Multi-Objective Learning for Joint Optimization of Inference Latency, Energy Consumption, and Model Accuracy
-
DOI:
https://doi.org/10.67228/30715717/IJDEIC-V9I2P102Published 04-08-2026
Multi-Objective Learning, Inference Latency, Energy Consumption, Model Accuracy, Pareto Optimization, Hardware-Aware Neural Architecture Search, Edge Artificial Intelligence, Model Compression, Resource-Constrained Inference Issue
Section
ArticlesAbstract
Widespread adoption of artificial intelligence in edge computing, mobile devices, autonomous systems, and IoT devices has placed a growing importance on efficient inference models. Common optimization techniques focus mainly on the predictive accuracy or computational complexity, with little attention to the influence of inference latency, energy consumption, or model quality. This paper suggests a multi-objective learning framework for optimizing them, all competing dimensions of performance, together. The proposed framework combines hardware-aware profiling characteristics, model compression, adaptive loss weighting, Pareto-based candidate selection, and deployment-aware validation. The normalized objective function includes a trade-off between latency, energy and classification error and it keeps application-specific resource constraints. In the optimization approach, non-dominated model configurations are identified, instead of just a single weighted configuration. The framework has been designed to enable deployment in a heterogeneous computing environment. An illustrative evaluation protocol is used to illustrate how percentage-based comparisons can quantify trade-offs among competing strategies. This methodology creates a repeatable framework for deploying machine learning models while maintaining accuracy, minimizing latency, and reducing energy consumption.
References
[1] S. Ruder, “An overview of gradient descent optimization algorithms,” arXiv preprint arXiv:1609.04747, 2016.
[2] Syed, Y. (2023). Polyglot Persistence and Enterprise Data Migration: Architectures, Consistency, and Practice. American International Journal of Computer Science and Technology, 5(3), 61-66. https://doi.org/10.63282/3117-5481/AIJCST-V5I3P106
[3] D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” in Proc. Int. Conf. Learn. Representations (ICLR), 2015.
[4] Fung, W., Guo, L., Hilliard, D., Casadei, A., Ratan, R., Bhaduri, S., ... & Rakshit, A. (2025). OpComm: A Reinforcement Learning Framework for Adaptive Buffer Control in Warehouse Volume Forecasting. arXiv preprint arXiv:2512.19738.
[5] Syed, Y. (2025). Generative AI Integration and RAG-Based Enterprise Tooling: Architectures, Techniques, and Organizational Impact. International Journal of Scientific Research in Science, Engineering and Technology, 12(6), 600-606. DOI: https://doi.org/10.32628/IJSRSET2613255
[6] S. Han, H. Mao, and W. J. Dally, “Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding,” in Proc. Int. Conf. Learn. Representations (ICLR), 2016.
[7] Nangi, P. R., Obannagari, C. K. R. N., & Settipi, S. (2022). Enhanced Serverless Micro-Reactivity Model for High-Velocity Event Streams within Scalable Cloud-Native Architectures. International Journal of Emerging Research in Engineering and Technology, 3(3), 127-135.
[8] Syed, Y. (2024). DevSecOps and Enterprise-Scale CI/CD Pipeline Engineering: Governing Delivery Velocity, Embedded Security, and Release Automation Across Polyglot Environments. International Journal of Scientific Research in Science, Engineering and Technology, 11(5), 475-484. DOI: https://doi.org/10.32628/IJSRSET2513156
[9] Bhaduri, S., Mullapudi, P., Dietrich, S., DeWaters, S., Ratan, R., Kashyap, B., ... & Rakshit, A. (2025). Beyond mere automation: A techno-functional framework for reimagining gen-AI in supply chain operations.
[10] G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” arXiv preprint arXiv:1503.02531, 2015.
[11] Akinapalli, S. (2025). Metadata-driven data integration framework: Automating enterprise data integration through declarative approaches. European Modern Studies Journal, 9(4), 9. http://www.journal-ems.com
[12] Pavan Nithin Mullapudi (2024). An Adaptive Explainability Framework for Machine Learning Predictions of Deals in Cloud Computing . International Journal of Multidisciplinary Research and Growth Evaluation (IJMRGE), 5(3), 1027-1034. DOI: https://doi.org/10.54660/.IJMRGE.2024.5.3.1027-1034
[13] A. Polino, R. Pascanu, and D. Alistarh, “Model compression via distillation and quantization,” in Proc. Int. Conf. Learn. Representations (ICLR), 2018.
[14] Obannagari, C. K. R. N., & Nangi, P. R. (2021). Predictive Autoscaling for Kubernetes Using Multivariate Time-Series Forecasting. International Journal of Emerging Research in Engineering and Technology, 2(2), 73-83.
[15] K. Deb, A. Pratap, S. Agarwal, and T. Meyarivan, “A fast and elitist multiobjective genetic algorithm: NSGA-II,” IEEE Trans. Evol. Comput., vol. 6, no. 2, pp. 182–197, Apr. 2002.
[16] Nangi, P. R., & Obannagari, R. N. CK, & Settipi, S.(2022). Predictive SQL Query Tuning Using Sequence Modeling of Query Plans for Performance Optimization. International Journal of AI, BigData, Computational and Management Studies, 3(2), 104-113.
