Bit of a shakey start to the day, burnt my mouth on pancakes…again. Becuase I was getting stressed about just having them. I need to really start being conscious of this with my food. I add far too much anxiety to the whole occasion.
I think I need to devote a few days to reading literature. Use the ‘Gwern method’ if I can, what do I expect to see. I can’t just be reading passively. There seem to be lots of meta studies. How do I get a feel for how people actually performed the uncertainty analysis?
Feeling a bit fatigued today, coffee two around 12pm. Think it could be the workouts because I think I got enough sleep (although was very tough get up this morning).
I feel frustrated that I haven’t more experience with general statistical methods.
Bit of pain just below right delt today.
Percentage of unknowns from knowns.
5kg for hydrogen needed for Excel.
I need to spend some time mapping out options of where to focus time for the coming weeks so I’m not faced with the notion that I’ve to study everyday (unless I feel its necessary of course).
”My first reaction was, its useless to make anything” Feynman on the bomb.
Research
Could check out CMLCA.
It seems that uncertainty propogation could make use of the matrix formalism of the LCA. Things like stochaistic modelling seemed to be represented as eigen vectors and values.
Matrix representation seems to try and view the LCA as a system, or at least the product flow as a system which is interesting. It seems like a field that is well explored if you think of it like a system. It’s where this overlap of system dynamics and operations design/research seem to overlap.
Representation seems to be the key word here. Do all representations have the same power? Presumably not. I think thats generally the tieing idea for the machine learning or the matrix or whatever. Generally, the exploration of types of representation of the LCA is something that would be a phD in itself. You could learn about system dynamics, graph models, matrix pertubation theory, collections of differential equations etc.
Representation is the gateway to leveraging computation. So I suppose thats the initial challenge I was faced with. How to represent the model in a way I’m familiar with and also that allows us to leverage computation.
Knowledge representation
Another idea I haven’t explored fully is leveraging all the work in developing agent based systems to analysise LCA. For scenario based uncertainty it could be useful. This overlaps with the ‘tacit knowledge’ aspects I think but just probably more in the machine learning space.
Computational
While I thought that viewing an LCA in computational terms was a bit of a stretch it seems to make more and more sense. It’s really a system of checks and balances, the matrix representation does seem quite natural. The goal of any knowledge ‘created’ in the assessment aims to become explicit in some way. Once explicit it is a number with some operation.
How do non linear equations fit in here?
How do I put myself in a position to perform any kind of uncertainty analysis we would want to do?
- What are the key features of a model necessary to perform uncertainty propagation.
- Make sure it is an accurate replication. That its producing expected results (this would involve checking numbers with Liam).
The rank of a matrix is the dimensions of the vector space spanned by its rows.