aircompkit 1.0.0__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.


This version of aircompkit might be problematic. Click here for more details.

Files changed (39) hide show
  1. aircompkit/__init__.py +32 -0
  2. aircompkit/__main__.py +5 -0
  3. aircompkit/adversary/__init__.py +175 -0
  4. aircompkit/aircomp/__init__.py +41 -0
  5. aircompkit/aircomp/analytic.py +42 -0
  6. aircompkit/aircomp/power_control.py +183 -0
  7. aircompkit/aircomp/preprocess.py +48 -0
  8. aircompkit/bench.py +159 -0
  9. aircompkit/channel/__init__.py +209 -0
  10. aircompkit/cli.py +165 -0
  11. aircompkit/core/__init__.py +1 -0
  12. aircompkit/core/registry.py +123 -0
  13. aircompkit/core/result.py +45 -0
  14. aircompkit/core/rng.py +47 -0
  15. aircompkit/core/scenario.py +226 -0
  16. aircompkit/defense/__init__.py +156 -0
  17. aircompkit/defense/evaluate.py +48 -0
  18. aircompkit/devices/__init__.py +79 -0
  19. aircompkit/learning/__init__.py +32 -0
  20. aircompkit/learning/aggregate.py +49 -0
  21. aircompkit/learning/array.py +140 -0
  22. aircompkit/learning/bank.py +86 -0
  23. aircompkit/learning/data.py +128 -0
  24. aircompkit/learning/fl.py +249 -0
  25. aircompkit/learning/models.py +89 -0
  26. aircompkit/learning/partition.py +58 -0
  27. aircompkit/metrics/__init__.py +36 -0
  28. aircompkit/metrics/privacy.py +36 -0
  29. aircompkit/plot/__init__.py +5 -0
  30. aircompkit/receiver/__init__.py +244 -0
  31. aircompkit/reproduce/__init__.py +26 -0
  32. aircompkit/reproduce/yang2020.py +182 -0
  33. aircompkit/reproduce/zhu2020.py +205 -0
  34. aircompkit-1.0.0.dist-info/METADATA +167 -0
  35. aircompkit-1.0.0.dist-info/RECORD +39 -0
  36. aircompkit-1.0.0.dist-info/WHEEL +5 -0
  37. aircompkit-1.0.0.dist-info/entry_points.txt +2 -0
  38. aircompkit-1.0.0.dist-info/licenses/LICENSE +21 -0
  39. aircompkit-1.0.0.dist-info/top_level.txt +1 -0
@@ -0,0 +1,205 @@
1
+ """Reproduction: G. Zhu, Y. Wang, K. Huang, "Broadband analog aggregation for low-latency
2
+ federated edge learning", IEEE Trans. Wireless Commun., 19(1), 2020 (arXiv:1812.11494).
3
+
4
+ Paper settings (Sec. VI-A): K = 200 devices uniform in a disc of radius R = 100 m, path-loss
5
+ exponent alpha = 3, M = 1000 sub-channels, average transmit power P0 = 0.1 W per device
6
+ (P0/M per sub-channel), noise N0 = -80 dBm per sub-channel, Rayleigh fading, truncated channel
7
+ inversion with an average power budget, MNIST split IID (200 equal shares) or non-IID
8
+ (400 label-sorted shards of 150, two per device).
9
+
10
+ * **Fig. 4** (SNR-truncation trade-off): ``rho0 = P0 / (M r_max^alpha E1(g_th))`` against the
11
+ truncation ratio ``1 - exp(-g_th)``, with r_max = 100. Checked against Monte Carlo: the
12
+ empirical truncation ratio, the empirical *average* transmit power (must equal P0/M) and
13
+ ``eta/N0`` from the simulator's power control.
14
+ * **Fig. 7(b)** (alpha = 3): test accuracy against the cell-interior radius ratio R_in/R.
15
+ The paper uses a 582k-parameter CNN, an unstated number of rounds and a per-point grid
16
+ search over g_th; here an MLP (784-64-10), a fixed number of rounds and g_th = 0.1 are used,
17
+ so only the *trend* is compared. Values read by eye from the paper's bar chart (±0.02).
