The document discusses the integration of machine learning (ML) with simulation models to enhance their predictive capabilities and adaptability in changing environments. It explores key ideas such as learning system behavior, identifying influential variables, and employing incremental learning for adaptive behavior in simulations. The author emphasizes the potential of ML-aided simulations to reduce human input while improving the accuracy and confidence of simulation outcomes.

1. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
ML-Aided Simulation:A Conceptual Framework for
Integrating Simulation Models with Machine Learning
Mahmoud Elbattah, Owen Molloy

2. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Integrating M&S with ML: Why, When, How?
2
Image Source:
Robinson, S. (2004). Simulation: The Practice of Model Development and Use. Chichester: Wiley.

3. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Basic View on Systems & Simulations
3
Source: Zeigler, B.P. and Sarjoughian, H.S., 2012. Guide to modeling and simulation of systems of systems.
Springer Science & Business Media.

4. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Motivational Questions
• How simulation models can learn about changes in the actual
system behaviour with minimal human input?
• Is it possible to integrate simulation models with ML models to
enable that learning process to happen in an automated
manner? If so, how?
• Can the integration with ML lead to a higher level of confidence
in simulations, given by a more measurable accuracy of ML
models?
4

5. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Key Idea 1:
Learning to Predict the System Behaviour
5

6. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Key Idea 2:
Identify Predictable Influential Variables
6
• Influential Variable: A variable that has a significant influence
on the system behaviour with respect to the question(s) of
interest, whereas the variation of that variable can lead to a
change in policy, strategy, or decision-making.

7. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Key Idea 2:
Identify Predictable Influential Variables (cont’d)
7

8. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Key Idea 2:
Identify Predictable Influential Variables (cont’d)
8

9. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Key Idea 3:
Incremental Learning = Adaptive Behaviour
9
A Basic Feedback Loop 1
Feedback Loops Aided by ML.
Source: Forrester, J.W., 1968. Principles of Systems, Text and Workbook, Wright-Allen Press-US.

10. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Key Idea 3:
Incremental Learning = Adaptive Behaviour (cont’d)
10

11. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
How This Can Be Useful?
• The power of learning from data is that the entire process can
be automated with minimal, or without, involvement of human
input.
• This can be useful for modeling dynamic systems that exist in
rapidly changing environments (Concept Drift).
• Realising “self-adaptive” simulation models that can adapt their
behaviour based on ML predictions.
• May help reduce the epistemic uncertainty1 attributed to the
subjective interpretation of system knowledge.
• Works effectively in situations where the system behaviour can
be largely described and learned by examples.
11
Source: Oberkampf, W.L., DeLand, S.M., Rutherford, B.M., Diegert, K.V. and Alvin, K.F., 2002. Error and uncertainty in modeling
and simulation. Reliability Engineering & System Safety, 75(3), pp.333-357.

12. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Further Directions: More Complex ML for More
Complex Systems
12

13. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
Closing Thought
• Machine Learning: The subfield of computer science that
gives computers the ability to learn without being explicitly
programmed (Arthur Samuel 1959).
• ML-Aided Simulations: Simulation models given the
ability to adapt to new knowledge without being explicitly
informed by modellers.
13
Source: Samuel, A.L., 1959. Some Studies in Machine Learning Using the Game of Checkers. IBM Journal of Research and
Development, 3(3), pp.210-229.

14. ACM 2018 Conference on Principles of Advanced Discrete Simulation (PADS)
THANK YOU!
Mahmoud Elbattah
mahmoud.elbattah@nuigalway.ie