Research Article

Adaptive Accuracy and Cost Control under Distribution Shift in Model Cascades

Authors

Renas Rajab Asaad * Department of Information Technology, Technical College of Informatics–Akre, Akre University for Applied Sciences, Duhok, Iraq.
Viktor P. Sokolov Department of Computer Science, Kazan Federal University, Kazan, Russia.

https://doi.org/10.48314/ceti.v3i3.92

Downloads

Abstract

Streaming prediction under distribution shift requires decisions about inference cost and model refresh. We examine ABACUS, a confidence-gated cascade with drift-triggered refitting and feedback on average proxy cost, and outline a separate conditional design for admission under per-item deadlines. The supplied study reports ten seeded simulations of 10,000 items each on rotating-hyperplane and Streaming Ensemble Algorithm (SEA) generators with two scheduled abrupt changes, quality perturbations, and a 200-item label delay. At nominal budget B=3, reported mean accuracies are 0.9148 and 0.9397 at assigned mean costs 2.99 and 2.94, compared with 0.6775 and 0.8987 at costs 3.00 and 2.85 for a learned static cascade. Reported areas under accuracy-versus-nominal-budget curves are 0.853 and 0.916 versus 0.651 and 0.883. The ablations attribute the observed gain at this operating point to drift-triggered refitting; budget feedback mainly tracks the target mean cost. These are author-reported simulation results: the named experiment script, shared generator, seed-level results, and raw JavaScript Object Notation (JSON) file were not furnished with the materials reviewed. The assigned cost units are not measured wall-clock latency. The experiment did not exercise worst-case stage admission, preemption, pool-weight updates, projected-gradient threshold control, or an input-channel detector. It therefore establishes neither a hard per-item deadline nor the comparative cost of enforcing one. Independent reproduction and a matched, refitted single-model baseline are required before a general performance claim.

Keywords

Distribution shift Model cascade Drift-triggered refitting Average compute cost Conditional deadline admission Synthetic stream

References (38)

[1] Babu, S., & Widom, J. (2001). Continuous queries over data streams. ACM Sigmod Record, 30(3), 109–120. https://doi.org/10.1145/603867.603884

[2] Akidau, T., Balikov, A., Bekiroglu, K., Chernyak, S., Haberman, J., Lax, R., Mills, D., & Paul Nordstrom, S. W. (2013). MillWheel: Fault-tolerant stream processing at internet scale. In procceding 6th biennial conf (pp. 1033–1044). Innovative Data Systems Research (CIDR). https://www.vldb.org/pvldb/vol6/p1033-akidau.pdf

[3] Akidau, T., Bradshaw, R., Chambers, C., Chernyak, S., Fernández Moctezuma, R. J., & Lax, R. (2015). The dataflow model: A practical approach to balancing correctness, latency, and cost in massive-scale, unbounded, out-of-order data processing. Proceedings of the VLDB Endowment, 8(12), 1792–1803. https://www.vldb.org/pvldb/vol8/p1792-Akidau.pdf %28Google

[4] Carbone, P., Katsifodimos, A., Ewen, S., Markl, V., Haridi, S., & Tzoumas, K. (2015). Apache flink^TM: Stream and batch processing in a single engine. IEEE., 38(4), 28–38. https://research.tudelft.nl/en/publications/apache-flink-stream-and-batch-processing-in-a-single-engine/

[5] Tatbul, N., Çetintemel, U., Zdonik, S., Cherniack, M., & Stonebraker, M. (2003). Load shedding in a data stream manager. Proceedings 2003 VLDB conference (pp. 309–320). Elsevier. https://doi.org/10.1016/B978-012722442-8/50035-5

[6] Dean, J., & Barroso, L. A. (2013). The tail at scale. Communications of the ACM, 56(2), 74–80. http://doi.acm.org/10.1145/2408776.2408794