18
+
19
+ Known outcome (see the Reproductions page of the documentation): the data-limited rise reproduces; the SNR-limited
20
+ decline at large R_in does not appear with the stated SNRs, because SGD averages the
21
+ zero-mean channel noise over rounds (the paper's own Fig. 10 remarks on this robustness).
22
+ """
23
+
24
+ from __future__ import annotations
25
+
26
+ import json
27
+ from pathlib import Path
28
+ from typing import Any
29
+
30
+ import numpy as np
31
+
32
+ from aircompkit.aircomp.power_control import exp1
33
+ from aircompkit.core.registry import register
34
+ from aircompkit.core.rng import SeedTree
35
+ from aircompkit.core.scenario import Scenario
36
+
37
+ K, R, ALPHA, M = 200, 100.0, 3.0, 1000
38
+ P0, N0 = 0.1, 1e-11 # W; -80 dBm
39
+ FIG4_TRUNC = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 0.95, 0.99])
40
+ FIG7_RATIOS = (0.1, 0.3, 0.5, 0.7, 0.9)
41
+ PAPER_FIG7B = {"noniid": [0.72, 0.80, 0.85, 0.83, 0.79], "iid": [0.80, 0.85, 0.90, 0.86, 0.81]}
42
+
43
+
44
+ def rho0_db(trunc: float, r_max: float = R, alpha: float = ALPHA) -> float:
45
+ g = -np.log(1 - trunc)
46
+ return float(10 * np.log10(P0 / (M * r_max**alpha * exp1(g)) / N0))
47
+
48
+
49
+ def scenario(threshold: float, alpha: float = ALPHA, K_: int = K, seed: int = 0) -> Scenario:
50
+ return Scenario(
51
+ devices={"name": "uniform_disc", "K": K_, "p_max": P0 / M, "r_min": 1.0, "r_max": R},
52
+ channel={"name": "rayleigh", "pathloss": {"exponent": alpha}},
53
+ power_control={"name": "truncated_inversion", "threshold": threshold, "budget": "average"},
54
+ receiver="eigen",
55
+ noise_var=N0,
56
+ symbols="real",
57
+ seed=seed,
58
+ )
59
+
60
+
61
+ def fig4(trials: int = 20000, seed: int = 0) -> dict[str, Any]:
62
+ """Closed form vs. Monte Carlo for one device at r_max = R."""
63
+ out: dict[str, Any] = {"truncation_ratio": FIG4_TRUNC.tolist(), "closed_form_db": [],
64
+ "sim_rho0_db": [], "sim_truncation": [], "sim_avg_power_over_budget": []}
65
+ for z in FIG4_TRUNC:
66
+ g = float(-np.log(1 - z))
67
+ sc = Scenario(devices={"name": "fixed", "K": 1, "p_max": P0 / M, "distances": [R]},
68
+ channel={"name": "rayleigh", "pathloss": {"exponent": ALPHA}},
69
+ power_control={"name": "truncated_inversion", "threshold": g, "budget": "average"},
70
+ receiver="eigen", noise_var=N0, symbols="complex", seed=seed)
71
+ h_act, h_est, beta = sc.sample_channels(SeedTree(seed), trials, np.array([R]))
72
+ lay = sc.design(h_act, h_est, beta, sc.devices.power_budgets())
73
+ out["closed_form_db"].append(rho0_db(z))
74
+ out["sim_rho0_db"].append(float(10 * np.log10(lay["eta"].mean() / N0)))
75
+ out["sim_truncation"].append(float(np.mean(lay["p"][:, 0] == 0)))
76
+ out["sim_avg_power_over_budget"].append(float(lay["p"][:, 0].mean() / (P0 / M)))
77
+ return out
78
+
79
+
80
+ def fig7(data, rounds: int, threshold: float = 0.1, seed: int = 0,
81
+ ratios=FIG7_RATIOS, verbose: bool = False) -> dict[str, Any]:
82
+ from aircompkit.learning import FederatedConfig, train
