Awesome-Integrated-Sensing-and-Communications/code/baselines/isac_capacity_distortion/reproducibility.yaml at main · yuanhao-cui/Awesome-Integrated-Sensing-and-Communications · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
---
schema_version: 1
baseline: isac_capacity_distortion
reference:
  doi: 10.1109/TIT.2023.3284449
  manuscript: https://arxiv.org/abs/2204.06938
evidence_level: educational-surrogate
paper_figure_parity: false
model: >-
  Gaussian MIMO covariance rate, generic BFIM and CRB, known-unit-gain ULA
  angle information, water filling, and a local log-determinant surrogate.
assumptions:
  - Total covariance trace equals power per transmitter times transmitter count.
  - Communication rate is reported in nats per channel use.
  - The local angle model has fixed angle, known unit gain, and one path.
  - Communication coefficients are normalized by sqrt(noise variance) before products.
  - The certificate uses a seeded two-mode synthetic diagonal channel.
parameters:
  seed: 20260717
  paper_anchors:
    transmit_antennas: 10
    sensing_receive_antennas: 10
    communication_receive_antennas: 1
    array_spacing_wavelengths: 0.5
    max_sensing_receive_snr_db_per_antenna: 20
    max_communication_receive_snr_db_per_antenna: 33
    angle_prior_distribution: von-Mises
    angle_prior_mean_degrees: 30
    angle_prior_standard_deviation_degrees: 5
    figure_scenario_communication_bearing_degrees: 42
    figure_scenario_coherent_interval: 3
    reported_subspace_overlap: approximately-0.61
  dynamic_range_channels:
    - extreme-scale channel/noise pair
    - representable subnormal Gram eigenvalues
    - 1.0e154 scalar channel for the covariance-surrogate quadratic root
numeric_domain:
  arithmetic: IEEE-754 binary64
  guarantees:
    - scale-safe BFIM scaling
    - direct received-waveform rate evaluation
    - ratio-scaled covariance-surrogate quadratic root
    - explicit geometry-domain rejection before phase construction
  rejected:
    - nonrepresentable Hermitian sums, Gram products, or received waveforms
    - array phases outside the declared reliable-cycle domain
oracle:
  type: dense-grid-analytic-and-active-set
  expected:
    water_filling_matches_500001_point_grid: true
    covariance_surrogate_matches_grid_and_kkt: true
    diagonal_bfim_crb_and_siso_rate_match_closed_form: true
    subnormal_gain_active_set_is_correct: true
    invalid_or_unrepresentable_domains_are_explicit: true
tolerances:
  dense_grid_steps: 1.1
  objective_absolute: 1.0e-10
  kkt_absolute: 2.0e-12
  analytic_absolute: 2.0e-14
  dynamic_range_relative: 2.0e-15
  subnormal_gain_relative: 1.0e-12
  steering_derivative_relative: 1.0e-8
  physical_constraint_absolute: 2.0e-12
evidence:
  command: >-
    python -m
    code.baselines.isac_capacity_distortion.examples.verify_surrogate --json
  certificate_checks:
    - bfim_analytic
    - bfim_scale_safe
    - crb_analytic
    - dynamic_range_scaling
    - physical_postconditions
    - received_waveform_cancellation_permutation_invariant
    - received_waveform_rate_scale_safe
    - seed_repeatability
    - siso_rate_analytic
    - steering_derivative
    - subnormal_gain_scaling
    - surrogate_discriminant_scale_safe
    - surrogate_power_gain_product_scale_safe
    - surrogate_grid_objective
    - surrogate_grid_power
    - surrogate_kkt
    - unrepresentable_outputs_are_explicit
    - water_grid_objective
    - water_grid_power
    - weak_information_is_finite
  checks:
    - 500001-point simplex grid for water filling
    - 500001-point simplex grid and KKT residual for covariance surrogate
    - closed-form diagonal BFIM, CRB, weak-information CRB, and SISO rate
    - extreme-scale channel/noise water-filling and rate oracle
    - subnormal-gain active-set water-filling and positive-rate oracle
    - overflow-free covariance-surrogate root at squared gain 1.0e308
    - >-
      finite two-mode KKT solution when power budget times gain exceeds binary64
    - explicit CRB-infinity and oversized-power-budget checks
    - central finite-difference ULA steering derivative
    - seeded bitwise-repeatability check
    - scale-safe BFIM scaling across extreme finite channel amplitudes
    - direct received-waveform rate without an overflowing Gram intermediate
    - explicit geometry-domain rejection before phase construction
limitations:
  - The paper unknown-gain nuisance-parameter Bayesian angle FIM is absent.
  - Paper prior averaging, fading ensembles, and overlap coefficient are absent.
  - No published curve or signaling theorem is numerically certified.
  - Noise-normalized coefficients, energies, and total trace must fit binary64.
  - CRB values beyond binary64 are represented by positive infinity.