level 2: input -> transformations -> output
Table of Contents
The second level in .
Essentially, take the one simple input->output association in level 1: input -> output, and decomposite it to several simple level 1: input -> output associations.
input1 -> [output1 = input2 -> output2 = input3 -> output3 = input3] -> output3
1. examples
- backprop neural net training =
init weight + observation vector + loss function -> better weight:
- observation + init weight + feed forward -> prediction vector
- predition vector + loss function -> error
- error + loss function + feed forward intermediate results + init weight -> gradient of each weight
- gradient of each weight + init weight + learning rate -take a step-> better weight
Backlinks
simulation method
(simple version [use this])
- imagine the whole topic with level 1: input -> output simulation (chopping tree = axe + swing -> tree down + hurray!)
- imagine how the simulation would be carried out step by step with level 2: input -> transformations -> output simulation, in which each step will still be a level 1: input -> output simulation.
- deal with each step seperately with level 2: input -> transformations -> output simulation, which will give you more level 1: input -> output simulations
And when you encounter ideas that are under the same topic but not sequential, like apple and banana are both fruits, just do a simple outline with parallel simulation, like imagining the taxonomy
simulation method
(simple version [use this])
- imagine the whole topic with level 1: input -> output simulation (chopping tree = axe + swing -> tree down + hurray!)
- imagine how the simulation would be carried out step by step with level 2: input -> transformations -> output simulation, in which each step will still be a level 1: input -> output simulation.
- deal with each step seperately with level 2: input -> transformations -> output simulation, which will give you more level 1: input -> output simulations
And when you encounter ideas that are under the same topic but not sequential, like apple and banana are both fruits, just do a simple outline with parallel simulation, like imagining the taxonomy
simulation method
(detailed Process)
The process assumes one topic that all materials falls under, or in term of understanding, one genral scenario that we are observing/imagining(like woodcraft for simulations of chopping wood, sawing, glueing, etc.)
It starts by describing the biggest topic with level 1: input -> output, and perform level 2: input -> transformations -> output decomposition on it, which will give you more level 1: input -> output relations, which you can perform level 2: input -> transformations -> output further on
simulation method
(detailed Process)
The process assumes one topic that all materials falls under, or in term of understanding, one genral scenario that we are observing/imagining(like woodcraft for simulations of chopping wood, sawing, glueing, etc.)
It starts by describing the biggest topic with level 1: input -> output, and perform level 2: input -> transformations -> output decomposition on it, which will give you more level 1: input -> output relations, which you can perform level 2: input -> transformations -> output further on
simulation method
(detailed Process > 2. level 2: input -> transformations -> output)
How you’d imagine the process being carried out in steps/stages
from an effort it requires, the ability to simulate:
Basically, there is only those 2. You can practice level 2 decomposition many times, and each time you’ll get a handful of level 1 simulations (components of the level 2 simulation) that you probably can do more level 2 decomposition on, giving more level 1 input -> output simulations.
You’ll always have level 1 simulations, and in a way you can only have level 1 simulations, cause that’s what simulations are: somethnig happened to A and it changed a little bit to become B. We just remembered that it happens all the time