83
+
84
+ out: dict[str, Any] = {"ratios": list(ratios), "rounds": rounds, "threshold": threshold}
85
+ for split, part in (("iid", "iid"), ("noniid", {"name": "shards", "shards_per_device": 2})):
86
+ ota, ideal, n_sched = [], [], []
87
+ for q in ratios:
88
+ sc = scenario(threshold, seed=seed)
89
+ cfg = {"dataset": data, "partition": part, "model": {"name": "mlp", "hidden": [64]},
90
+ "rounds": rounds, "eval_every": rounds, "lr": 0.1, "batch_size": 50,
91
+ "subchannels": M, "max_distance": q * R, "seed": seed}
92
+ try:
93
+ r = train(sc, FederatedConfig(**cfg))
94
+ ri = train(sc, FederatedConfig(**{**cfg, "aggregation": "ideal"}))
95
+ except ValueError: # nobody inside R_in for this placement
96
+ ota.append(float("nan"))
97
+ ideal.append(float("nan"))
98
+ n_sched.append(0)
99
+ continue
100
+ ota.append(float(r["test_acc"][-1]))
101
+ ideal.append(float(ri["test_acc"][-1]))
102
+ n_sched.append(int(r["n_scheduled"][0]))
103
+ if verbose:
104
+ print(f" {split:6s} R_in/R={q:.1f} devices={n_sched[-1]:3d} "
105
+ f"OTA {ota[-1]:.4f} ideal {ideal[-1]:.4f}")
106
+ out[split] = {"ota": ota, "ideal": ideal, "n_scheduled": n_sched, "paper": PAPER_FIG7B[split]}
107
+ return out
108
+
109
+
110
+ @register("reproduction", "zhu2020")
111
+ def run(trials: int | None = None, quick: bool = False, out: str | Path | None = None,
112
+ data_root: str | Path | None = None, rounds: int | None = None,
113
+ seed: int = 0) -> dict[str, Any]:
114
+ """Fig. 4 always; Fig. 7(b) when MNIST is available locally (see ``learning.data``)."""
115
+ res: dict[str, Any] = {"fig4": fig4(trials or (4000 if quick else 20000), seed)}
116
+ _print_fig4(res["fig4"])
117
+ try:
118
+ from aircompkit.learning.data import mnist
119
+
120
+ data = mnist(root=data_root)
121
+ except FileNotFoundError as e:
122
+ res["fig7"] = None
123
+ print(f"\nFig. 7(b) skipped: {str(e).splitlines()[0]}")
124
+ else:
125
+ n_rounds = rounds or (15 if quick else 100)
126
+ print(f"\nFig. 7(b): accuracy vs R_in/R (alpha=3, MLP, {n_rounds} rounds, g_th=0.1)")
127
+ res["fig7"] = fig7(data, n_rounds, seed=seed, verbose=True)
128
+ _print_fig7(res["fig7"])
129
+ if out is not None:
130
+ out = Path(out)
131
+ out.mkdir(parents=True, exist_ok=True)
132
+ (out / "zhu2020.json").write_text(json.dumps(res, indent=2))
133
+ _plot(res, out / "zhu2020_fig4_fig7.png")
134
+ return res
135
+
136
+
137
+ def _print_fig4(f4):
138
+ print("Zhu et al. 2020 Fig. 4: receive SNR vs truncation ratio (r_max = 100 m)")
139
+ print(" trunc closed-form dB sim dB sim trunc avg power / budget")
140
+ for row in zip(f4["truncation_ratio"], f4["closed_form_db"], f4["sim_rho0_db"],
141
+ f4["sim_truncation"], f4["sim_avg_power_over_budget"]):
142
+ print(" {:4.2f} {:10.3f} {:8.3f} {:7.4f} {:8.4f}".format(*row))
143
+
144
+
145
+ def _print_fig7(f7):
146
+ print(" R_in/R " + "".join(f"{q:>16.1f}" for q in f7["ratios"]))
147
+ for split in ("iid", "noniid"):
148
+ d = f7[split]
149
+ print(f" {split:6s} " + "".join(f" {o:5.3f}/{i:5.3f}/{p:4.2f}"
