esc

The rabbit hole.

Bialgebras, Categories and Homology.

10On arXiv
5Peer-Reviewed
14Talks
23Repositories

Research

  1. Programming Languages H∙(X;ΩX)

    Homology of Programming Languages

    A programming language is a syntax with rules that say how its programs run, and the programs obey laws, such as the commutativity of a sum. Homology measures these laws. For polynomial behaviour, differentiating the rules gives every model a linearisation, and a complex drawn on the model itself computes the homology. Its first group embeds into the relations of a presentation modulo their consequences.

  2. Behavioural Distances d(c[s],c[t])≤mc(d(s,t))

    GSOS Laws, Measured

    Programs with the same behaviour should be interchangeable in every context. When behaviour is compared by a distance, the distance of two composites should be bounded by a modulus of the distances of their parts. Over quasi-pseudometric spaces, a GSOS law gives every model a least distance for which its coalgebra is nonexpansive, and this distance obeys the bound without any final coalgebra. For programs that flip coins the rules compute the moduli, which are optimal for the standard process algebra.

  3. Operational Semantics λ:ΣB⇒BΣ

    Turi–Plotkin Homology

    Rules in the GSOS format are natural transformations, and their models are bialgebras. The free bialgebras give a comonad, and its bar resolution turns every model into a chain complex. For linear rules one Smith normal form computes the homology, and degree one finds laws that hold in a model but not in the free one, such as ρ(ρ(x))=x.

  4. Quantitative Algebra TUA

    Magnitude Meets Persistence

    A quantitative equational theory U gives every metric space A a free algebra TUA. Filtering its length nerve by length yields persistent homology, and the associated graded is magnitude homology. A long exact sequence ties the two together, and the barcode in degree n moves by at most (n+1)δ under a perturbation of size δ.

  5. Categories and Learning LanJF

    Kan Extensions for Transfer

    A change of task is a functor J, and the invariants of a learned representation form a functor F. The left Kan extension LanJF is what the source forces on the target, and its distance to the observed invariant measures the transfer. For persistent homology it is a bottleneck distance, and a functorial version is stable under interleavings.

  6. Topological Groupoids H∙(G;A)

    Homology of Ample Groupoids

    Ample groupoids carry a homology built from compactly supported functions on their nerve. With discrete coefficients a universal coefficient theorem holds, and on the Cantor set this homology differs from that of the classifying space already in degree zero. Laminar presentations cut finite chain models out of its chain complex, with ranks bounded by the size of the cover.

  • bialgebras
  • comonads
  • distances
  • free algebras
  • adjunctions
  • homology
  • derived colimits