Implementation reference
The same mathematical answer can be computed in several ways, with different intermediate data, costs and opportunities for reuse. These accounts follow TamerOp's choices from the operation being requested to the objects retained and the checks that justify its result. Each develops the notation it needs and links the mathematical or usage context at the point where it is useful. You can begin with the computation you want to understand; there is no prescribed reading order.
Persistence and filtered complexes
- Relative persistence of filtered pairs — Implementation account. How are quotient persistence and its connecting maps computed for a filtered pair?
- Retaining representatives and tracing them to source cells — Implementation account. Which representations and algorithms underlie retaining representatives and tracing them to source cells, and why?
Finite descriptions and parameter queries
- Axis-aligned box encoding — Implementation account. Which representations and algorithms underlie axis-aligned box encoding, and why?
- Change of poset, refinement and Kan extensions — Implementation account. Which representations and algorithms underlie change of poset, refinement and Kan extensions, and why?
Modules, maps and derived algebra
- From selected images to submodules and quotients — Implementation account. Which representations and algorithms underlie kernels, quotients and universal constructions, and why?
- Interval multiplicities and certified decomposition — Implementation account. Which representations and algorithms underlie interval multiplicities and certified decomposition, and why?
- Yoneda products and validated Tor algebra — Implementation account. Which representations and algorithms underlie yoneda products and validated Tor algebra, and why?
Geometric inputs and filtrations
- Exact rhomboid multicover construction — Implementation account. Which representations and algorithms underlie exact rhomboid multicover construction, and why?
Invariants, slices and distances
- Computing generalized rank and GRIL — Implementation account. How do selected finite diagrams and continuous grid worms reduce to exact limit-to-colimit ranks?
Features and data analysis
- Aligned ensembles and statistical uncertainty computations — Implementation account. How are repeated outputs aligned and which computations support the reported uncertainty?
- From sampled landscapes to ambient queries — Implementation account. How are retained line landscapes placed on an ambient grid, with their sampling and error contracts intact?
Visualization and interaction
- How a mathematical view becomes a figure — Implementation account. How does a prepared mathematical view retain its coefficients, coordinates and meaning while it is drawn or exported?
- Linked inspection sessions and browser execution — Implementation account. Which representations and algorithms underlie linked inspection sessions and browser execution, and why?
Data, saving and interchange
- Serialization, interchange and mathematical provenance — Implementation account. Which representations and algorithms underlie serialization, interchange and mathematical provenance, and why?
Fields, grades and numerical arithmetic
- Exact rational coordinates — Implementation account. Which representations and algorithms underlie exact rational coordinates, and why?
Supporting resources:
- Implementation bibliography. Which academic and software sources support the implementation accounts?
Exact rational coordinates follows a vector into a chosen basis. An invertible set of rows gives candidate coordinates; checking the remaining rows decides whether those coordinates describe the supplied vector. The same idea specializes to repeated homology queries.
Axis-aligned box encoding explains how the grid retains closed points, faces and corners, carries their maps, and certifies exact representatives without invoking general polyhedral feasibility.
Linked inspection sessions explains how related views change together after a selection, which computations can be reused and the difference between live Julia queries and captured browser states. Serialization and mathematical provenance follows those states into portable files while keeping the mathematical source, its coordinate bases and the selected drawing separate.
Computing generalized rank and GRIL follows compatible sections through a common quotient, then explains reconstruction on a chosen finite family, barcodes of paths along interval boundaries and exact widths for the regions used in GRIL.
Exact rhomboid multicover construction follows site labels through constrained radii and signed boundaries into finite encodings. It explains lifted cell geometry, closed depth caps and the separation between source sampling and exact selected-parameter algebra.
Retaining representatives and tracing them to source cells explains filtered chain maps, contractions between cells of equal grade, the reversal of direction when lifting cocycles, and a search for representatives using fewer source cells. Read it when a source overlay needs a mathematical correspondence between computed and original cell bases.
Change of poset, refinement and Kan extensions explains how order maps, actual linear maps and certificates relating finite source data to the transported result are retained when the parameter poset changes.
From selected images to submodules and quotients follows selected coordinates through the image engine and induced maps, then through the quotient projection. It explains why a generated submodule need not be a direct summand.
Yoneda products and validated Tor algebra follows comparison lifts and quotient coordinates, then explains what supplied chain products must satisfy before they define multiplication on Tor homology.
The landscape sampling account explains stored line parameters, ambient grid queries, missing samples and conditional interpolation bounds.
Aligned ensembles and uncertainty computations follows repeated observables through probe alignment, empirical reductions, confidence bands for bounded means and resampling of whole observations, then explains how visual selections use the prepared results.
Relative persistence of filtered pairs explains when one quotient complex suffices, when delayed subcomplex entry needs a filtered cone, and how the actual quotient bases and connecting maps are recovered.
Interval multiplicities and certified decomposition follows ambient Hom maps into a scalar pairing, then endpoint pruning and simultaneous splitting maps. It explains when a residual certifies that a module cannot be decomposed into intervals and when a selected family leaves that question open.