loss', 'loss_func': 'cross_entropy_bayes_risk',
\n": "'loss_func'\uff1a'max_imilihood_loss'\uff0c'loss_func'\uff1a'cross_entropy_bayes_risk '\uff0c
\n", "\n": "\u4ea4\u53c9\u71b5\u8d1d\u53f6\u65af\u98ce\u9669
\n", "\n": "KL \u5206\u6b67\u6b63\u5219\u5316
\n", "\n": "\u6700\u5927\u4f3c\u7136\u635f\u5931
\n", "\n": "\u5e73\u65b9\u8bef\u5dee\u8d1d\u53f6\u65af\u98ce\u9669
\n", "Stats module for tracking
\n": "\u7528\u4e8e\u8ddf\u8e2a\u7684\u7edf\u8ba1\u6a21\u5757
\n", "_^_0_^_ max-pooling
\n": "_^_0_^_max-pooling
\n", "Apply dropout
\n": "\u7533\u8bf7\u9000\u5b66
\n", "Apply final layer and return
\n": "\u5e94\u7528\u6700\u540e\u4e00\u5c42\u7136\u540e\u8fd4\u56de
\n", "Apply first convolution and max pooling. The result has shape _^_0_^_
\n": "\u5e94\u7528\u7b2c\u4e00\u4e2a\u5377\u79ef\u548c\u6700\u5927\u6c60\u3002\u7ed3\u679c\u6709\u5f62\u72b6_^_0_^_
\n", "Apply hidden layer
\n": "\u5e94\u7528\u9690\u85cf\u5c42
\n", "Apply second convolution and max pooling. The result has shape _^_0_^_
\n": "\u5e94\u7528\u7b2c\u4e8c\u4e2a\u5377\u79ef\u548c\u6700\u5927\u6c60\u3002\u7ed3\u679c\u6709\u5f62\u72b6_^_0_^_
\n", "Calculate KL Divergence regularization loss
\n": "\u8ba1\u7b97 KL \u80cc\u79bb\u6b63\u5219\u5316\u635f\u5931
\n", "Calculate gradients
\n": "\u8ba1\u7b97\u68af\u5ea6
\n", "Calculate loss
\n": "\u8ba1\u7b97\u635f\u5931
\n", "Clear the gradients
\n": "\u6e05\u9664\u6e10\u53d8
\n", "Create a relative piecewise schedule
\n": "\u521b\u5efa\u76f8\u5bf9\u7684\u5206\u6bb5\u65f6\u95f4\u8868
\n", "Create configurations
\n": "\u521b\u5efa\u914d\u7f6e
\n", "Create experiment
\n": "\u521b\u5efa\u5b9e\u9a8c
\n", "Dropout
\n": "\u8f8d\u5b66
\n", "Dropout for the hidden layer
\n": "\u9690\u85cf\u56fe\u5c42\u7684\u9000\u51fa
\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": "\u6700\u540e\u4e00\u4e2a\u5b8c\u5168\u8fde\u63a5\u7684\u5c42\uff0c\u7528\u4e8e\u8f93\u51fa_^_0_^_\u8bfe\u5802\u8bc1\u636e\u3002RelU \u6216 Softplus \u6fc0\u6d3b\u5728\u6a21\u578b\u4e4b\u5916\u5e94\u7528\u4e8e\u6b64\uff0c\u4ee5\u83b7\u5f97\u975e\u8d1f\u9762\u8bc1\u636e
\n", "First _^_0_^_ convolution layer
\n": "\u7b2c\u4e00\u4e2a_^_0_^_\u5377\u79ef\u5c42
\n", "First fully-connected layer that maps to _^_0_^_ features
\n": "\u7b2c\u4e00\u4e2a\u6620\u5c04\u5230\u8981\u7d20\u7684\u5b8c\u5168\u8fde\u63a5\u7684_^_0_^_\u56fe\u5c42
\n", "Flatten the tensor to shape _^_0_^_
\n": "\u5c06\u5f20\u91cf\u5c55\u5e73\u6210\u5f62\u72b6_^_0_^_
\n", "Get evidences _^_0_^_
\n": "\u83b7\u53d6\u8bc1\u636e_^_0_^_
\n", "Get model outputs
\n": "\u83b7\u53d6\u6a21\u578b\u8f93\u51fa
\n", "KL Divergence loss coefficient _^_0_^_
\n": "KL \u80cc\u79bb\u635f\u5931\u7cfb\u6570_^_0_^_
\n", "KL Divergence regularization coefficient schedule
\n": "KL \u53d1\u6563\u6b63\u5219\u5316\u7cfb\u6570\u65f6\u95f4\u8868
\n", "Load configurations
\n": "\u88c5\u8f7d\u914d\u7f6e
\n", "Module to convert the model output to non-zero evidences
\n": "\u7528\u4e8e\u5c06\u6a21\u578b\u8f93\u51fa\u8f6c\u6362\u4e3a\u975e\u96f6\u8bc1\u636e\u7684\u6a21\u5757
\n", "Move data to the device
\n": "\u5c06\u6570\u636e\u79fb\u52a8\u5230\u8bbe\u5907
\n", "One-hot coded targets
\n": "\u4e00\u70ed\u7f16\u7801\u76ee\u6807
\n", "ReLU activation
\n": "\u6fc0\u6d3b ReLU
\n", "ReLU to calculate evidence
\n": "RelU \u6765\u8ba1\u7b97\u8bc1\u636e
\n", "Save the tracked metrics
\n": "\u4fdd\u5b58\u8ddf\u8e2a\u7684\u6307\u6807
\n", "Second _^_0_^_ convolution layer
\n": "\u7b2c\u4e8c\u4e2a_^_0_^_\u5377\u79ef\u5c42
\n", "Set tracker configurations
\n": "\u8bbe\u7f6e\u8ddf\u8e2a\u5668\u914d\u7f6e
\n", "Softplus to calculate evidence
\n": "Softplus \u7528\u4e8e\u8ba1\u7b97\u8bc1\u636e
\n", "Start the experiment and run the training loop
\n": "\u5f00\u59cb\u5b9e\u9a8c\u5e76\u8fd0\u884c\u8bad\u7ec3\u5faa\u73af
\n", "Take optimizer step
\n": "\u91c7\u53d6\u4f18\u5316\u5668\u6b65\u9aa4
\n", "Total loss
\n": "\u603b\u4e8f\u635f
\n", "Track statistics
\n": "\u8ffd\u8e2a\u7edf\u8ba1\u6570\u636e
\n", "Train the model
\n": "\u8bad\u7ec3\u6a21\u578b
\n", "Training/Evaluation mode
\n": "\u8bad\u7ec3/\u8bc4\u4f30\u6a21\u5f0f
\n", "Update global step (number of samples processed) when in training mode
\n": "\u5728\u8bad\u7ec3\u6a21\u5f0f\u4e0b\u66f4\u65b0\u5168\u5c40\u6b65\u957f\uff08\u5904\u7406\u7684\u6837\u672c\u6570\uff09
\n", "