Figure 2.

The architecture of our proposed framework, the neural information squeezer plus (NIS+), building upon our previous model, NIS [10,44]. It accepts a variety of time series as inputs and enables us to derive the degree of CE, the learned macro-dynamics, captured emergent patterns and the strategy for coarse graining. Within the framework, the boxes symbolize functions or neural networks, while the arrow pointing to a cross signifies the operation of information discarding. For further details regarding these mathematical symbols, see Section S2.2.