APEX: A Network-Native Time-Series Foundation Model for Forecasting and Anomaly Detection for Wireless Edge Operations

In three linesAPEX is a transformer foundation model specialized for wireless network telemetry. Pre-trained on 10-channel multivariate data from ~4,500 production networks (100K time series), it reduces MAE by 18% vs Toto and 38% vs SARIMA on 192-step DHCP forecasting. Two versions: APEX-Large (269M, cloud) and APEX-Edge (10.5M, edge) with F1=0.93 for anomaly detection.

Summary generated by Claude — human-verified