Tata Institute of Fundamental Research
Rate Distortion in the One-shot Setting
STCS Student Seminar
Speaker:
Sayantan Chakraborty
Organiser:
Aditya Nema
Date:
Friday, 22 Jun 2018, 17:15 to 18:15
Venue:
A-201 (STCS Seminar Room)
(Scan to add to calendar)
Abstract:
We shall discuss classical rate distortion, i.e., message compression with a distortion criterion, in the one-shot setting with and without side information at the decoder. The rate distortion problem with side information at the decoder is also called Wyner Ziv problem in classical information theory, which is well know in IID asymptotic setting.
This talk will mainly focus on establishing achievable schemes for simple point to point case without side information and for Wyner Ziv problem as well, in one shot regime. The achievable rates are optimal as well (but optimality won't be a part of the talk). Along the way we will be highlighting the use of convex split lemma (for Wyner Ziv case). If time permits we will also look at an informal proof for the convex split lemma. The only required tool for the talk would be basic probability theory.
| Speaker: | Sayantan Chakraborty |
| Organiser: | Aditya Nema |
| Date: | Friday, 22 Jun 2018, 17:15 to 18:15 |
| Venue: | A-201 (STCS Seminar Room) |
(Scan to add to calendar)
Abstract:
We shall discuss classical rate distortion, i.e., message compression with a distortion criterion, in the one-shot setting with and without side information at the decoder. The rate distortion problem with side information at the decoder is also called Wyner Ziv problem in classical information theory, which is well know in IID asymptotic setting.
This talk will mainly focus on establishing achievable schemes for simple point to point case without side information and for Wyner Ziv problem as well, in one shot regime. The achievable rates are optimal as well (but optimality won't be a part of the talk). Along the way we will be highlighting the use of convex split lemma (for Wyner Ziv case). If time permits we will also look at an informal proof for the convex split lemma. The only required tool for the talk would be basic probability theory.
This talk will mainly focus on establishing achievable schemes for simple point to point case without side information and for Wyner Ziv problem as well, in one shot regime. The achievable rates are optimal as well (but optimality won't be a part of the talk). Along the way we will be highlighting the use of convex split lemma (for Wyner Ziv case). If time permits we will also look at an informal proof for the convex split lemma. The only required tool for the talk would be basic probability theory.