[17] M. Tan et al., “MnasNet: Platform-aware neural architecture search for mobile,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2019, pp. 2820–2828.
[18] T.-J. Yang, A. Howard, B. Chen, X. Zhang, A. Go, M. Sandler, V. Sze, and H. Adam, “NetAdapt: Platform-aware neural network adaptation for mobile applications,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2018, pp. 285–300.
[19] H. Cai, C. Gan, T. Wang, Z. Zhang, and S. Han, “Once-for-All: Train one network and specialize it for efficient deployment,” in Proc. Int. Conf. Learn. Representations (ICLR), 2020.
[20] Nangi, P. R., & Settipi, S. (2023). A cloud-native serverless architecture for event-driven, low-latency, and AI-enabled distributed systems. International Journal of Emerging Research in Engineering and Technology, 4(4), 128-136.
[21] P. Ren et al., “A comprehensive survey of neural architecture search: Challenges and solutions,” ACM Comput. Surv., vol. 54, no. 4, Art. no. 76, 2021, doi: 10.1145/3447582.
[22] M. Tan and Q. V. Le, “EfficientNet: Rethinking model scaling for convolutional neural networks,” in Proc. Int. Conf. Mach. Learn. (ICML), 2019, pp. 6105–6114.
[23] Nangi, P. R., Obannagari, C. K. R. N., & Settipi, S. (2022). Self-auditing deep learning pipelines for automated compliance validation with explainability, traceability, and regulatory assurance. International Journal of Artificial Intelligence, Data Science, and Machine Learning, 3(1), 133-142.
[24] H. Cai, L. Zhu, and S. Han, “ProxylessNAS: Direct neural architecture search on target task and hardware,” in Proc. Int. Conf. Learn. Representations (ICLR), 2019.
[25] C. H. Lin, C. Y. Cheng, Y. C. Chen, H. C. Lee, and D. C. Juan, “MCUNet: Tiny deep learning on IoT devices,” in Advances in Neural Information Processing Systems (NeurIPS), vol. 33, 2020, pp. 11711–11722.
[26] Nangi, P. R., & Obannagari, C. K. R. N. (2024). High-Performance Distributed Database Partitioning Using Machine Learning-Driven Workload Forecasting and Query Optimization. American International Journal of Computer Science and Technology, 6(2), 11-21.
[27] T.-J. Yang, Y.-L. Liao, and V. Sze, “NetAdaptV2: Efficient neural architecture search with fast super-network training and architecture optimization,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2021, pp. 2402–2411.
[28] Nangi, P. R. (2022). Multi-Cloud Resource Stability Forecasting Using Temporal Fusion Transformers. International Journal of Artificial Intelligence, Data Science, and Machine Learning, 3(3), 123-135.
[29] X. Lu, W. Dong, X. Li, J. Wu, L. Li, and G. Shi, “Adaptive search-and-training for robust and efficient network pruning,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 45, no. 8, pp. 9325–9338, Aug. 2023, doi: 10.1109/TPAMI.2023.3248612
[30] S. Suganya, Dr K. Govind, "Relationship Between Social Media Addiction and Self-Esteem among Working and Non Working Women," Purakala, vol. 31, no. 5, pp. 29-33, 2020.
[31] Reddy Nangi, P., & Reddy Nala Obannagari, C. K. (2023). Scalable End-to-End Encryption Management Using Quantum-Resistant Cryptographic Protocols for Cloud-Native Microservices Ecosystems. International Journal of Emerging Trends in Computer Science and Information Technology, 4(1), 142-153.
[32] Virendra Jangid. (2025). AI-Powered Cross-Border Securities Processing and ADR Platform Optimization. International Journal of Leading Research Publication, 6(11). https://doi.org/10.70528/IJLRP.v6.i11.2233
[33] Nangi, P. R., Obannagari, C. K. R. N., & Settipi, S. (2025). Large Language Model–Driven Intelligent Observability Frameworks for Serverless Applications in Event-Driven Cloud Architectures. International Journal of Emerging Research in Engineering and Technology, 6(1), 100-110.
[34] Y. Syed, “Enterprise Observability and Software Delivery Performance: From Telemetry to Engineering Intelligence”, IJADSMC, vol. 5, no. 2, pp. 01–08, Sep. 2022, doi: 10.67228/30713498/IJADSMC-2022PIIKP2M.
[35] Gantikota, S. (2023). Reducing HL7 Processing Errors through Automated File Creation and Ingestion Pipelines: A Production Case Study in EHR Data Integration. International Journal of Emerging Trends in Computer Science and Information Technology, 4(4), 241-245. https://doi.org/10.63282/3050-9246.IJETCSIT-V4I4P125
[36] Nangi, P. R., & Obannagari, C. K. R. N. (2026). Federated ai-driven query optimization for distributed cloud databases. International Journal of Artificial Intelligence, Data Science, and Machine Learning, 7(1), 91-99.
[37] K. Bhagyasri, Education vs Entertainment on the mobile screen time spent per day by students in Higher Education in Tamil Nadu, Salud, Ciencia y Tecnologia - Serie de Conferencias, vol. 3, no. 714, pp. 1-11, 2024.
Downloads