150
+ for o, i, p in zip(d["ota"], d["ideal"], d["paper"])))
151
+ print(" (OTA / ideal aggregation / paper)")
152
+
153
+
154
+ def _plot(res, path):
155
+ try:
156
+ import matplotlib.pyplot as plt
157
+ except ImportError:
158
+ return
159
+ f4, f7 = res["fig4"], res["fig7"]
160
+ ncols = 2 if f7 else 1
161
+ fig, axes = plt.subplots(1, ncols, figsize=(4.8 * ncols, 3.6))
162
+ a = axes[0] if ncols == 2 else axes
163
+ a.plot(f4["truncation_ratio"], f4["closed_form_db"], "k-", label="closed form (paper eq. 17)")
164
+ a.plot(f4["sim_truncation"], f4["sim_rho0_db"], "o", mfc="none", label="AirCompKit Monte Carlo")
165
+ a.set_xlabel("truncation ratio")
166
+ a.set_ylabel("receive SNR (dB)")
167
+ a.set_title("(a) cf. Zhu et al. Fig. 4", fontsize=9)
168
+ a.grid(alpha=0.3)
169
+ a.legend(fontsize=7)
170
+ if f7:
171
+ b = axes[1]
172
+ x = np.asarray(f7["ratios"])
173
+ for split, c in (("iid", "tab:green"), ("noniid", "tab:blue")):
174
+ b.plot(x, f7[split]["ota"], "o-", color=c, label=f"AirCompKit OTA, {split}")
175
+ b.plot(x, f7[split]["ideal"], ":", color=c, label=f"ideal aggregation, {split}")
176
+ b.plot(x, f7[split]["paper"], "x--", color=c, lw=0.8, label=f"paper (CNN), {split}")
177
+ b.set_xlabel(r"$R_{in}/R$")
178
+ b.set_ylabel("test accuracy")
179
+ b.set_title(rf"(b) cf. Zhu et al. Fig. 7(b), $\alpha$=3, {f7['rounds']} rounds", fontsize=9)
180
+ b.grid(alpha=0.3)
181
+ b.legend(fontsize=6)
182
+ fig.tight_layout()
183
+ fig.savefig(path, dpi=160)
184
+ print(f"saved {path}")
185
+
186
+
187
+ def check(res: dict[str, Any]) -> list[tuple[str, bool, str]]:
188
+ """Tolerance checks used by ``aircompkit reproduce zhu2020 --check`` (reproduction CI)."""
189
+ f4 = res["fig4"]
190
+ drho = max(abs(a - b) for a, b in zip(f4["sim_rho0_db"], f4["closed_form_db"]))
191
+ dtr = max(abs(a - b) for a, b in zip(f4["sim_truncation"], f4["truncation_ratio"]))
192
+ dpow = max(abs(a - 1) for a, b in zip(f4["sim_avg_power_over_budget"], f4["truncation_ratio"]) if b <= 0.9)
193
+ out = [("Fig. 4: designed rho_0 = closed form (0.1 dB)", drho <= 0.1, f"max {drho:.3g} dB"),
194
+ ("Fig. 4: truncation ratio as designed (0.03)", dtr <= 0.03, f"max {dtr:.3f}"),
195
+ ("Fig. 4: average power = budget (25 %, ratio <= 0.9)", dpow <= 0.25, f"max {dpow:.3f}")]
196
+ f7 = res.get("fig7")
197
+ if f7:
198
+ gaps = [abs(a - b) for s in ("iid", "noniid") for a, b in zip(f7[s]["ota"], f7[s]["ideal"])
199
+ if np.isfinite(a)]
200
+ out.append(("Fig. 7(b): OTA-FL within 2 points of ideal FedAvg", max(gaps) <= 0.02,
201
+ f"max gap {max(gaps):.3f}"))
202
+ acc = dict(zip(f7["ratios"], f7["noniid"]["ota"]))
203
+ out.append(("Fig. 7(b): non-IID accuracy rises with R_in (data-limited regime)",
204
+ acc[0.9] > acc[0.3], f"{acc[0.3]:.3f} -> {acc[0.9]:.3f}"))
205
+ return out
@@ -0,0 +1,167 @@
1
+ Metadata-Version: 2.4
2
+ Name: aircompkit
3
+ Version: 1.0.0
4
+ Summary: Security-aware simulator for over-the-air computation and over-the-air federated learning.
