The paper maps the two frameworks as follows:
X โ source material;T โ kernel;Y โ reconstructive target;I(T;X) โ retained source information;I(T;Y) โ reconstructive capacity;The correspondence is explicitly structural, not yet mathematical.
The central thesis is regime underdetermination. Information Bottleneck measures compression and target relevance, but not:
The Three Compressions framework adds those dimensions through Lossy, Predatory, and Witness regimes.
The paper distinguishes four statuses:
A research program would require formal variables, encoders/decoders, controlled kernel tasks, and comparison of back-projection performance with estimated information frontiers.