[7] Viola, P., & Jones, M. J. (2004). Robust real-time face detection. International Journal of Computer Vision, 57(2), 137–154. https://doi.org/10.1023/B:VISI.0000013087.49260.fb%0A%0A

[8] Teerapittayanon, S., McDanel, B., & Kung, H. T. (2016). Branchynet: Fast inference via early exiting from deep neural networks. 2016 23rd international conference on pattern recognition (ICPR) (pp. 2464–2469). IEEE. https://doi.org/10.48550/arXiv.1709.01686

[9] Huang, G., Chen, D., Li, T., Wu, F., Van Der Maaten, L., & Weinberger, K. Q. (2018). Multi-scale dense networks for resource efficient image classification, 1, 1–14. https://doi.org/10.48550/arXiv.1703.09844

[10] Agarwal, S., Mozafari, B., Panda, A., Milner, H., Madden, S., & Stoica, I. (2013). BlinkDB: Queries with bounded errors and bounded response times on very large data. Proceedings of the 8th ACM european conference on computer systems (pp. 29–42). ACM. https://doi.org/10.1145/2465351.2465355

[11] Gama, J., Žliobaite, I. ., Bifet, A., Pechenizkiy, M., & Bouchachia, A. (2014). A survey on concept drift adaptation. ACM Computing Surveys (CSUR), 46(4), 1–37. https://doi.org/10.1145/2523813

[12] Lu, J., Liu, A., Dong, F., Gu, F., Gama, J., & Zhang, G. (2018). Learning under concept drift: A review. IEEE Transactions on Knowledge and Data Engineering, 31(12), 2346–2363. https://doi.org/10.1109/TKDE.2018.2876857

[13] Quiñonero Candela, J., Sugiyama, M., Schwaighofer, A., & Lawrence, N. D. (2008). Dataset shift in machine learning. Mit Press. https://www.researchgate.net/publication/234817678

[14] Sugiyama, M., & Kawanabe, M. (2012). Machine learning in non-stationary environments: Introduction to covariate shift adaptation. MIT press. https://doi.org/10.7551/mitpress/9780262017091.001.0001

[15] Pan, S. J., & Yang, Q. (2009). A survey on transfer learning. IEEE Transactions on Knowledge and Data Engineering, 22(10), 1345–1359. https://doi.org/10.1109/TKDE.2009.191

[16] Domingos, P., & Hulten, G. (2000). Mining high-speed data streams. Proceedings of the sixth ACM sigkdd international conference on knowledge discovery and data mining (pp. 71–80). ACM. https://doi.org/10.1145/347090.347107

[17] Hulten, G., Spencer, L., & Domingos, P. (2001). Mining time-changing data streams. Proceedings of the seventh ACM sigkdd international conference on knowledge discovery and data mining (pp. 97–106). ACM. https://doi.org/10.1145/502512.502529

[18] Street, W. N., & Kim, Y. (2001). A streaming ensemble algorithm (SEA) for large-scale classification. Proceedings of the seventh ACM sigkdd international conference on knowledge discovery and data mining (pp. 377–382). ACM. https://doi.org/10.1145/502512.502568

[19] Oza, N. C., & Russell, S. J. (2001). Online bagging and boosting. International workshop on artificial intelligence and statistics (pp. 229–236). Proceedings of Machine Learning Research (PMLR). https://proceedings.mlr.press/r3/oza01a.html

[20] Kotler, J., & Maloof, M. (2003). Dynamic weighted majority: A new ensemble method for tracking concept drift. IEEE international conference on data mining (pp. 123–130). IEEE Computer Society, Washington, DC, USA. https://dl.acm.org/doi/pdf/10.5555/1314498.1390333

[21] Elwell, R., & Polikar, R. (2011). Incremental learning of concept drift in nonstationary environments. IEEE Transactions on Neural Networks, 22(10), 1517–1531. https://doi.org/10.1109/TNN.2011.2160459

[22] Page, E. S. (1954). Continuous inspection schemes. Biometrika, 41(1/2), 100–115. https://doi.org/10.2307/2333009

