A study is presented of a trigger for b and c quark production in high-energy collider experiments, based on a feed-forward artificial neural network. The recently approved BTeV experiment at the Fermilab collider is used as a case study. The input to the network is provided by the track processors which are proposed for the default trigger, using information from a pixel vertex detector. The approach, which attempts to exploit global features and multi-track correlations in heavy-quark events, is compared to the canonical detached-track triggering scheme of the experiment. The feasibility of the method as an alternative first-level online trigger, using an independent processor for each of its units, is discussed. Other network architectures, are also briefly
mentioned, in particular self-organizing feature maps. Finally, other possibilities are considered, such as implementation at a higher-level trigger, incorporating information from additional detectors and the vertex processor.