5
+ Author: M. A. Moyeen, Kuljeet Kaur, Anjali Agarwal
6
+ License: MIT
7
+ Project-URL: Repository, https://github.com/moyeendev/AirCompKit
8
+ Project-URL: Documentation, https://moyeendev.github.io/AirCompKit/
9
+ Project-URL: Changelog, https://github.com/moyeendev/AirCompKit/blob/main/CHANGELOG.md
10
+ Keywords: over-the-air computation,AirComp,federated learning,wireless,6G,Byzantine robustness,physical-layer security,simulation
11
+ Classifier: Development Status :: 5 - Production/Stable
12
+ Classifier: Programming Language :: Python :: 3
13
+ Classifier: Programming Language :: Python :: 3.10
14
+ Classifier: Programming Language :: Python :: 3.11
15
+ Classifier: Programming Language :: Python :: 3.12
16
+ Classifier: Programming Language :: Python :: 3.13
17
+ Classifier: License :: OSI Approved :: MIT License
18
+ Classifier: Operating System :: OS Independent
19
+ Classifier: Intended Audience :: Science/Research
20
+ Classifier: Topic :: Scientific/Engineering
21
+ Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
22
+ Classifier: Topic :: Communications
23
+ Requires-Python: >=3.10
24
+ Description-Content-Type: text/markdown
25
+ License-File: LICENSE
26
+ Requires-Dist: numpy>=1.24
27
+ Requires-Dist: pyyaml>=6
28
+ Provides-Extra: torch
29
+ Requires-Dist: torch>=2.1; extra == "torch"
30
+ Provides-Extra: plot
31
+ Requires-Dist: matplotlib>=3.7; extra == "plot"
32
+ Provides-Extra: test
33
+ Requires-Dist: pytest>=7; extra == "test"
34
+ Provides-Extra: dev
35
+ Requires-Dist: pytest>=7; extra == "dev"
36
+ Requires-Dist: pytest-cov>=4; extra == "dev"
37
+ Requires-Dist: ruff>=0.5; extra == "dev"
38
+ Requires-Dist: mypy>=1.8; extra == "dev"
39
+ Requires-Dist: types-PyYAML; extra == "dev"
40
+ Requires-Dist: matplotlib>=3.7; extra == "dev"
41
+ Provides-Extra: docs
42
+ Requires-Dist: mkdocs-material>=9.5; extra == "docs"
43
+ Requires-Dist: mkdocstrings[python]>=0.25; extra == "docs"
44
+ Requires-Dist: mkdocs-jupyter>=0.24; extra == "docs"
45
+ Requires-Dist: nbclient>=0.10; extra == "docs"
46
+ Requires-Dist: ipykernel>=6; extra == "docs"
47
+ Requires-Dist: matplotlib>=3.7; extra == "docs"
48
+ Provides-Extra: all
49
+ Requires-Dist: aircompkit[plot,torch]; extra == "all"
50
+ Dynamic: license-file
51
+
52
+ # AirCompKit
53
+
54
+ [![CI](https://github.com/moyeendev/AirCompKit/actions/workflows/ci.yml/badge.svg)](https://github.com/moyeendev/AirCompKit/actions/workflows/ci.yml)
55
+ [![Reproductions](https://github.com/moyeendev/AirCompKit/actions/workflows/reproduce.yml/badge.svg)](https://github.com/moyeendev/AirCompKit/actions/workflows/reproduce.yml)
56
+ [![Docs](https://github.com/moyeendev/AirCompKit/actions/workflows/docs.yml/badge.svg)](https://moyeendev.github.io/AirCompKit/)
57
+ [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
58
+ [![Python 3.10-3.13](https://img.shields.io/badge/python-3.10--3.13-blue.svg)](pyproject.toml)
59
+
60
+ **Security-aware simulator for over-the-air computation (AirComp) and over-the-air federated learning (OTA-FL).**
61
+
62
+ > Status: **v1.0.0**. AirComp over AWGN, Rayleigh and Rician
63
+ > channels (multi-antenna server, spatial correlation, path loss/shadowing, imperfect CSI,
64
+ > phase errors), four power-control schemes, four receive beamformers, over-the-air
65
+ > federated learning (FedSGD/FedAvg, NumPy models, MNIST/Fashion-MNIST/synthetic data,
66
+ > gradient bank), a threat model with seven attacks, and defences from grouped over-the-air
67
+ > aggregation to robust digital rules, plus an interface for detection-based defences. Validated against closed forms and published results, re-checked
68
+ > weekly in CI (`aircompkit reproduce NAME --check`).
