4.3 KiB
4.3 KiB
Presentation Structure
List of all content
| Slide | Content | TODO |
|---|---|---|
| 4 | Background: Motivation | - |
| 5 | Background: Previous Work | X |
| 6 | Background: Presumptions & Notation | - |
| 7 | Background: Optimization vs graph based methods | - |
| 9-11 | Proximal: Theory | - |
| 12 | Proximal: Sim: Able to replicate literature | - |
| 13 | Proximal: Sim: Choice of gamma | X |
| 14 | Proximal: Sim: FER due to DFR | - |
| 15-16 | Proximal: Oscillation of estimate | - |
| 17 | Proximal: Improved: Emprirical FoM | - |
| 18 | Proximal: Improved: Algorithm | - |
| 19 | Proximal: Improved: Results | - |
| 21-24 | LP Decoding: General Theory | - |
| 25-26 | LP Decoding: ADMM Theory | - |
| 27 | LP Decoding: ADMM Sim: Able to replicate literature | - |
| 28-29 | LP Decoding: ADMM Sim: Choice of mu and rho | S |
| 31 | Comparison: Number of required iterations | S |
| 32 | Comparison: Time Complexity | S |
| 33 | Comparison: FER and BER | S |
| 34 | Comparison: FER and BER for more codes | S |
| 35 | Comparison: Algorithm structure | - |
| 36 | Comparison: Algorithm structure: Message passing | X |
| 38 | Conclusion and outlook | R |
| 41 | Supplementary slides: Choice of gamma for more codes | - |
| 42 | Supplementary slides: Choice of mu and rho for more codes | - |
| - | Current and new approaches for improving proximal decoging | TODO |
Questions
- Is a notational slide that important in this case?
- Yes, otherwise the listeners won't understand the following slides
- Should LP Decoding and ADMM even be mentioned?
- Yes, the basis for the presentation is still the thesis. Also, the title of the presentation is "-- optimization --", not "-- proximal decoding --"
- Mention approaches not taken (e.g., Adam for proximal decoding)?
- Yes: "Zur Debatte stellen"
- Present current state of research on improved algorithm (e.g., questions
raised in last meeting with Jonathan)?
- Yes, at least partly mention current approaches and insights not present in the thesis
- Compare proximal and LP decoding using ADMM?
- Maybe at least mention that the general structure of the objective function is similar and why we would expect proximal decoding to work better and why it's surprising it doesn't.
- Present insights on parameter choice?
- Maybe only in supplementary slides?
- Corroborate results with slides showing additional simulations for multiple
codes?
- Maybe only in supplementary slides?