[23] Gama, J., Medas, P., Castillo, G., & Rodrigues, P. (2004). Learning with drift detection. Brazilian Symposium on Artificial Intelligence (pp. 286–295). Springer. https://doi.org/10.1007/978-3-540-28645-5_29%0A%0A

[24] Baena Garcia, M., del Campo Ávila, J., Fidalgo, R., Bifet, A., Gavalda, R., & Morales Bueno, R. (2006). Early drift detection method. Fourth international workshop on knowledge discovery from data streams (Vol. 6, pp. 77–86). ACM. https://www.researchgate.net/publication/245999704

[25] Bifet, A., & Gavalda, R. (2007). Learning from time-changing data with adaptive windowing. Proceedings of the 2007 SIAM international conference on data mining (pp. 443–448). SIAM. https://doi.org/10.1137/1.9781611972771.42

[26] Bifet, A., Holmes, G., Pfahringer, B., Kranen, P., Kremer, H., Jansen, T., & Seidl, T. (2010). Moa: Massive online analysis, a framework for stream classification and clustering. Proceedings of the first workshop on applications of pattern analysis (pp. 44–50). PMLR. https://www.researchgate.net/publication/220320178

[27] Zilberstein, S. (1996). Using anytime algorithms in intelligent systems. AI Magazine, 17(3), 73. https://doi.org/10.1609/aimag.v17i3.1232

[28] Hellerstein, J. M., Haas, P. J., & Wang, H. J. (1997). Online aggregation. Proceedings of the 1997 ACM sigmod international conference on management of data (pp. 171–182). ACM. https://doi.org/10.1145/253260.253291

[29] Crankshaw, D., Wang, X., Zhou, G., Franklin, M. J., Gonzalez, J. E., & Stoica, I. (2017). Clipper: A low-latency online prediction serving system. 14th usenix symposium on networked systems design and implementation (NSDI 17) (pp. 613–627). USENIX Association. https://www.usenix.org/conference/nsdi17/technical-sessions/presentation/crankshaw

[30] Littlestone, N., & Warmuth, M. K. (1994). The weighted majority algorithm. Information and Computation, 108(2), 212–261. https://doi.org/10.1006/inco.1994.1009

[31] Cesa-Bianchi, N., & Lugosi, G. (2006). Prediction, learning, and games. Cambridge University Press. https://www.researchgate.net/profile/Gabor-Lugosi/publication/220690817

[32] Bubeck, S., & Cesa Bianchi, N. (2012). Regret analysis of stochastic and nonstochastic multi-armed bandit problems. Foundations and Trends^® In Machine Learning, 5(1), 1–122. https://doi.org/10.1561/2200000024

[33] Shalev Shwartz, S. (2025). Online learning and online convex optimization. Foundations and Trends®In Machine Learning, 4(2), 107–194. https://doi.org/10.1561/2200000018

[34] Breiman, L. (2001). Random forests. Machine learning, 45(1), 5–32. https://doi.org/10.1023/A:1010933404324%0A%0A

[35] Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., & Duchesnay, É. (2011). Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12, 2825–2830. https://jmlr.org/papers/volume12/pedregosa11a/pedregosa11a.pdf

[36] Wilcoxon, F. (1945). Individual comparisons by ranking methods. Biometrics Bulletin, 1(6), 80–83. https://doi.org/10.2307/3001968

[37] Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics, 65–70. https://doi.org/10.2307/4615733

[38] Demšar, J. (2006). Statistical comparisons of classifiers over multiple data sets. Journal of Machine Learning Research, 7(Jan), 1–30. https://www.jmlr.org/papers/volume7/demsar06a/demsar06a.pdf

Article Metrics

Published 2026-08-30
Issue Vol. 3 No. 3 (2026): Computational Engineering and Technology Innovations
Section Articles
Document Type Research Article
Views 0
Downloads 0

