Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces APEX, a network-native, decoder-only transformer specifically designed for forecasting and anomaly detection in wireless edge operations, addressing the unique challenges posed by bursty and zero-inflated telemetry signals. Evaluated on a DHCP degradation benchmark, APEX-Large outperforms existing models, achieving an 18% reduction in mean absolute error compared to the best foundation model and a 38% improvement over SARIMA, while maintaining a high anomaly-detection F1 score of 0.93. APEX-Edge further enables efficient, privacy-preserving inference on edge hardware, demonstrating the effectiveness of network-native pre-training for proactive management of wireless networks.
APEX-Large outperforms traditional models by 18% in forecasting accuracy, showcasing the power of network-native pre-training for wireless telemetry.
Generic time-series foundation models transfer poorly to wireless network telemetry whose signals are bursty, zero-inflated, and coupled across protocol layers. We present APEX, a network-native, decoder-only transformer for forecasting enterprise AP telemetry, and evaluate it on DHCP degradation as a representative network task. APEX is pre-trained on 10-channel multivariate telemetry from ~4,500 production wireless networks (~100K AP time series, 34 metrics per AP), and is available as APEX-Large (269M, cloud) and APEX-Edge (10.5M, edge). On a 192-step (4-day) DHCP degradation benchmark, APEX-Large reduces MAE by 18% over the strongest foundation-model baseline (Toto) and 38% over SARIMA, with anomaly-detection F1 = 0.93, while APEX-Edge enables sub-second, privacy-preserving inference on AP-class edge hardware. These results suggest network-native pre-training is a practical foundation for proactive wireless operations.