Network refactor and reluctant Activity driven model #28
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Helps to write commit messages, that are conforming to CONTRIBUTING.md.
Pointed out that Seldon contains a commit template under `.gitmessage`. Might be useful for contributors. Co-authored-by: Amrita Goswami <[email protected]>
Change Network to a class templated on the AgentType. All functions were moved to the hpp file. Co-authored-by: Moritz Sallermann <[email protected]>
WIP commit, with simulation run functions and IO commented out. But hey presto, it compiles! Co-authored-by: Moritz Sallermann <[email protected]>
AgentType has to be included in simulation and model. TODO: Fix IO functions Co-authored-by: Moritz Sallermann <[email protected]>
Some of the tests work now. However, IO does not yet work. TODO: Models need to use agents from Network. Co-authored-by: Moritz Sallermann <[email protected]>
We had to change the models to take agents from network. Some tests work now. TODO: Activity driven model test. Also agents should be private not public, in network. Co-authored-by: Moritz Sallermann <[email protected]>
TODO: Activity driven model test
TODO: Fix the bot and single agent test. x(t) somehow does not work. Co-authored-by: Moritz Sallermann <[email protected]>
The analytical solution was not being computed correctly because the of a reference that should have been a copy (and was before the refactor)
- cleaned up the code after running clang-format. - using NetworkT instead of using Network, to not shadow the Network class - also put the `typename` in `using typename NetworkT::WeightT` etc. - renamed unittests so that they dont have any whitepsace, i.empty "Test_Activity Driven" -> "Test_Activity_Driven", because such paths are nicer to work with
Started to remove unique_ptr in Simulation class. Broken at the moment
Since model actually relies on polymorphic behaviour, it was wrong to not access it via a pointer to the base class. Co-authored-by: Amrita Goswami <[email protected]>
The ModelBase class is unnecessary now that simulation is templated on the agent type. Therefore, it has been removed. Further, `model.n_iterations()` is now a getter function for `model._n_iterations` and there is a new `mode.initialize_iterations()` function that contains behaviour that needs to be exectued before a `simulation.run()` is started.
Since Simulation is now templated on the agent type, this is not needed. Co-authored-by: Amrita Goswami <[email protected]>
Forgot to run the simulations in main function so far. Yes, really. Created SimulationInterface to run simulations with different agent types via a common interface. Co-authored-by: Amrita Goswami <[email protected]>
Some TODO comments that were no longer needed have been removed. Co-authored-by: Amrita Goswami <[email protected]>
Straight forward and may be useful in the future. Default type is double Co-authored-by: Amrita Goswami <[email protected]>
Indentation of a comment in network and std::is_same_v<T> instead of std::is_same<T>::value
Template Network on the agent type
Just to be consistent with the rest of the nomenclature. Co-authored-by: Amrita Goswami <[email protected]>
The ActivityDrivenModel driven model has been extended with a concept, that we call reluctance. This quantifies how receptive to others' opinions an agent is. Reluctance is an per agent quantity and therefore the IO had to be changed slightly. The bots can be thought of as agents with infinite reluctance. TODO: generation of distributions for reluctance, possibly correlated with activities. Co-authored-by: Amrita Goswami <[email protected]>
Will be used later to draw the reluctances, it's a normal distribution truncated to the interval [eps, infty].
ActivityDrivenModel::generate_agents_from_power_law(), now also generates the reluctances from a truncated normal distribution. Optionally they can be correlated with the activities. TODO: parsing of the config file to control these options Co-authored-by: Amrita Goswami <[email protected]>
Reluctance options are not part of ConfigParser. Defaults are set and some checks are completed. Also added an example with the reluctance. TODO: Make a unit test on the ActivityDrivenModel with reluctances. Co-authored-by: Moritz Sallermann <[email protected]>
We modified the unit test with one bot and one agent to now include the reluctance (of the agent). Co-authored-by: Moritz Sallermann <[email protected]>
Otherwise, we can't figure out why the tests fail on the CI Co-authored-by: Moritz Sallermann <[email protected]>
We now set the width of the distribution (before truncating) for the reluctant agents to a smaller value, so that tests on the CI are comparable Co-authored-by: Moritz Sallermann <[email protected]>
Testing has shown that our implementation of the covariance factor was unsatisfactory in several regards. Since, at the time, it is not so trivial to improve, we have removed it. The activities and reluctances can, of course, still be arbitrarily correlated -- simply by reading in an initial opinions file. Co-authored-by: Amrita Goswami <[email protected]>
Seldon can now optionally print files starting from a user-defined value which is greater than 0. Right now the zeroth iteration files are always created (if outputs are enabled).
Now the user can optionally disable outputting the initial opinions and network. By default, the initial opinions and network are always printed.
Changed the variable deciding when to start outputting files to the more intuitive `start_output`, as opposed to the old `start_print_iteration` Co-authored-by: Moritz Sallermann <[email protected]>
Small io feature
Reluctant activity driven model
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.