Cyclotron Seminar ALICE/NA61/SPbSU - ML lectures by Michail

→ Europe/Moscow
SPbSU

SPbSU

    • 11:00 → 11:30
      EM-shower recognition (Opera) 30m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 11:30 → 12:00
      PID: DNN and BDT models with high efficiency 30m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 12:00 → 12:30
      Tea break 30m
    • 12:30 → 12:45
      Pi0-gamma separation at LHCb 15m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 12:45 → 13:00
      CMS anomalies detection using ML 15m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 13:00 → 13:10
      LHCb robo-shifter based on AdaBoost 10m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 13:10 → 13:25
      Detector optimization: gaussian processes 15m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 13:25 → 14:00
      Discussion 35m
    • 14:00 → 15:30
      Lunch 1h 30m
    • 15:30 → 15:45
      Fakes rejection at LHCb 15m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 15:45 → 16:00
      Noise hits rejection at COMET 15m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 16:00 → 16:15
      Artificial retina for LHCb VELO 15m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 16:15 → 16:30
      Cellular automaton + ML at Bell2 15m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 16:30 → 16:45
      RNN, CNN approaches in track recognition 15m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 16:45 → 17:15
      Dicussion 30m
    • 17:15 → 18:00
      Tea break 45m
    • 18:00 → 18:30
      Intro to ML + bBDT 30m
      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 18:30 → 19:00
      Model selection & hyperparameters tuning 30m

      a. Train/Validation/Test split (5) b. Bias vs variance problem (5)
      c. Hyperparameters tuning based on train/validation scores (10) 
d. ANNs vs BDTs (10)

      Speaker: Mikhail Hushchyn (Yandex School of Data Analysis (RU))
    • 19:00 → 19:30
      Discussion 30m