Hostile Neural Networks

Data Models are the core unit of the mod. A Data Model is an item that stores information about one entity and is used to generate drops from that entity. To obtain one, a player first crafts a Data Model Framework and uses it on the target entity, then builds the model. A model is trained once it is built, either through a Deep Learner or through a Simulation chamber. A Deep Learner stores several models at once, collecting data from kills while they train.

Models start in a Faulty state and must be trained up in a Basic Deep Learner before the simulation chamber can run simulations on them. Simulations always produce a generalized prediction, and its accuracy depends on the model's tier, rising up to 99.5% for a Self Aware model. Running simulations needs energy and a catalyst item, usually a Prediction Matrix. The resulting Mob Predictions are converted into mob drops by a Loot Fabricator, which also uses power. Drops come from a pre-defined list of items for the selected entity type, and a fabricator remembers the chosen type and can take predictions from several types.