69
+ >
70
+ > **Documentation:** <https://moyeendev.github.io/AirCompKit/> (user guide, six executed
71
+ > tutorial notebooks in [docs/tutorials](docs/tutorials), API reference, reproductions,
72
+ > benchmarks). See also [FEATURES.md](FEATURES.md).
73
+
74
+ ## Security in three lines
75
+
76
+ ```python
77
+ from aircompkit.learning import FederatedConfig, train
78
+
79
+ threat = {"attack": {"name": "collude", "boost_db": 15}, "n_attackers": 6,
80
+ "knowledge": ["own_data", "own_csi", "honest_mean"]} # assumptions are explicit
81
+ res = train(scenario, FederatedConfig(dataset="mnist", clip=0.2, threat=threat,
82
+ defense={"name": "digital_robust", "rule": "median"}))
83
+ ```
84
+
85
+ `scenario` needs a multi-antenna channel and `symbols="complex"` (the array-level path); see
86
+ [examples/secure_otafl.py](examples/secure_otafl.py) for a full setup.
87
+
88
+ AirComp lets many devices transmit at once so that the wireless channel computes their sum.
89
+ Because signals superpose before the server sees them, per-client defences no longer apply.
90
+ AirCompKit models the **channel, the AirComp transceiver, the learning loop, the adversary and
91
+ the defence** as interchangeable components, and it ships a suite that reproduces published
92
+ AirComp results.
93
+
94
+ ## Install
95
+
96
+ From PyPI once v1.0.0 is released: `pip install aircompkit` (add `[plot]` for figures).
97
+ For development:
98
+
99
+ ```bash
100
+ git clone https://github.com/moyeendev/AirCompKit.git
101
+ cd AirCompKit
102
+ python -m pip install -e ".[dev]"
103
+ pytest
104
+ ```
105
+
106
+ ## Quickstart
107
+
108
+ ```python
109
+ import aircompkit as ac
110
+
111
+ sc = ac.Scenario(
112
+ devices={"name": "uniform_disc", "K": 20, "p_max": 1.0, "r_min": 10, "r_max": 100},
113
+ channel={"name": "rayleigh", "pathloss": {"exponent": 3.0, "ref_gain_db": -30}},
114
+ power_control="optimal", # truncated_inversion | full_power | channel_inversion
115
+ target="mean",
116
+ noise_var=1e-10,
117
+ seed=0,
118
+ )
119
+ res = sc.run(trials=2000, dim=32) # analytical + signal-level Monte Carlo
120
+ print(res.summary()["mse"]) # (mean, 95% CI half-width)
121
+ res.save("results/demo") # .npz arrays + .json with the full config
122
+ ```
123
+
124
+ The same scenario as YAML is in [configs/rayleigh_optimal.yaml](configs/rayleigh_optimal.yaml).
125
+ [examples/mse_vs_snr.py](examples/mse_vs_snr.py) compares the power-control schemes against
126
+ the closed-form MSE of truncated channel inversion.
127
+ [examples/secure_otafl.py](examples/secure_otafl.py) compares the defences under three attacks on MNIST
128
+ (about 45 minutes; `--quick` for a smoke test).
129
+
130
+ ## Command line
131
+
132
+ ```bash
133
+ aircompkit --version
134
+ aircompkit list # registered components per kind
135
+ aircompkit run configs/rayleigh_optimal.yaml --trials 5000 --out results/run1
136
+ aircompkit run configs/mimo_maxmin_csi.yaml --out results/mimo
137
+ aircompkit reproduce yang2020 --out results/yang2020 # Yang et al. TWC 2020, Figs. 4-5
138
+ aircompkit train configs/otafl_synthetic.yaml --out results/otafl
139
+ aircompkit reproduce zhu2020 --out results/zhu2020 # Zhu et al. TWC 2020, Figs. 4, 7(b)
140
+ aircompkit reproduce yang2020 --quick --check # compare with the paper; exit 1 on failure
141
+ aircompkit reproduce --all --quick --check # all reproductions (~1 min with MNIST)
142
+ aircompkit bench --out results/bench # performance benchmarks
143
+ python -m aircompkit list # same as above, if your system blocks pip-generated launchers
144
+ ```
145
+
146
+ MNIST and Fashion-MNIST are never downloaded automatically. Put `mnist.npz` (Keras layout)
147
+ or the four IDX files in a folder and pass `--data-root` or set `AIRCOMPKIT_DATA`.
