A. Forward, processing, optimization, and backward
B. Processing, pre-processing, and post-processing
C. Forward, loss calculation, optimization, and backward
D. Processing, loss calculation, optimization, and post-processing
A. Loss function, optimizer, and parameters
B. Hyperparameters, parameters, and operations
C. Hyperparameters, loss function, and operations
D. Parameters, operations, and loss function
3. After executing the training algorithm on all dataset samples, the training process has completed a training what?
A. Evolution
B. Epoch
C. Step
D. Generation
A. Dataset collections
B. Dataset steps
C. Dataset epochs
D. Dataset instances
5. Which hyperparameter is more likely to increase the computational burden of the training process?
A. Batch size
B. Optimizer
C. Number of epochs
D. Learning rate
6. A training set has 2,500 instances. By defining a batch size equal to 1 and 50, the number of steps executed during the training process is, respectively, which of the following?
A. 500 and 5
B. 2,500 and 1
C. 2,500 and 50
D. 500 and 50
7. The profiling of a training process showed that the most time-consuming operation was aten::mkldnn_convolution. In this case, what is the heavier computing phase of the training process?
A. Backward
B. Forward
C. Loss calculation
D. Optimization
8. A model has two convolutional layers and two fully connected layers. If we add two more convolutional layers to the model, it will increase the number of what?
A. Hyperparameters
B. Training steps
C. Parameters
D. Training samples