← all systems · see it in The System
Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked
In mathematics, a matrix is a rectangular array of numbers or other mathematical objects with elements or entries arranged in rows and columns, usually satisfying certain properties of addition and multiplication.
Properties published scienceour proposal
Sourced fields, and our own readings.
| aliases | Matrices (arrays standing for linear maps) |
| born at | 10th-2nd century BCE (arrays); 1850-1858 (the matrix proper) |
| born precision | centuries; contested |
| boundary | Inside: the practice of writing quantities in rows and columns and computing with them (adding, scaling, transposing, multiplying, reducing rows, taking determinants, finding eigenvalues, factoring), by hand, by human computers and by machines with matrix units, and the structure worked on: the matrix as an array over a field or ring, its algebra (associative, distributive, not commutative), its invariants (determinant, trace, rank, eigenvalues, characteristic polynomial), its correspondence with linear maps, its decompositions, matrix groups, and random matrices. Outside: the vector spaces and linear maps it represents (linear algebra, vector space), the systems it models (nuclei, ecosystems, populations, networks, neural networks), the people and machines that carry it, and the writing and memory that hold its entries. |
| nested level | Part of linear algebra (its explicit, computational form) and, as matrix theory, a branch of mathematics that grew into graph theory, algebra, combinatorics and statistics; runs on writing and on computer memory; carried by people (rung 11), computers (rung 12) and trained networks (rung 13). |
| scale | social |
Measures (16) published science
| characteristic clock: TPU weight-tile shift | 256 clock cycles |
| characteristic clock: one matrix-vector product per iteration (power iteration, Markov step, population projection) | 1 product per step |
| TPU matrix unit size | 65,536 (256 × 256) 8-bit multiply-accumulators |
| TPU peak throughput | 92 tera-operations per second |
| share of cycles waiting for weights (CNN1 on TPU) | about 35 percent |
| useful multiply-accumulators on active cycles (CNN1) | about half of 65,536 MACs |
| TPU speed and efficiency against contemporary GPU or CPU | about 15-30 times faster; TOPS/Watt about 30-80 times higher ratio |
| multiplications for a 2 × 2 product (Strassen 1969) | 7 (not 8) multiplications |
| multiplications for a 4 × 4 product in arithmetic mod 2 (AlphaTensor 2022) | 47 (Strassen two-level: 49) multiplications |
| matrix multiplication exponent: best upper bound (August 2026) | ω < 2.371177 exponent |
| laser method on CW5: limit on provable ω | 2.3078 exponent |
| Leibniz formula: number of terms for n = 5 | 120 (n!) terms |
| single-cell data: random-matrix noise / sparsity artefact / biological signal | about 95 / 3 / 2 percent of information |
| May's stability threshold for random community matrices | complexity σ√(SC) > 1 gives probability of stability near 0 (large S) dimensionless |
| fish species with age-structured Leslie models in one study | 30 species |
| dates: Binet's product 1812; Cauchy real eigenvalues 1829; Sylvester's name 1850; Cayley's memoir 1858; Cullis notation 1913; matrix mechanics 1925; transfer matrix 1941; Strassen 1969 | 1812-1969 year |
Open contentions (4) open contention
Where sources or readings disagree. Both stay on the map.
- C-413 Where do thoughts come from?
booked: No theory is proven, so the map lists the main ones side by side and doesn't pick one. It shows where thinking is handled in the brain, not where thoughts start.
the other reading: the theories listed in C-413 - C-1188 Which parts of the brain do advanced mathematics?
booked: The map records that experts use a network across the frontal, parietal and temporal lobes, not language areas, while noting this is inferred from related fields and not tested for each field by name.
the other reading: no imaging study names matrices: inferred from neighbouring domains - C-1198 When was China's Nine Chapters on the Mathematical Art written?
booked: The map uses parts written as early as about 150 BC, and keeps the wider 10th-2nd century BCE range alongside.
the other reading: parts written as early as about 150 BC, first reference by title 179 AD (Gaussian elimination article) - C-2993 Do the zeros of the Riemann zeta function follow the same law as random-matrix eigenvalues?
booked: The map keeps the competing readings; none is ruled.
