Midterm Flashcards
modeling real world scenario
- e.g. traffic congestion, weather forecast, real-time laws
Modeling and Simulation
- done inside a confined/isolated environment which does not affect any production system
- In IT and CS, LAB Simulations are usually utilized
LAB SIMULATION
is used to ready your workers
LAB SIMULATION
- crucial tool in various fields of works
- FOCUS: Stepped Time Simulation, Event-based Simulation
SIMULATION TIME HANDLING
are in fixed term intervals
- e.g. weather - temperature, wind speed, precipitation, time
- details for reportings
- for continuous development
- e.g. every two weeks, there will be a report
STEPPED TIME SIMULATION
- determine the behavior based on given time
- ease of implementation
- suitability for continuous system
Advantages of Stepped time simulation
- computational inefficiency: too slow to compute big data
- difficulty in handling rare events - abnormalities in the linear progression
- accuracy trade off
LIMITATIONS of stepped time simulation
has an event trigger
- e.g. earthquake, stock market
EVENT-BASED SIMULATION
- computational efficiency
- higher accuracy
- scalability
Advantages of event-based
- complex implementation
- not suitable for continuous system
LIMITATIONS of event based
- not applicable at the current time
- combination of stepped-time and event-based
HYBRID SIMULATION
- creation of mathematical representation of real world problems/systems
- to study a behavior
- predict future states of the thing you are representing
MODELING
- use of models to study the behavior of a system over time
- e.g. crash - we need to evaluate the timing, precision, activation of the system (e.g. airbags)
- e.g. in it - creating bugs before publicly engaging in the implementation of the project
SIMULATION
similar to stepped time
: changes occur at a specific interval
: e.g. queuing system (depends on the amount of customers), cellular automata (spatially and
temporally finite-state discrete computational systems composed of a finite set of cells evolving
in parallel at discrete time steps.), digital circuits (uses binary signals to process data)
DISCRETE MODEL
changing smoothly and continuously overtime
: e.g. physics based simulation (e.g. boiling point), fluid dynamics (weather prediction),
population growth (predict how much population will grow)
: should not be bothered by intervals
CONTINUOUS MODEL
- countable = discrete
- evolution or changes coming smoothly = continuous
understand the nature of the system
- event-driven data = discrete
- precise measurements available that changes overtime = continuous
data availability
- logical computation = discrete
- based on the availability of data
- complex computation using numerical method = continuous
computational resources
- some studies need d+c approach = HYBRID MODEL
- e.g. heat transfer (continuous), 3 stoves (discrete), capacity of the stoves to heat water faster
from point a to point b (0 to boiling point)
accuracy
solves scientific problem
- whenever analytical methods are not feasible
NUMERICAL METHODS
based on differential methodologies
- get first the center
e.g. heat conduction
FINITE DIFFERENCE METHOD
- complex engineering equations
- CUT THROUGH ONE SEGMENT
- find out the strength of the material/product
FINITE ELEMENT Method
a mathematical technique that uses random numbers to
simulate possible outcomes of an uncertain event
MONTECARLO SIMULATION
approximating the roots of an equation through repetitive
numerical methods
ROOT FINDING SIMULATIONS
ensuring that the computational model is correct
- the mathematical formula you used is correct
VERIFICATION
what you used aligns with real world data
VALIDATION
Various type of validation
Theoretical
Code Verification
Experimental Validation
- there are established theory
Theoretical
ensure that the numerical methods used are correct
- benchmarking
- e.g. standard for stuffies
Code Verification
you do various experiments to validate your claims
Experimental Validation
COMMON MISTAKES
Overfitting
Numerical Instability
Poor Data Quality -
data is not fit with the dataset
Overfitting
there is diversity in numerical data
Numerical Instability
there many variables you should be cleaning (especially the ones you
don’t need)
Poor Data Quality
