Sign Up

Zhishen Huang, Department of Applied Mathematics, Doctoral Candidate

'Randomisation for Statistical Machine Learning'

Supervised learning and reinforcement learning problems are eventually formulated as optimisation problems for training. The optimisation algorithms themselves bear interest from the mathematical point of view. This talk discusses the usage of randomisation in optimisation, which renders what used to be impossible for deterministic optimisation algorithms possible.

  The first part of the talk considers the minimisation of nonconvex and nonsmooth objectives, where we give probabilistic guarantees for the perturbed proximal gradient descent algorithm to converge to local minima. A variation of the randomisation format is discussed later where Gaussian noise is injected to each gradient descent step. We point out the ergodicity property of such variation, which is not available for a deterministic version of gradient descent, thus revealing the potential of randomised algorithms for global optimisation.

  The second part of the talk considers using the sketching technique to compress data and evaluate statistics solely based on the sketched dataset. We give theoretical guarantees for the evaluation accuracy of autocorrelation from data sketches and demonstrate numerical performance on the methanol ensemble MD simulation data and the synthetic data.

  • Mr. Sathish Compscience - IADCA

1 person is interested in this event


Join Zoom Meeting
https://cuboulder.zoom.us/j/93718841472

Meeting ID: 937 1884 1472
Password: 444801
One tap mobile
+12532158782,,93718841472#,,,,0#,,444801# US (Tacoma)
+13462487799,,93718841472#,,,,0#,,444801# US (Houston)

Dial by your location
        +1 253 215 8782 US (Tacoma)
        +1 346 248 7799 US (Houston)
        +1 669 900 6833 US (San Jose)
        +1 301 715 8592 US (Germantown)
        +1 312 626 6799 US (Chicago)
        +1 646 558 8656 US (New York)
Meeting ID: 937 1884 1472
Password: 444801
Find your local number: https://cuboulder.zoom.us/u/asYaweJJ3

Join by SIP
93718841472@zoomcrc.com

Join by H.323
162.255.37.11 (US West)
162.255.36.11 (US East)
115.114.131.7 (India Mumbai)
115.114.115.7 (India Hyderabad)
213.19.144.110 (EMEA)
103.122.166.55 (Australia)
209.9.211.110 (Hong Kong SAR)
64.211.144.160 (Brazil)
69.174.57.160 (Canada)
207.226.132.110 (Japan)
Meeting ID: 937 1884 1472
Password: 444801

User Activity

No recent activity