148
+
149
+ ## Repository layout
150
+
151
+ | Path | Contents |
152
+ |------|----------|
153
+ | `src/aircompkit/core` | registry, seed tree, Scenario, Result |
154
+ | `src/aircompkit/{devices,channel,aircomp,receiver}` | physical layer |
155
+ | `src/aircompkit/{adversary,defense}` | threat model, attacks, defences |
156
+ | `src/aircompkit/learning` | OTA-FL loop, datasets, gradient bank |
157
+ | `src/aircompkit/{metrics,reproduce}`, `bench.py` | evaluation, paper reproductions with tolerance checks, benchmarks |
158
+ | `configs/`, `examples/` | scenarios, examples |
159
+ | `docs/` | MkDocs site; `docs/tutorials/*.ipynb` executed tutorials (`python docs/execute_notebooks.py`) |
160
+
161
+ ## Citing
162
+
163
+ See [CITATION.cff](CITATION.cff). A SoftwareX paper is in preparation.
164
+
165
+ ## License
166
+
167
+ MIT. See [LICENSE](LICENSE).
@@ -0,0 +1,39 @@
1
+ aircompkit/__init__.py,sha256=UbdqzYnos-LVyYw1PKzP23t0PeRDyF5QcKuUMuqrnks,675
2
+ aircompkit/__main__.py,sha256=oLKlibDWgeXWoSdLhuizKVa3OsgLm5kCUNjYyN3d39M,62
3
+ aircompkit/bench.py,sha256=ftn_ghwJYU0J3SEaQ9cKNxP8PzaDAVkUEI8kxzdlaeU,6846
4
+ aircompkit/cli.py,sha256=AxgWgFNsMsKSjgB9iDKaX1hY8BlsXntAA7VWrN6CR0A,6717
5
+ aircompkit/adversary/__init__.py,sha256=h-CLKhFw_5SNcEVE7tTFZ0Bca1O5ckOajRMEnjbyw0I,6244
6
+ aircompkit/aircomp/__init__.py,sha256=gPpOT98hIzPUX6cSiW3ybjOWNbswDL2jF4FE9dPyFdo,828
7
+ aircompkit/aircomp/analytic.py,sha256=OEuv70l6FJZn43fufvXZj1NahycFHLwG-CH8TGxE_g0,1492
8
+ aircompkit/aircomp/power_control.py,sha256=eE3EybRWw-ev0RCoU4x1wb2EqqT1zxwhWbpK9A5S2V8,6966
9
+ aircompkit/aircomp/preprocess.py,sha256=vH2YfteZM2PK8gbcbP_UDh4Lo7Sem0R7WSC1JrrVas8,2004
10
+ aircompkit/channel/__init__.py,sha256=Gt694Sr78qeIdnBGU2dWem-2GsC853EeuvN2zWIk8Zk,7530
11
+ aircompkit/core/__init__.py,sha256=v7m95EVkCpNM99NYx3JZgGc7S8CwHuI5Xi_9Kf-6ejQ,93
12
+ aircompkit/core/registry.py,sha256=Xg5NuoMxbyL9OWs4Z4kkLlaXgznkFzqsVFrvNLZiO2Y,3682
13
+ aircompkit/core/result.py,sha256=PnXLsgXVbLB2Tm0eGzFOEAcCiOuZjf8STNhkWLISAVg,1723
14
+ aircompkit/core/rng.py,sha256=MRVl3pcy8-H6dt6sQejlubmvrVNf_ejObKvMQR6cMMI,1836
15
+ aircompkit/core/scenario.py,sha256=DXzNCNOnmlMpuXxFyJ5sGdz9-JFeIbZ2a-0K5eBde7c,9327
16
+ aircompkit/defense/__init__.py,sha256=uroC0Burcp8x79v76YWQQa5fefGdVLnmjJhQkhJsIpY,5537
17
+ aircompkit/defense/evaluate.py,sha256=hH3vMSBubZqolA4TzSDKLJqfQpbdmKVCkiWF_UE4H-8,2222
