EMMA-BEAM-v0.2¶

v0.2 · proposed-contract

Task. E-CH-BEAM - see Tasks. What you predict. the best beam index. Target domain. COMMS - 3GPP AI/ML beam management.

Input¶

Raw complex multi-antenna I/Q (in-phase and quadrature), shape (num_rx, 2, N) float32, with per-scene metadata. 32- or 64-element massive MIMO (multiple-input multiple-output) array.

Labels¶

Best-beam index.

Splits¶

train and dev are public and regenerable; holdout is private and scored by prediction submission. Sample count to be pinned at v0.2.

Get it¶

emma download --dataset EMMA-BEAM-v0.2 --split dev --out ./scenes

Metric¶

TopKAccuracy (top-k categorical accuracy over the beam codebook). See Metrics.

What this tests¶

Beam management predicts the best beam from the codebook under an unseen scenario and array, the 3GPP AI/ML use case that gates millimeter-wave link recovery.

How it is made¶

Generated by rfgen. The exact configuration (commit, emitters, channel, label extraction, seeds) is in the developer recipe.