How to Cite

Rajab Asaad, R., & P. Sokolov, V. (2026). Adaptive Accuracy and Cost Control under Distribution Shift in Model Cascades. Computational Engineering and Technology Innovations, 3(3), 193-207. https://doi.org/10.48314/ceti.v3i3.92

Download Citation

Similar Articles

Madasu Hanmandlu, Kharabela Swain, Wolfgang H. Müller, Mechanism Informed Early Warning of Simulated Dynamical Transitions , Computational Engineering and Technology Innovations: Vol. 3 No. 3 (2026): Computational Engineering and Technology Innovations
Maedeh Fasihi, Azadeh Shemshad, A Taxonomy and Evidence Framework for Hybrid Metaheuristics in Smart Manufacturing and Supply Chains , Computational Engineering and Technology Innovations: Vol. 3 No. 1 (2026): Computational Engineering and Technology Innovations
Mahmoud Alimoradi, Eisa Abdolmaleki, Hashem Saberi Najafi, Hadi Faghihmaleki, Latency Aware Metaheuristics for Earthquake Early Warning Network Design , Computational Engineering and Technology Innovations: Vol. 3 No. 3 (2026): Computational Engineering and Technology Innovations
Omar Mar Cornelio, Mingyue Wang, Rania A. Mahmoud, Exact Predictive Ventilation Control with Causal Occupancy Forecasts and Concentration Guarantees , Computational Engineering and Technology Innovations: Vol. 3 No. 2 (2026): Computational Engineering and Technology Innovations
Victoria Nozick, Fatemeh Rasoulpour, Milos Milovancevic, View Coverage and Evidence Gating for Vision-Based Robotic Quality Inspection , Computational Engineering and Technology Innovations: Vol. 3 No. 3 (2026): Computational Engineering and Technology Innovations
Zeinab Parandavar, Arman Daliri, Occupancy Inference from Aggregate Smart-Meter Data Across Sampling Resolutions: A Controlled Synthetic Privacy Study , Computational Engineering and Technology Innovations: Vol. 3 No. 3 (2026): Computational Engineering and Technology Innovations
Kharabela Swain, Marzieh Lamtar Gholipoor, Sogol Motallebi, Equilibrium Preservation and Finite Memory Bias in Fractional Monod Chemostat Models , Computational Engineering and Technology Innovations: Vol. 3 No. 3 (2026): Computational Engineering and Technology Innovations
Habibeh Karimi, Khatereh Bagherzadeh Asl, Samira Torabi, A Geometry-Gated Coordination of Reconfigurable Intelligent Surfaces for Maritime Surveillance Communications , Computational Engineering and Technology Innovations: Vol. 3 No. 2 (2026): Computational Engineering and Technology Innovations
Mahmoud Alimoradi, Setayesh Zahiri, Maryam Khani, RaDOD-Net: A Data-Contracted CNN–LSTM–Attention Architecture for Radar Enhancement and Airport Obstacle Detection , Computational Engineering and Technology Innovations: Vol. 3 No. 1 (2026): Computational Engineering and Technology Innovations
Aniekan Essienubong Ikpe, Imoh Ime Ekanem, Jephtar Uviefovwe Ohwoekevwo, Computational Modelling of Complex Physical Phenomena in TIG Weld Profile based on relationship between Coefficient of Thermal Expansion and Transient Temperature Variations in Mild Steel Plate , Computational Engineering and Technology Innovations: Vol. 1 No. 4 (2024): Computational Engineering and Technology Innovations

1-10 of 52 Next

You may also start an advanced similarity search for this article.

Most read articles by the same author(s)

Renas Rajab Asaad, Madasu Hanmandlu, Kharabela Swain, AISR: A Protocol-Closed Architecture for Interference-Aware Channel-Mask Control in Industrial LoRaWAN , Computational Engineering and Technology Innovations: Vol. 3 No. 2 (2026): Computational Engineering and Technology Innovations