Additionally, changing training batch size could lead to further reduction in training time.
In every episode of the exploration, randomly choosing the command would increase the variety of the training batch.
The loss in Fig. 4 is the average value of the whole training batch in related iteration step.
Because we chose the batch randomly in every step, the different training batch sets between the continues gradient descent steps led the apparent fluctuation in Fig. 4 in both of the environments.
Every time a training batch arrives, a new decision tree is trained and it replaces the tree in the ensemble with the major number of mistakes up to that point.
In our tests we did not optimize the batch size for different representations; however, as data has a smaller memory footprint using log-m embedding, training batch sizes can be increased which could reduce I/O operations between the CPU and GPU.
Hence, samples with high homogeneity are suggested to be placed into the same training batch [ 26].
Application of iPF to the first training batch identified clusters of patients consisting of homogenous disease phenotypes as well as clusters with intermediate disease characteristics.
Having obtained the initial solution, the remaining subset of the data can be trained batch-by-batch without keeping the previously trained set to minimize computational time and complexity.
Physicians are enrolled through a one-week long induction training, in batches of approximately 20.