the other reading: proved only in restricted ranges and under the Riemann hypothesis (Montgomery 1973; Baluyot et al. 2025 remove RH under weaker conditions); sums of two ordinates show no level repulsion (Banks 2025): the GUE link holds for the ordinates, not every statistic built on them
Partners and parts published science
Built from / builds (7)
- Linear algebra (the practice of solving, transforming and taking apart linear systems with vectors and matrices, and the structures it works with)
- Eigenvalues and eigenvectors
- Matrix decompositions (LU, QR, SVD, eigendecomposition)
- Matrix multiplication
- Entries (elements) indexed by row and column
- Determinant
- Competition between matrix mechanics and wave mechanics as the formulation of quantum theory (1925-26)
Sits inside / holds (3)
- Vector space: sets whose elements can be added and scaled, as mathematics describes it, with bases and dimension, linear maps, inner products, and their use as the state spaces of physics and data
- Mathematics (the field: the practice and the structures it studies)
- Linear algebra (the practice of solving, transforming and taking apart linear systems with vectors and matrices, and the structures it works with)
Trades with (valence) (16)
- Vector space: sets whose elements can be added and scaled, as mathematics describes it, with bases and dimension, linear maps, inner products, and their use as the state spaces of physics and data via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Atomic nucleus via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Human (Homo sapiens) via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Computer memory via a matrix in a machine: entries held in memory and streamed through a matrix multiply unit · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked mixed
- Quantum mechanics via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Ising spins on a causal set via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Habitat and ecosystem via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Large language model via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Recursion: defining or computing a thing in terms of smaller versions of itself, as computer science and mathematics describe it, with its base cases, call stack, recursive data, induction, and its sightings in language and nature via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Mathematics (the field: the practice and the structures it studies) via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Computer via a matrix in a machine: entries held in memory and streamed through a matrix multiply unit · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Linear algebra (the practice of solving, transforming and taking apart linear systems with vectors and matrices, and the structures it works with) via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Riemann zeta function via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Fish via sys:written-record · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
- Series (mathematics): sums of infinitely many terms, and the practice of summing, testing and using them via sys:sensory-environment · Matrix (mathematics): the rectangular array of numbers and its algebra, determinants and eigenvalues, matrices as linear maps, their history, and their use in physics, graphics and machine learning; the practice and the structure both booked gain
expressed by (9)
- Recombination/Composition
- Templating / Standardization
- Interoperability/Transduction (Bridging)
- Encapsulation & Modularity (Abstraction)
- Amplification/Scale-Up (Manufacture & Distribution)
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Surprise / Change-Point Detection
- Prepositioning & Latency
floor filled by (21)
- Vector space: sets whose elements can be added and scaled, as mathematics describes it, with bases and dimension, linear maps, inner products, and their use as the state spaces of physics and data
- Atomic nucleus
- Human (Homo sapiens)
- Computer memory
- Quantum mechanics
- Habitat and ecosystem
- Large language model
- Recursion: defining or computing a thing in terms of smaller versions of itself, as computer science and mathematics describe it, with its base cases, call stack, recursive data, induction, and its sightings in language and nature
- Computer
- Statistics: the practice of collecting and reasoning from data, the intuitive statistics it grows from, and the structure (samples, distributions, estimators) it works on
- Linear algebra (the practice of solving, transforming and taking apart linear systems with vectors and matrices, and the structures it works with)
- Systems of linear equations
- Eigenvalues and eigenvectors
- Matrix decompositions (LU, QR, SVD, eigendecomposition)
- Matrix multiplication
- Computer graphics and visualization
- Entries (elements) indexed by row and column
- Determinant
- Square matrix
- Random matrix
- Tensor processing unit
Kinds (10)
- Square matrix
- Symmetric and Hermitian matrix
- Orthogonal matrix
- Stochastic (Markov transition) matrix
- Adjacency matrix
- Sparse matrix
- Random matrix
- Infinite matrix
- Leslie (age-structured population) matrix
- Transformation matrix (affine and projective, 4 × 4 in graphics)
Notes published science
- In July 1925 Heisenberg wrote down a strange rule for multiplying tables of atomic transition amplitudes and was unsettled that x times y need not equal y times x; Max Born recognised it as matrix multiplication, which Heisenberg did not know, and quantum mechanics' first full form was born within months (src:aitchison-2004-heisenberg).
- The spacings between energy levels of heavy nuclei, and the conjectured spacings between zeros of the Riemann zeta function, follow the same law as eigenvalues of random matrices, which repel each other; Wigner first guessed the law standing up at a 1956 conference (src:livan-2017-rmt, src:wp:random-matrix).
- Robert May used random matrices to show that big, richly connected ecosystems should be unstable; Allesina and Tang found that predator-prey pairs (one positive, one negative effect) stabilise a community while mutualists and competitors destabilise it, because the sign of each pair's product flips the effect (src:allesina-tang-2012).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
a matrix as mathematicians describe it: numbers arranged in rows and columns, combined by the rules of matrix algebra, read for its determinant and eigenvalues, and used as a linear map (reader's view)
- L1: Quantities that depend on two indices: coefficients of simultaneous equations, the moves of a transformation, data tables, links in networks, transition amplitudes of quantum systems.
- L2: The array taken in: m rows by n columns of entries, each found by its row and column.
