I think I’m OK to just practice my presentation tomorrow morning. I have the main points I want to hit. It’s hard to practice today due to being in the library.
I’m a bit wired today, think its the coffee and just meeting mom. Maybe excitement or something. I’m sure the rain will sort me out in this respect.
I’ve been slowly working my way through understanding the philsophy of science. It’s really dense.
I have made some progress in understanding Poincares epistemic equivalent explanations example and also plan on reading some Hume.
Maybe I should look at an academic writing workshop?
Need to prepare for meeting with Mark on Friday.
An exercise could be to try and develop interpretations of what I’m doing that are totally tangental to what I have so far.
If I was describe what i see as the potential for LCA stuff
- Machine learning, by which I mean using techniques for knowledge representation from CS to represent knowledge for LCA’s. How could I get more concrete here.
Three main themes emerging so far from the uncertainty analysis.
It seems that all the following are due to the fact that I have programmed a version of the TCD LCA model.
- Transparency of assumptions (due to Python).
- Visualisation of assumptions.
- Making explicit tacit knowledge.
- Inherent limitations of replication of scientific knowledge in software?
- Insight into the model that has been created (due to things like sensitivity analysis).
- A quality not obvious from the surface level.
- Prodding and probing the model.
- Both of these display the quality (value) of the information that the LCA provides.
- The value of information in decision theory.
- The risk we can take on with this information (decision theory).
The dissertation will spend some time explaining the problems that an uncertainty analysis hopes to solve and then why the particular method chosen solves those problems and more. There is a element of persuasion involved in proving that those problems need to be solved. There is an element of informing in how the surface level problems (that is, just doing a basic uncertainty analysis) was solved.
Structure
- Introduction
- what is uncertainty in LCA’s.
- Getting familiar with the LCA that this paper will tackle.
- Methodology/ approach to doing an uncertainty analysis of said LCA.
- Methodology for uncertainty analysis.
- Methodology for model representation.
- Results, what did we find out about the model.
- What this methodology revealed, new problems.
- Why are these problems relevant.
To use the 80/20 rule which I just learned about today, the uncertainty analysis might produce 80% of the actual output but be 20% of the work. Although, if you count the coding representation of the model as work towards it then the principle doesn’t apply.
Could it be considered two projects in one. There are problems that spin off from the approach taken, that is programming a model. And there are problems due to the uncertainty analysis.
Methodology would technically be all involved in development of representation. Future problems section could contain the related material to do with the model but a thorough detail of the model under methodology probably makes sense.
Great Work (Continued)
I like the idea of trying to be the best in the world at what you do. More than likely you won’t be but it simplifies things, theres no middle ground. For instance, he says, build things for people in 100 years time. That’s related to my notion of what I’m coding now, whats its lifecycle, caring is important to good design. … Post is massive need another try.
Stochaistic modelling
Eigenvalues and eigen vectors for repeated processes. Transition matrix.