About

A binary matrix is a matrix whose entries belong exclusively to the set {0, 1}. These structures appear in graph adjacency representations, digital image masks, combinatorial optimization, error-correcting codes, and finite automaton transition tables. Getting the density parameter d wrong in a Monte Carlo simulation or network adjacency test produces statistically invalid results. A truly random binary matrix of dimension m × n with density d assigns each cell independently: a_ij = 1 with probability d, and 0 otherwise. This tool generates matrices up to 128 × 128 using a seeded Linear Congruential Generator for full reproducibility.

Eight structural patterns are supported: pure random, symmetric (where a_ij = a_ji), identity, checkerboard, diagonal, Toeplitz (constant along diagonals), sparse, and block diagonal. The tool approximates ideal randomness via LCG. For cryptographic applications, use a CSPRNG instead. Results are exportable as PNG, SVG, or plain text for direct integration into papers, code, or presentations.

Formulas

Each cell a_ij of a random binary matrix is determined by comparing a pseudorandom number against the density threshold:

a_ij = {

1 if rand(i, j) < d0 otherwise

The pseudorandom function uses a Linear Congruential Generator (LCG):

X_n+1 = (a ⋅ X_n + c) mod m

where a = 1664525, c = 1013904223, m = 2³² (Numerical Recipes constants). The normalized output r = X_n ÷ m falls in [0, 1).

For a symmetric matrix, only the upper triangle is generated randomly. The lower triangle mirrors it: a_ji = a_ij for all i < j. The expected number of 1s in a random matrix is E = m × n × d, where d ∈ [0, 1] is the density parameter.

Variable legend: m = row count, n = column count, d = density (probability of 1), X₀ = seed value, a_ij = cell value at row i, column j.

Reference Data

Pattern	Definition	Typical Use Case	Symmetry	Density Control	Deterministic
Random	a_ij = 1 if r < d	Monte Carlo simulation, fuzz testing	No	Yes	With seed
Identity	a_ij = 1 iff i = j	Linear algebra neutral element	Yes	No (fixed)	Yes
Symmetric	a_ij = a_ji	Undirected graph adjacency	Yes	Yes	With seed
Diagonal	a_ij = 0 if i ≠ j	Scaling operators, masks	Yes	Yes	With seed
Checkerboard	a_ij = (i + j) mod 2	Texture generation, parity checks	Anti-symmetric	No (≈50%)	Yes
Toeplitz	Constant along each diagonal	Convolution, signal processing	No	Yes	With seed
Sparse	Density forced ≤ 15%	Large-scale graph storage, CSR input	No	Capped	With seed
Block Diagonal	Non-zero blocks along diagonal	Disconnected subgraph modeling	Partial	Yes	With seed
Upper Triangular	a_ij = 0 if i > j	DAG adjacency, Gaussian elimination	No	Yes	With seed
Lower Triangular	a_ij = 0 if i < j	Cholesky factor structure	No	Yes	With seed
Anti-diagonal	a_ij = 1 iff i + j = n − 1	Permutation matrices, reflections	Yes	No (fixed)	Yes
Hadamard-like	Rows are orthogonal binary vectors	Error-correcting codes (Walsh)	Yes	No (50%)	Yes
Circulant	Each row is a cyclic shift of the first	Cyclic codes, DFT structure	No	Yes	With seed

Frequently Asked Questions

The seed initializes the LCG's internal state X₀. Identical seed, dimensions, density, and pattern will always produce the exact same matrix. This is critical for reproducible experiments. Changing the seed by even 1 produces a completely different output due to the multiplicative constant a = 1664525 amplifying small differences across iterations.

Sparse mode hard-caps density at 15% regardless of the slider position, ensuring matrices suitable for compressed sparse row (CSR) storage testing. Random mode with low density achieves similar visual results but doesn't enforce the cap. If you set density to 0.05 in Random mode, you get approximately the same output. Sparse mode is a convenience guard.

The identity matrix I_n is deterministic: a_ij = 1 iff i = j. Its density is always 1÷n. Applying an external density parameter would violate the definition. The same applies to Checkerboard and Anti-diagonal patterns.

No. The LCG used here (with constants from Numerical Recipes) is a fast PRNG but is not cryptographically secure. Its internal state can be predicted after observing a few outputs. For cryptographic binary matrices, use crypto.getRandomValues() or a CSPRNG. This tool is designed for simulation, education, and testing purposes.

A Toeplitz matrix satisfies a_ij = a_i+1,j+1. The generator creates m + n − 1 random binary values (one per diagonal) and maps each cell to its diagonal index j − i. This structure appears naturally in discrete convolution and autocovariance matrices.

At 128 × 128 (16384 cells), the LCG generates all values in under 2ms on modern hardware. The canvas renderer draws each cell as a colored rectangle. Text output at this scale produces a 16384-character string. The tool remains responsive. Performance degrades only if you attempt to copy extremely large text outputs to clipboard, which browsers handle asynchronously.