Search papers, labs, and topics across Lattice.
The paper introduces Geometric Manifold Analysis (GeMA), a novel framework for benchmarking complex systems by representing the production set as the boundary of a low-dimensional manifold learned via a productivity-manifold variational autoencoder (ProMan-VAE). GeMA addresses limitations of classical frontier methods like DEA and SFA by directly modeling heterogeneity, scale effects, and non-convex frontiers in a latent space. Validated on synthetic and real-world datasets including rail systems, national economies, and wind farms, GeMA demonstrates comparable performance to existing methods under classical assumptions while offering enhanced insights in complex scenarios.
Benchmarking complex systems just got a geometric upgrade: GeMA learns latent manifold frontiers to reveal hidden inefficiencies and technological structures, outperforming traditional methods when heterogeneity and scale bias muddy the waters.
Benchmarking the performance of complex systems such as rail networks, renewable generation assets and national economies is central to transport planning, regulation and macroeconomic analysis. Classical frontier methods, notably Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA), estimate an efficient frontier in the observed input-output space and define efficiency as distance to this frontier, but rely on restrictive assumptions on the production set and only indirectly address heterogeneity and scale effects. We propose Geometric Manifold Analysis (GeMA), a latent manifold frontier framework implemented via a productivity-manifold variational autoencoder (ProMan-VAE). Instead of specifying a frontier function in the observed space, GeMA represents the production set as the boundary of a low-dimensional manifold embedded in the joint input-output space. A split-head encoder learns latent variables that capture technological structure and operational inefficiency. Efficiency is evaluated with respect to the learned manifold, endogenous peer groups arise as clusters in latent technology space, a quotient construction supports scale-invariant benchmarking, and a local certification radius, derived from the decoder Jacobian and a Lipschitz bound, quantifies the geometric robustness of efficiency scores. We validate GeMA on synthetic data with non-convex frontiers, heterogeneous technologies and scale bias, and on four real-world case studies: global urban rail systems (COMET), British rail operators (ORR), national economies (Penn World Table) and a high-frequency wind-farm dataset. Across these domains GeMA behaves comparably to established methods when classical assumptions hold, and provides additional insight in settings with pronounced heterogeneity, non-convexity or size-related bias.
