What does Zero Explanation Tax and Artifact Gravity conclude, in numbers, about “Projection Loss: the Real Bill for Rank Truncation, Retelling Chains and Random Projection”?

Answer: Chapter 1 uses the theorem of Eckart and Young to show that a steep-spectrum work keeping 6 dimensions loses only 0.0446 of its energy, and that three retellings leave 0.4277 bits of each bit.

Scope of the numbers: All parameters are illustrative and all calculations are generated by the book's programs. The book is general business information, not individual advice, psychological counselling or clinical judgement; it makes no promise of income, and rendering an idea means only turning it into a deliverable, verifiable artifact.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.0446

A. population spectra, d12, n = 600 points
rank r steep lost flat lost (exact, population)
10.5991 0.8775
20.3586 0.7662
30.2143 0.6641
40.1277 0.5699
60.0446 0.4007
80.0147 0.2521
B. sample covariance, Jacobi eigenvalues (top 4)
steep1.0409 0.6071 0.3750 0.2108
flat0.9802 0.9122 0.8710 0.8164
C. sample lost share at r3: steep 0.2045, flat 0.6657
D. steep: smallest r losing <0.05 of energy is 6
D. flat: smallest r losing <0.05 of energy is 12
Show the full program
# Chapter 1. p11: energy lost by a rank-r projection (Eckart-Young).
# A work is a cloud of n points in d = 12 need-dimensions. Its sample
# covariance is diagonalised by a hand-written cyclic Jacobi method; the
# share of energy lost by keeping the top r directions equals the sum of
# the discarded eigenvalues over the trace (Eckart and Young, 1936).
# Two illustrative spectra: "steep" (0.6 ** i) and "flat" (1 / (1 + 0.1 i)).
# All data are synthetic; seeds are fixed literals.
import math
import random

D = 12
N = 600


def gram_schmidt(vectors):
    basis = []
    for v in vectors:
        w = v[:]
        for b in basis:
            dot = math.fsum(x * y for x, y in zip(w, b))
            w = [x - dot * y for x, y in zip(w, b)]
        norm = math.sqrt(math.fsum(x * x for x in w))
        basis.append([x / norm for x in w])
    return basis


def jacobi_eigenvalues(a, tol=1e-18, sweeps=60):
    n = len(a)
    a = [row[:] for row in a]
    for _ in range(sweeps):
        off = math.fsum(a[i][j] ** 2 for i in range(n)
                        for j in range(i + 1, n))
        if off < tol:
            break
        for p in range(n - 1):
            for q in range(p + 1, n):
                if abs(a[p][q]) < 1e-300:
                    continue
                theta = (a[q][q] - a[p][p]) / (2.0 * a[p][q])
                sign = 1.0 if theta >= 0 else -1.0
                t = sign / (abs(theta) + math.sqrt(theta * theta + 1.0))
                c = 1.0 / math.sqrt(t * t + 1.0)
                s = t * c
                for k in range(n):
                    akp, akq = a[k][p], a[k][q]
                    a[k][p] = c * akp - s * akq
                    a[k][q] = s * akp + c * akq
                for k in range(n):
                    apk, aqk = a[p][k], a[q][k]
                    a[p][k] = c * apk - s * aqk
                    a[q][k] = s * apk + c * aqk
    return sorted((a[i][i] for i in range(n)), reverse=True)


def sample_cov(spectrum, rng):
    raw = [[rng.gauss(0.0, 1.0) for _ in range(D)] for _ in range(D)]
    q = gram_schmidt(raw)
    pts = []
    for _ in range(N):
        z = [rng.gauss(0.0, math.sqrt(lam)) for lam in spectrum]
        pts.append([math.fsum(z[i] * q[i][j] for i in range(D))
                    for j in range(D)])
    mean = [math.fsum(p[j] for p in pts) / N for j in range(D)]
    cov = [[math.fsum((p[i] - mean[i]) * (p[j] - mean[j]) for p in pts)
            / (N - 1) for j in range(D)] for i in range(D)]
    return cov


def lost_share(eig, r):
    return math.fsum(eig[r:]) / math.fsum(eig)


if __name__ == "__main__":
    rng = random.Random(43211)
    steep = [0.6 ** i for i in range(D)]
    flat = [1.0 / (1.0 + 0.1 * i) for i in range(D)]
    print("A. population spectra, d = {}, n = {} points".format(D, N))
    print("   rank r   steep lost   flat lost   (exact, population)")
    for r in (1, 2, 3, 4, 6, 8):
        print("   {:>6}   {:>10.4f}   {:>9.4f}".format(
            r, lost_share(steep, r), lost_share(flat, r)))
    es = jacobi_eigenvalues(sample_cov(steep, rng))
    ef = jacobi_eigenvalues(sample_cov(flat, rng))
    print("B. sample covariance, Jacobi eigenvalues (top 4)")
    print("   steep: " + "  ".join("{:.4f}".format(x) for x in es[:4]))
    print("   flat:  " + "  ".join("{:.4f}".format(x) for x in ef[:4]))
    print("C. sample lost share at r = 3: steep {:.4f}, flat {:.4f}"
          .format(lost_share(es, 3), lost_share(ef, 3)))
    for name, e in (("steep", es), ("flat", ef)):
        need = next(r for r in range(1, D + 1)
                    if lost_share(e, r) <= 0.05)
        print("D. {}: smallest r losing <= 0.05 of energy is {}"
              .format(name, need))

Single source and citation

Fan, Y. (2026). Zero Explanation-Tax Dynamics: Abandoning Validation-Seeking, Full-Power Reality Rendering and Proof-by-Artifact Gravity (Version 1.0.0), Chapter 1: Projection Loss: the Real Bill for Rank Truncation, Retelling Chains and Random Projection. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-432 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026432,
  author    = {Fan, Yafu},
  title     = {Zero Explanation-Tax Dynamics: Abandoning Validation-Seeking, Full-Power Reality Rendering and Proof-by-Artifact Gravity},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book