Setting up the DQN-CRLMM model
This section describes how to set up the previous chapter's model for this project and add a few functions.
A DQN-CRLMM model contains a convolutional neural network (CNN) and a Markov Decision Process (MDP) linked together by an optimizer.
A conceptual representation learning meta-model contains:
- A CNN
- An optimizer linking it to an MDP
- An MDP function
This system will now be referred to as a CRLMM.
Training the CRLMM
In previous chapters, the CRLMM program CNN_STRATEGY_MODEL.py was trained to identify Γ (gamma concept) in outputs on the conveyor belt of a food processing factory. The end of the previous chapter brought Γ up to a higher abstraction level.
As long as a member of γ (gamma) of the Γ dataset is in an undetermined state, its generalization encompasses ambivalent but similar concepts. Up to this chapter, these are the concepts Γ has learned (conceptual representation learning).
Γ = { a gap, no gap, a load, no load, not enough load, enough load, too much load...
