20th Prediction Science Seminar
- Date
- November 24th, 2023 (Fri.) 14:00-15:30 (JST)
- Language
- Engish
- Place
- C107(R-CCS) or online(zoom)
To join the seminar, please contact the Prediction Science Seminar Office: prediction-seminar[remove here]@ml.riken.jp
Program
| Time | Content | Speaker |
|---|---|---|
| 14:00-15:00 | Implementing Reservoir Computing in Practice | Dr. Kengo NAKAI (Okayama University) |
| 15:00-15:30 | Discussion | - |
Abstract
It has been reported that reservoir computing is effective in the inference of time-series and some characteristics. A reservoir is a recurrent neural network whose internal parameters are not adjusted to fit the data in the training process. What is done is to train the reservoir by feeding it an input time-series and fitting a linear function of the reservoir state variables to a desired output time-series. Due to this approach of reservoir computing we can save a great amount of computational costs, which enables us to deal with a complex deterministic behavior. In this talk, we show how to learn time-series data by using reservoir computing and discuss the prediction results using the constructed models.
Organizer
- Prediction Science Laboratory (RIKEN CPR)
- RIKEN Center for Biosystems Dynamics Research (BDR）
Co-organizer
- Data Assimilation Research Team (R-CCS)
- RIKEN Interdisciplinary Theoretical and Mathematical Sciences Program
- Environmental Metabolic Analysis Research Team (RIKEN CSRS)
- Computational Climate Science Research Team (R-CCS)
- Medical Data Deep Learning Team (R-IH)
- Medical Data Mathematical Reasoning Team (R-IH)
- Laboratory for Physical Biology (RIKEN BDR)