Course Flashcards
(40 cards)
1
Q
What is a model? Describe it
A
- Description of a phenomenon
- Representation of a system
- Simplification
- Set of instructions for generating a behavior
2
Q
What different types of behavior in a model is there?
A
- Model behavior - what the model does
* Model structure - What makes it do what it does?
3
Q
What kind of purpose does a model have?
A
- Conceptual frame
- Communication tool
- Summarizing a large amount of data
- Provide understandning
- Identify missing knowledge
4
Q
What is a system?
A
Collection of entities that act and interact together
5
Q
What is a simulation?
A
- Imitation
- Developing an executable model
- Experimenting with a model
- Category of problem solvning
6
Q
There are several ways of studying a system - what are they?
A
- Experimenting with the actual system
- Experimenting with model of a system
- -> Physical
- -> Mathematical (analytical, statistical, simulation)
7
Q
Why would you like to do a model and not try out the actual (physical) system?
A
- Not accessible (to small, large etc)
- Not existing yet
- Unknown structure
- To expensive
8
Q
What are the 7 dimensions for simulation models?
A
- Object behind the model
- Static vs. Dynamic model
- Domain of the simulation time
- Determinism vs. Stochastic
- How entities are represented
- Kind of dependency
- Domain/range of variables
9
Q
What is the domain of the simulated time?
A
- Continuous - change in small intervals
* Discrete - changes in particular points
10
Q
Describe the deterministic approach vs stochastic
A
- Deterministic - future events can be predicted from simulated history
- Stochastic - probability that a particular event occurs
11
Q
Describe how different entities can be represented
A
- Macro models - describes a single object whichs’s properties are put into relation to each other
- Micro models - several objects whichs properties and relation between are analyzed
- Multi-level-models - objects of one model are aggregated into higher-level objectives and tacked as a model on the higher level. Properties and relations on the higher level may be derived from properties and relations on the lower level objectives.
12
Q
Whats the characteristics of ABM?
A
- Dynamic
- Usually time-discrete
- Mostly Stochastic
- Non-linear
- Micro models, also multi-level models
- All kind of objectives
13
Q
Whats the alternative microsimulation approaches?
A
- Queueing systems (discrete event simulation)
- Object-oriented simulation
- Cellular automata
14
Q
Describe discrete event simulation
A
- Simulation state is only updated during event processing
- Very efficient; easy to distribute
- Event queue contains events that will be done –> advanced time, change state according to event type, generate events and put into a queue.
15
Q
Describe object oriented simulation
A
- Sequence of events as input, triggering state change
16
Q
Describe cellular automata
A
- Every cell is equal
- State transition is local in time and space
- Macro-level phenomena can be observed that cannot be derived from micro-level transition functions
- Parallel update of all cells
17
Q
What is a multi-agent model?
A
- Representation of an original system of the metaphor of a multi-agent system
- The active entities are “agents”
18
Q
What is a multi-agent simulation?
A
Is running a multi-agent model
19
Q
What is an agent?
A
An entity in an environment that is able to perform actions in the environment for fulfilling its goals
20
Q
What is the element of an agent-based simulation model?
A
- Agents
- Interactions
- Environment
21
Q
What is the characteristics of an agent?
A
- Autonomous
- Bounded rationality
- Heterogeneous
- Adaptive
22
Q
Describe a soup?
A
- No structure
- No restriction
- starting point before relationships evolve
23
Q
Describe a map
A
- Explicit space with a coordinate system
* Discrete or continuous
24
Q
What is a good model?
A
- Sufficient
- Reproductible
- Simple
- Comprehensible
- Feasible
- Flexible
- Maintainable/extandable
25
Describe the modelling strategies
* KISS - start simple, extend
* KIDS - start descriptive, simplify
* TAPAS - use existing model and adapt
* Pattern-oriented - identify patterns, modify model untill you capture all data
26
What 7 elemens are included in model design?
1. Agent type
2. Agent properties
3. Environment
4. Agent behaivour
5. Design time step
6. Choose parameters
7. Choose measures
27
What is agent-centered approach?
Observe. agents i real world, develop model, simulate
28
What is interaction-centered approach?
take birds eye view and look for interactions, describe interactions and create agent that interact
29
What is the process-centered approach?
start by describe overall process, identify actors ab distribute sub-process actors
30
What is environmental-centred approach?
if objects react to different environemtn, figure out why
31
What is an error?
* Model does not comply with spec.
32
What is an artefact?
* Core assumptions and accessory assumptions making the model
* There are significant and non-significant assumption
33
What is validity?
* Process of determining if the simulation is accurate
| * Without data exclude artefact by testing/sensitivity analysis
34
what different validity types are there?
* Face validity
* Empirical validity
* Behavioral validity
* Structural validity
35
What is sensitivity analysis?
For testing the impact of parameter settings
| * Factor screening/local sensitivity (small changes) / global sensitivity (entire range)
36
What is calibration?
* Optimization toward intended output
37
What may cause bad calibration?
* Too many degrees of freedom, model brittleness, dependencies
38
What is the relation between sensitivity and calibration?
* Sensitivity between the model output and parameters
| * Calibration is about optimizing the output towards the intended output
39
Whats the traditional do-not's for the simulation?
1. Not enough well-defined objective
2. Inappropriate level of model detail
3. Failure to get the appropriate data
4. Inappropriate output data analysis
40
What should you watch out for during a simulation?
1. Model complexity and brittleness
2. Scalability
3. technical scalability
4. Efficient calibration
5. appropriate level of resolution
