... HNFA

One could also think in terms of e.g. a mixture model.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... relations.

In mixture of experts (MoE), it is enough when only one of the experts explains the data even if all the other disagree. The ignored experts will not pass information on. This explains why the author did not consider MoE as a combination rule.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.