Hostile Neural Networks is a mob based on Deep Mob Learning for 1.16.5 and higher. It allows you to train data models and run simulations, producing mob predictions, which can be used to generate mob loot.
The fundamental unit of this mod is the Data Model - an item which holds information about an entity that you can utilize to generate their drops. To get a Data Model, you must first craft a Data Model Framework, and use it on the target entity to build the model. Once you have your model, you will have to train it.
Training a Data Model can be done in one of two ways, with a Deep Learner or with a Simulation Chamber. The deep learner allows you to store a few data models within, and kills you receive will begin to train the model. All models start out in the Faulty state, and must be trained up to Basic in the deep learner.
Once your model is Basic or higher, you can use the simulation chamber to run simulations of your model. Simulations will always produce a generalized prediction, and may also produce a Mob Prediction, based on the accuracy of the model. The higher tier of the model, the higher the accuracy, up to 99.5% at Self Aware. Running simulations requires energy (based on the model type) and a catalyst item, which is usually a Prediction Matrix.
Mob Predictions will need to be processed in a Loot Fabricator, a device which uses power to convert predictions to mob drops. The drops which are available are based on a list of pre-defined items, and a target can be selected for each type of model. The loot fabricator will remember your choices for each type, so you can feed multiple types of predictions into a single fabricator.
Custom Data Models can be added via a Datapack. Each Data Model is represented as a single file under the data_models subfolder. An example is the Witch Data Model file located in the base mod.
Published datapacks for certain mods may be listed here in the future as they are created.
