Posts by Collection

publications

As-Encountered Prediction of Tunnel Boring Machine Performance Parameters using Recurrent Neural Networks

Published in Transportation Research Record, 2020

This paper explores the use of deep learning to predict excavation performance parameters during tunnel construction.

Recommended citation: Nagrecha, K., Fisher, L., Mooney, M., Rodriguez-Nikl, T., Mazari, M., & Pourhomayoun, M. (2020). As-encountered prediction of tunnel boring machine performance parameters using recurrent neural networks. Transportation Research Record, 2674(10), 241-249. https://journals.sagepub.com/doi/abs/10.1177/0361198120934796

Incremental and Approximate Computations for Accelerating Deep CNN Inference

Published in ACM Transactions on Database Systems, 2020

This paper proposes Krypton, a new multi-query-optimization technique for “incremental inference” for convolutional neural networks. Empirical evaluations show up to 34X speedups for occlusion-based explanations and up to 5X speedups for video analytics.
Invited paper.

Recommended citation: Nakandala, S., Nagrecha, K., Kumar, A., & Papakonstantinou, Y. (2020). Incremental and approximate computations for accelerating deep CNN inference. ACM Transactions on Database Systems (TODS), 45(4), 1-42. https://dl.acm.org/doi/abs/10.1145/3397461

Satellite Image Atmospheric Air Pollution Prediction through Meteorological Graph Convolutional Network with Deep Convolutional LSTM

Published in 2020 International Conference on Computational Science and Computational Intelligence (CSCI), 2020

This paper proposes a deep learning technique utilizing satellite imagery provide predictions of air pollution levels several hours in advance.

Recommended citation: Muthukumar, P., Cocom, E., Nagrecha, K., Holm, J., Comer, D., Lyons, A., ... & Pourhomayoun, M. (2020, December). Satellite Image Atmospheric Air Pollution Prediction through Meteorological Graph Convolutional Network with Deep Convolutional LSTM. In 2020 International Conference on Computational Science and Computational Intelligence (CSCI) (pp. 521-526). IEEE.

Sensor-Based Air Pollution Prediction using Deep CNN-LSTM

Published in 2020 International Conference on Computational Science and Computational Intelligence (CSCI), 2020

This paper proposes a deep learning technique utilizing a 1-D CNN-LSTM to predict air pollution levels using low-cost sensors.

Recommended citation: Nagrecha, K., Muthukumar, P., Cocom, E., Holm, J., Comer, D., Burga, I., & Pourhomayoun, M. (2020, December). Sensor-Based Air Pollution Prediction using Deep CNN-LSTM. In 2020 International Conference on Computational Science and Computational Intelligence (CSCI) (pp. 694-696). IEEE.

Cerebro: A Layered Data Platform for Scalable Deep Learning

Published in 11th Annual Conference on Innovative Data Systems Research (CIDR 21), 2021

This vision paper introduces Cerebro, a layered data platform to support scalable model selection and building procedures.

Recommended citation: Kumar, A., Nakandala, S., Zhang, Y., Li, S., Gemawat, A., & Nagrecha, K. (2021, January). Cerebro: A Layered Data Platform for Scalable Deep Learning. In 11th Annual Conference on Innovative Data Systems Research (CIDR 21) https://par.nsf.gov/servlets/purl/10227388

Model-Parallel Model Selection for Deep Learning Systems

Published in Proceedings of the 2021 International Conference on Management of Data, 2021

This competition abstract proposes a new system for parallelized training of multiple deep learning models that do not fit into the memory of a single GPU.
Solo-authored. Won first place at the ACM SIGMOD Student Research Competition.

Recommended citation: Nagrecha, K. (2021, June). Model-Parallel Model Selection for Deep Learning Systems. In Proceedings of the 2021 International Conference on Management of Data (pp. 2929-2931). https://arxiv.org/pdf/2107.06469

Gradient-based Algorithms for Machine Teaching

Published in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021

This paper proposes a new machine teaching algorithm that uses gradients to select data samples that are most effective in educating human or machine learners.

Recommended citation: Wang, P., Nagrecha, K., & Vasconcelos, N. (2021). Gradient-based Algorithms for Machine Teaching. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (pp. 1387-1396). http://openaccess.thecvf.com/content/CVPR2021/papers/Wang_Gradient-Based_Algorithms_for_Machine_Teaching_CVPR_2021_paper.pdf

Hydra: A System for Large Multi-Model Deep Learning (Preprint)

Published in Arxiv Preprint, 2021

This paper introduces Hydra, a system for parallel training of large-scale deep learning models. Empirical evaluations demonstrate up to 80% improved performance over the state-of-the-art with near-linear speedups over traditional techniques.

Recommended citation: Nagrecha, K., & Kumar, A. (2021). Hydra: A System for Large Multi-Model Deep Learning. arXiv preprint arXiv:2110.08633. https://arxiv.org/pdf/2110.08633

talks

teaching