The natural place to start a modularization is by thinking in terms of a top down design, that is, to address a task at successively lower levels of detail.
The first method, top-down domain adaption multi-task (TDMT SVRR, successively trains more specific models along a supplied taxonomy.
The algorithm, then, merges the closer tasks into successively larger clusters, until only one cluster is left.
This task consists of successively displacing a reward inside an opaque container (or a closed hand) across multiple opaque containers and dropping the reward inside one of them.
The solution of the recognition task is progressive, processing successively different objects, and using different pieces of knowledge about the object and about relationships between objects.
Half of the subjects performed successively the Interference Stroop task, the two control tasks and the Reverse Stroop task, and half performed the Reverse Stroop task, the control tasks and finally the Interference Stroop task.
Subsequently, rats were trained in a task that required discrimination of successively presented signals and "blank" trials with no signal presentation.
Participants successively performed three tasks, while EEG was recorded.
In our version of the task, mice were conditioned to run successively down two runaways differing in colors, one color runaway being always baited with food (Go trail) and the other never baited (No-Go trial).
Over the course of the task, the amounts of immediate rewards are successively modified, as is the duration of delay.
