APEX: A Network-Native Time-Series Foundation Model for Forecasting and Anomaly Detection for Wireless Edge Operations
AuthorsSwadhin Pradhan, Niloo Bahadori, Peiman Amini
Resources
APEX adapts foundation models to wireless network telemetry, showing that a smaller network-native transformer can forecast AP behavior and detect anomalies better than generic time-series baselines.
Key results
telemetry corpus used for pretraining
scale of AP telemetry in the training corpus
forecasting and evaluation window in steps
large network-native model size
edge-deployable model size
APEX-Large multivariate forecasting error on the DHCP benchmark
What the paper found
APEX is a Cisco Systems network-native time-series foundation model built specifically for wireless edge operations, where telemetry is bursty, zero-inflated, and tightly coupled across protocol layers, unlike the public corpora used by generic models. The paper trains a decoder-only patched transformer on co-collected multivariate telemetry from about 4,500 production wireless networks, covering roughly 100K AP time series with 34 metrics per AP, and uses a 10-channel causal-chain input to predict DHCP degradation. On a 192-step, 4-day benchmark, APEX-Large with 269M parameters cuts forecasting MAE to 2.98, beating Toto’s 3.64 by 18% and SARIMA’s 4.82 by 38%, while the 10.5M-parameter APEX-Edge retains strong performance with MAE 3.87. For anomaly detection, a single checkpoint with MC-dropout reaches F1 = 0.93, close to VAR-Mahalanobis at 0.94 and above Toto’s 0.85, showing that calibrated uncertainty can replace a separate detector. The edge model is designed for AP-class hardware: APEX-Edge runs in 202 ms on a Raspberry Pi 5 proxy, uses only 10.5M parameters, and keeps raw telemetry on-device, reducing uplink traffic from about 130 MB/day to compact alerts of about KB/day. The core novelty is that network-native pretraining, not just transformer architecture, closes the transfer gap for enterprise wireless forecasting and anomaly detection.
Original abstract
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.
Read the original paperMore in Foundation Models
Browse all 47 papers →How Much Is an AI Token Worth? Scaling Laws for Wild AI-Generated Web Text
Jenna Russell, Ben Glickenhaus, Katherine Thai, John Wieting, Mohit Iyyer, Max Spero, Bradley Emi
AI-generated web text can help language models at first, but beyond a tipping point it degrades performance on human writing, making data filtering and separate evaluation increasingly important.
TabFM: A Zero-Shot Foundation Model for Tabular Data
Weihao Kong, Erez Louidor Ilan, Shuxin Nie, Taman Narayan, Rajat Sen, Yichen Zhou, Deqing Fu, Samet Oymak, Abhimanyu Das
TabFM is a large synthetic-data-trained model that aims to make accurate tabular predictions instantly, without retraining for each new dataset.
When Do Biological Reasoning Models Use Their Biological Inputs?
Ada Fang, Nikitha Thoduguli, Lukas Fesser, Hanlin Zhang, Sham M. Kakade, Marinka Zitnik
The study finds that many biological reasoning systems appear to succeed without meaningfully using the biological inputs they were designed to reason over.