18
+ aircompkit/devices/__init__.py,sha256=U0Gyjqp-mjqD28jzriV6rg6qJZaUPFr3lNwnxQbj0ps,2643
19
+ aircompkit/learning/__init__.py,sha256=ZS1VgWJEw_zQ9vYo0khJaBvTpvfdk6QXuHg_EztQfUc,974
20
+ aircompkit/learning/aggregate.py,sha256=VV1K-Jqsv-EoEj0tlpd1bHCKBAZsWfz9H0j9LRscJRU,1972
21
+ aircompkit/learning/array.py,sha256=7Zms_sWtQ93ld4NfzqHdmBdMd4EWiDEnXdDUMF4ODEA,6477
22
+ aircompkit/learning/bank.py,sha256=kNalm3PCBvLGaW0PNy8xpX1jD41M5lHFMa-yKxrKLwE,3773
23
+ aircompkit/learning/data.py,sha256=yNqrksCKKRyil8cGyK0D08P-N0_d-x8BFQ_SgN107V0,4505
24
+ aircompkit/learning/fl.py,sha256=l3fE1aCn1o1tfcQNmEgRa3nNXvz5eDGYxURNBYidsgE,10582
25
+ aircompkit/learning/models.py,sha256=Rd_jh7B7FYpKotX3Uv-JEMfy-I8WGFdX1WzzHHCHBUQ,3282
26
+ aircompkit/learning/partition.py,sha256=EbodoMtrLCR9Vgf20c6wuOSaAFErvhpr459pqWjsRQ4,2128
27
+ aircompkit/metrics/__init__.py,sha256=wz974QJVRGTE0BRS8rSJJLgJe9n9J5DIAhY5aM-hw-U,1421
28
+ aircompkit/metrics/privacy.py,sha256=dE2I9P-htdt_c800Q9_zqhh_GUuNF17pvtjvaQe0mxw,1787
29
+ aircompkit/plot/__init__.py,sha256=cIRXdGfeaOu5VNkczmPIXsgP-oh8HE399_-IWCIwkjw,148
30
+ aircompkit/receiver/__init__.py,sha256=ciw_atwuuADiz9ZrkUrg9X7MM5g4fkJPeyWxwWVJ0c8,10522
31
+ aircompkit/reproduce/__init__.py,sha256=va-1Y-iRJAbhD-igiBrnZPnT3QJdB2QO6nuK1TfvYlU,1047
32
+ aircompkit/reproduce/yang2020.py,sha256=PZQ8w9k8N51drP0xds4i8eXUACzvSNf92_JkSwpffCs,8020
33
+ aircompkit/reproduce/zhu2020.py,sha256=MP4r0mXQu-lcQ-Gv1zAYnlQP6Lo9FSKHtQPQaYHA5FM,10261
34
+ aircompkit-1.0.0.dist-info/licenses/LICENSE,sha256=Io7Ug5Vk-OkW8qR1Y4pCenRP7phbqR_CUE8f4KprR7o,1066
35
+ aircompkit-1.0.0.dist-info/METADATA,sha256=uBFQrMf5zeI1pk-TsEQnOkx5EjOrAQQtt6by9XZW41o,7838
36
+ aircompkit-1.0.0.dist-info/WHEEL,sha256=mOdXR9C3wMVov7cSY8PY-XCzo3SmmWCHvGQr5csJoao,91
37
+ aircompkit-1.0.0.dist-info/entry_points.txt,sha256=icpnYhXDR6nEAlulAJPs82aOxR7Cjh_a8NbeSAbw-NQ,51
38
+ aircompkit-1.0.0.dist-info/top_level.txt,sha256=-QdTTuOc1HjtHGV5GrrBlcxhv33QAK7tAI4Xw4xoD6c,11
39
+ aircompkit-1.0.0.dist-info/RECORD,,
@@ -0,0 +1,5 @@
1
+ Wheel-Version: 1.0
2
+ Generator: setuptools (84.2.1)
3
+ Root-Is-Purelib: true
4
+ Tag: py3-none-any
5
+
@@ -0,0 +1,2 @@
1
+ [console_scripts]
2
+ aircompkit = aircompkit.cli:main
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2026 moyeendev
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
@@ -0,0 +1 @@
1
+ aircompkit