Mini-Batch Size Selection

Mini-batch size selection is the process of choosing the number of training examples to use in each iteration of stochastic gradient descent. This choice involves a trade-off between the speed of computation and the quality of the gradient estimate.

A smaller batch size provides a more stochastic gradient, which can help the model explore the parameter space and escape local minima, but it may increase the time required for convergence. A larger batch size provides a more accurate estimate of the gradient, leading to more stable updates, but it can be computationally expensive and may lead to sharp minima that generalize poorly.

In crypto-derivative trading, where data is abundant and time is of the essence, finding the right balance is key. The optimal batch size depends on the hardware architecture, the size of the model, and the nature of the financial data.

It is a critical hyperparameter that influences the final performance and robustness of the trading model.