It is also always worth seeing whether the process can be restructured to be more efficient. A review of a process may reveal that expensive operations are being performed repeatedly and unnecessarily. In this situation there are some operators that can help.

The Store and Retrieve operators allow objects to be stored in the repository. Objects such as example sets, models, and weights can be stored and retrieved in this way. Typically, a process will perform an expensive operation once and store it in the repository. Subsequent operators can retrieve the object whenever it is needed. These operators can also be used to implement a checkpoint regime within a big process. This is relatively complex to set up, but it may be worth having the process determine where it reached in a long processing step so that it starts up where it left off.

The Recall and Remember operators are similar except they do not persist data to the repository. These can be used in a way similar to the Store and Retrieve operators except that the type of the object must be specified, and when using the Recall operator, the option to remove it from the store can be specified. These operators will consume memory but are likely to be relatively quicker than those that interact with the repository.