3 commits

Author	SHA1	Message	Date
Graham McIntire	69b5caf876	Normalize ML features to prevent NaN gradient explosion ... Raw features had vastly different scales (pressure ~1013, sin/cos ~[-1,1]) causing gradient explosion. Normalize all atmospheric features to ~[0,1] using known physical bounds. Add Polaris dep for optimizer.	2026-04-01 09:08:33 -05:00
Graham McIntire	c12f8cf5ed	Use local solar time for time-of-day scoring, add PWAT factor and pressure refinements ... Score time-of-day per grid point using longitude/15 solar offset instead of hardcoded CST/CDT. Add PWAT as 10th scoring factor. Refine pressure thresholds. Update ML model and training pipeline to use local solar time.	2026-04-01 08:58:21 -05:00
Graham McIntire	8949920b7f	Add Nx/Axon/EXLA ML model skeleton for propagation prediction ... 13-feature feed-forward network (atmospheric + temporal + frequency). Includes build, init, predict, encode_features, save/load to disk. Model weights saved to priv/models/propagation_v1.nx (gitignored). Not yet trained — scaffolding only.	2026-03-31 16:26:34 -05:00