Sampling-based Algorithms for Fast and Deployable AI
We present sampling-based algorithms with provable guarantees to alleviate the increasingly prohibitive costs of training and deploying modern AI systems. At the core of this thesis lies importance sampling, which we use to construct representative subsets of inputs and compress machine learning mod...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2022
|
| Online Access: | https://hdl.handle.net/1721.1/139924 https://orcid.org/0000-0002-6776-9493 |
| _version_ | 1826202535339229184 |
|---|---|
| author | Baykal, Cenk |
| author2 | Rus, Daniela |
| author_facet | Rus, Daniela Baykal, Cenk |
| author_sort | Baykal, Cenk |
| collection | MIT |
| description | We present sampling-based algorithms with provable guarantees to alleviate the increasingly prohibitive costs of training and deploying modern AI systems. At the core of this thesis lies importance sampling, which we use to construct representative subsets of inputs and compress machine learning models to enable fast and deployable systems. We provide theoretical guarantees on the representativeness of the generated subsamples for a variety of objectives, ranging from eliminating data redundancy for efficient training of ML models to compressing large neural networks for real-time inference. In contrast to prior work that has predominantly focused on heuristics, the algorithms presented in this thesis can be widely applied to varying scenarios to obtain provably competitive results. We conduct empirical evaluations on real-world scenarios and data sets that demonstrate the practicality and effectiveness of the presented work. |
| first_indexed | 2024-09-23T12:09:02Z |
| format | Thesis |
| id | mit-1721.1/139924 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T12:09:02Z |
| publishDate | 2022 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1399242022-02-08T03:36:51Z Sampling-based Algorithms for Fast and Deployable AI Baykal, Cenk Rus, Daniela Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science We present sampling-based algorithms with provable guarantees to alleviate the increasingly prohibitive costs of training and deploying modern AI systems. At the core of this thesis lies importance sampling, which we use to construct representative subsets of inputs and compress machine learning models to enable fast and deployable systems. We provide theoretical guarantees on the representativeness of the generated subsamples for a variety of objectives, ranging from eliminating data redundancy for efficient training of ML models to compressing large neural networks for real-time inference. In contrast to prior work that has predominantly focused on heuristics, the algorithms presented in this thesis can be widely applied to varying scenarios to obtain provably competitive results. We conduct empirical evaluations on real-world scenarios and data sets that demonstrate the practicality and effectiveness of the presented work. Ph.D. 2022-02-07T15:12:58Z 2022-02-07T15:12:58Z 2021-09 2021-09-21T19:30:51.950Z Thesis https://hdl.handle.net/1721.1/139924 https://orcid.org/0000-0002-6776-9493 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Baykal, Cenk Sampling-based Algorithms for Fast and Deployable AI |
| title | Sampling-based Algorithms for Fast and Deployable AI |
| title_full | Sampling-based Algorithms for Fast and Deployable AI |
| title_fullStr | Sampling-based Algorithms for Fast and Deployable AI |
| title_full_unstemmed | Sampling-based Algorithms for Fast and Deployable AI |
| title_short | Sampling-based Algorithms for Fast and Deployable AI |
| title_sort | sampling based algorithms for fast and deployable ai |
| url | https://hdl.handle.net/1721.1/139924 https://orcid.org/0000-0002-6776-9493 |
| work_keys_str_mv | AT baykalcenk samplingbasedalgorithmsforfastanddeployableai |