- L3: Size and kind: m × n; square, diagonal, triangular, symmetric, orthogonal, stochastic, sparse; entries in a field, a ring or a semiring.
- L4: Operating: addition, scalar multiplication, transpose, row operations, the row-by-column product (associative, not commutative).
- L5: Invariants: determinant, trace, rank, eigenvalues and eigenvectors, characteristic polynomial; conditioning.
- L6: The structure: matrix as linear map, change of basis, decompositions, matrix groups, infinite matrices, random-matrix ensembles.
- L7: Use: solving systems, graphics, numerical computation, physics, statistics, networks, ecology, machine learning.
a matrix in a machine: entries held in memory and streamed through a matrix multiply unit
- L1: Entries as bits in off-chip memory and a 28 MiB software-managed on-chip memory.
- L2: Weight tiles loaded into the unit (64 KiB, double-buffered; 256 cycles to shift a tile in).
- L3: The systolic array: 256 × 256 multiply-accumulate cells; data from the left, weights from the top, combined as a diagonal wavefront.
- L4: Accumulators and the next layer: one layer's activations must finish before the next layer's products begin.
the boardroom (shared floors)
- L4: Operating: addition, scaling, transpose, row operations, and the row-by-column product (which composes maps and does not commute); done by hand, by elimination, by BLAS, by systolic matrix units, by Strassen-type and AlphaTensor algorithms.
- L5: The numbers read off a square matrix: determinant (invertibility, volume, orientation), trace, rank, eigenvalues and eigenvectors, the characteristic polynomial (Cayley-Hamilton), the largest eigenvalue for repeated application; conditioning for computation.
- L6: The structure: matrices as linear maps (one matrix per map and basis), matrix groups, decompositions (LU, SVD, eigendecomposition, Jordan form), infinite matrices and operators, random-matrix ensembles and their universal laws.
- L7: Use: solving systems, geometry and graphics, numerical computation, physics (matrix mechanics, normal modes, S-matrix, transfer matrices), statistics (covariance, Markov chains), networks, ecology (community and Leslie matrices), cryptography, and machine learning (weight matrices, attention).
way in: Data: measurements and records arranged as tables (word counts per document, gene counts per cell)
- L1: Many measurements of many variables (survey answers, word frequencies, RNA counts in single cells).
- L2: The data matrix: one row per sample, one column per variable; single-cell matrices are very sparse.
- L3: The covariance (Wishart) matrix of the data.
way in: Networks: who is linked to or acts on whom (graphs, web links, species interactions)
- L1: A network: vertices joined by edges, websites by hyperlinks, cities by roads, species by feeding and competition.
- L2: Each link or effect recorded by its two ends (row: affected, column: acting).
- L3: The adjacency matrix or community matrix.
way in: Physical systems: spectra, transition amplitudes and coupled motions (atoms, nuclei, magnets, vibrating molecules)
- L1: Microscopic systems whose observed quantities are frequencies and intensities of transitions, and heavy nuclei with dense energy levels.
- L2: Measured spectra and level spacings (Wigner's 1956 surmise for nuclear level spacings).
- L3: Matrix elements: the amplitudes X(n, n - α) arranged by initial and final state (Heisenberg 1925, recognised by Born).
way in: Populations by age or stage: survival and births from life tables
- L1: A population of one species structured by age (30 fish species around Gulf of Mexico oil platforms in one study).
- L2: Life-history data: survival and fecundity by age (FishBase, FishLife, size-dependent mortality).
- L3: The Leslie matrix: fecundities on the top row, survival probabilities below the diagonal.
way in: Sight: figures transformed (a unit square sent to a parallelogram, a scene rotated and projected)
- L1: Shapes in a plane or space: the unit square, a 3D object, a camera view.
- L2: The moved figure seen: the square's corners land on a parallelogram whose area is |det A|.
- L3: Coordinates: the images of the basis vectors become the columns; translations folded in with homogeneous coordinates (n + 1 dimensions).
way in: Sight: the array written (rows, columns, brackets) read on paper or screen
- L1: Marks on paper or screen: an m × n block of entries in brackets.
- L2: Reading by row and column: each entry found by its two indices; numbers and formulas activate a bilateral fronto-parietal-temporal network.
- L3: The notation A = [a_ij] and bracket form (Cullis 1913): the array becomes one named object.
way in: Simultaneous linear equations from a problem (survey, accounts, circuits)
- L1: A problem with several unknowns tied by linear conditions (the Nine Chapters' grain problems; Cardano 1545; Seki 1683).
- L2: The coefficients collected, one row per equation and one column per unknown.
- L3: The coefficient matrix and the matrix equation Ax = b.
way in: The machine route: weights and activations streamed from memory into a matrix unit
- L1: Weights and inputs held as bits in off-chip memory and on-chip buffers (28 MiB on the TPU).
