Table 2.

Summary of Projection Operators, Latent Variables, and Functionals in the Proposed Compression Model.

Symbol	Description	Dimensions	Role
	Input feature matrix		Raw sensor input
	Feature compression matrix		Projects input into latent space
	Instance selection matrix		τ-dependent sampling
	Compressed representation		Encoded output
	Latent decoding matrix		Decodes latent features
	Spatial reconstruction matrix		Restores instance structure
	Reconstructed matrix		Final approximation of input
,	Vectorised parameters		Used in gradient optimisation
	Entropy deviation functional	scalar	Penalises encoding complexity
	Directed divergence	scalar	Measures reconstruction loss
	Total loss functional	scalar	Objective to be minimised