- L2: Loading: weights fetched into the unit; the unit stalls when data is not ready.
- L3: A 64 KiB tile of weights in the unit, with a second tile double-buffered.
way in: Thought: a matrix and its properties held and turned over in the mind
- L1: What is already held (concepts, the problem in mind); where a mathematical thought comes from is not settled.
- L2: Mathematical judgement in experts: a bilateral prefrontal, parietal and inferior temporal network, sparing language areas.
- L3: The matrix as a mental object (an operator, a transformation) rather than an array; how this code is formed is not shown in the sources read.
way out: Decisions: management, conservation and design choices taken from matrix metrics
- L1: Harvest and protection applied to the population.
- L5: Eigen-metrics: dominant eigenvalue, damping ratio, sensitivity and elasticity matrices.
- L7: Metrics for managers (growth rate, damping ratio, elasticity).
way out: Machine output: products written back to memory, the next layer's inputs, a model's outputs
- L1: Results written to the buffer, read by the next layer.
- L4: Multiply-accumulate as a diagonal wavefront through the systolic array.
- L7: A network's layer or an attention step to compute.
way out: Prediction: spectra, energy levels and modes computed and checked against experiment
- L1: The prediction tested against measured lines and modes.
- L5: Eigenvalues computed: energies, normal-mode frequencies.
- L7: A physical question (the hydrogen spectrum, a molecule's vibrations).
way out: Rendering: transformed objects projected onto a screen
- L1: Pixels lit on a two-dimensional screen.
- L4: Chained 4 × 4 transforms (translate, rotate, scale, project).
- L7: A scene to show.
way out: Writing the result: the reduced array, the product, the solution vector
- L1: Marks on the page, read by the next person.
- L2: Written in brackets, entry by entry, or printed in a paper.
- L4: The computed array (a row-reduced form, a product, an inverse).
Sources (24) published science
Eigenvalues and eigenvectors (Wikipedia, via Wikimedia Enterprise) (behind its links)
Allesina and Tang, Stability criteria for complex ecosystems (arXiv:1105.2071; Nature 483, 2012) (behind its links)
Fujiwara et al., Constructing age-structured matrix population models for all fishes, PeerJ (2025), doi:10.7717/peerj.18387 (PMC11700495) (behind its links)
Jouppi et al., In-Datacenter Performance Analysis of a Tensor Processing Unit (arXiv:1704.04760; ISCA 2017) (behind its links)
Alman, Duan, Vassilevska Williams, Xu, Xu, Zhou, More Asymmetry Yields Faster Matrix Multiplication (arXiv:2404.16349v3, 2026-08) (behind its links)
Matrix multiplication (Wikipedia, via Wikimedia Enterprise) (behind its links)
Linear algebra (Wikipedia, via Wikimedia Enterprise snapshot 2026-11-01) (behind its links)
Matrix mechanics (Wikipedia, via Wikimedia Enterprise) (behind its links)
Baluyot, Goldston, Suriajaya, Turnage-Butterbaugh, Pair Correlation of Zeros of the Riemann Zeta Function I (arXiv:2501.14545) (behind its links)
Banks, Pair correlation for sums of two ordinates of zeros of the Riemann zeta function (arXiv:2502.20569) (behind its links)
Folk, Holovatch, Lipowski, Ising's Roots and the Transfer-Matrix Eigenvalues, Entropy 26 (2024), doi:10.3390/e26060459 (PMC11202636) (behind its links)
Fawzi A, Balog M, et al. (2022) Discovering faster matrix multiplication algorithms with reinforcement learning. Nature 610:47-53. doi:10.1038/s41586-022-05172-4 (behind its links)
Computational complexity of matrix multiplication (Wikipedia, via Wikimedia Enterprise) (behind its links)
Determinant (Wikipedia, via Wikimedia Enterprise) (behind its links)
Aparicio, Bordyuh, Blumberg, Rabadan, A Random Matrix Theory Approach to Denoise Single-Cell Data, Patterns 1 (2020), doi:10.1016/j.patter.2020.100035 (PMC7660363) (behind its links)
Transformation matrix (Wikipedia, via Wikimedia Enterprise) (behind its links)
Livan, Novaes, Vivo, Introduction to Random Matrices: Theory and Practice (arXiv:1712.07903, open textbook) (behind its links)
Aitchison, MacManus, Snyder, Understanding Heisenberg's 'magical' paper of July 1925 (arXiv:quant-ph/0404009; Am. J. Phys. 72, 2004) (behind its links)
Vaswani et al. 2017, Attention Is All You Need (arXiv:1706.03762) (behind its links)
Random matrix (Wikipedia, via Wikimedia Enterprise) (behind its links)
Vector space (Wikipedia, via Wikimedia Enterprise) (behind its links)