Anatomy of Dynamical Systems
D. Roy
University of Amsterdam
June 2, 2024
D. Roy (University of Amsterdam) ABM June 2, 2024 1 / 16
Dynamical System
D. Roy (University of Amsterdam) ABM June 2, 2024 2 / 16
Dynamical System
D. Roy (University of Amsterdam) ABM June 2, 2024 2 / 16
Dynamical System
D. Roy (University of Amsterdam) ABM June 2, 2024 2 / 16
Dynamical System
D. Roy (University of Amsterdam) ABM June 2, 2024 2 / 16
Dynamical System
D. Roy (University of Amsterdam) ABM June 2, 2024 2 / 16
Dynamical System
D. Roy (University of Amsterdam) ABM June 2, 2024 2 / 16
Dynamical System
Consider the following autonomous ordinary differential equation
(ODE):
dx
= F(x) (1)
dt
where F(x) is vector valued function of x.
D. Roy (University of Amsterdam) ABM June 2, 2024 3 / 16
Dynamical System
Consider the following non-autonomous ordinary differential equation
(ODE):
dx
= F(t, x) (2)
dt
where F(t, x) is a vector valued function of both t and x.
D. Roy (University of Amsterdam) ABM June 2, 2024 4 / 16
Dynamical System
Consider the following non-autonomous ordinary differential equation
(ODE) with control u(t) and parameters p:
dx
= F(t, x, u; p) (3)
dt
where F(t, x, u, p) is a function of time t, state x, control u, and
parameters p.
D. Roy (University of Amsterdam) ABM June 2, 2024 5 / 16
Navier-Stokes equation
The Navier-Stokes equation for an incompressible fluid in Cartesian
coordinates is given by:
∂u
ρ + (u · ∇)u = −∇p + ν∇2 u + f (4)
∂t
with the continuity equation:
∇·u=0 (5)
where u represents the velocity vector field, t is time, p is pressure, ρ is
density, ν is the kinematic viscosity, and f denotes external forces.
D. Roy (University of Amsterdam) ABM June 2, 2024 6 / 16
Schrödinger wave equation
The time-dependent Schrödinger wave equation is given by:
∂Ψ
iℏ = ĤΨ (6)
∂t
where i is the imaginary unit, ℏ is the reduced Planck constant, Ψ is
the wave function, t is time, and Ĥ is the Hamiltonian operator.
D. Roy (University of Amsterdam) ABM June 2, 2024 7 / 16
Challenges
1 Unknown ‘f’
D. Roy (University of Amsterdam) ABM June 2, 2024 8 / 16
Challenges
1 Unknown ‘f’
2 Non-linear ‘f’
D. Roy (University of Amsterdam) ABM June 2, 2024 8 / 16
Challenges
1 Unknown ‘f’
2 Non-linear ‘f’
3 High dimensional
D. Roy (University of Amsterdam) ABM June 2, 2024 8 / 16
Challenges
1 Unknown ‘f’
2 Non-linear ‘f’
3 High dimensional
4 Multi-scale
D. Roy (University of Amsterdam) ABM June 2, 2024 8 / 16
Challenges
1 Unknown ‘f’
2 Non-linear ‘f’
3 High dimensional
4 Multi-scale
5 Chaotic
D. Roy (University of Amsterdam) ABM June 2, 2024 8 / 16
Challenges
1 Unknown ‘f’
2 Non-linear ‘f’
3 High dimensional
4 Multi-scale
5 Chaotic
6 Noise, Stochastic
D. Roy (University of Amsterdam) ABM June 2, 2024 8 / 16
Challenges
1 Unknown ‘f’
2 Non-linear ‘f’
3 High dimensional
4 Multi-scale
5 Chaotic
6 Noise, Stochastic
7 Uncertainty
D. Roy (University of Amsterdam) ABM June 2, 2024 8 / 16
Uses of models
1 Predict using ‘f’ (ensemble)
D. Roy (University of Amsterdam) ABM June 2, 2024 9 / 16
Uses of models
1 Predict using ‘f’ (ensemble)
2 Design and Optimization (airplane, Traffic)
D. Roy (University of Amsterdam) ABM June 2, 2024 9 / 16
Uses of models
1 Predict using ‘f’ (ensemble)
2 Design and Optimization (airplane, Traffic)
3 Control
D. Roy (University of Amsterdam) ABM June 2, 2024 9 / 16
Uses of models
1 Predict using ‘f’ (ensemble)
2 Design and Optimization (airplane, Traffic)
3 Control
4 Understanding mechanisms (interpretable)
D. Roy (University of Amsterdam) ABM June 2, 2024 9 / 16
Uses of models
1 Predict using ‘f’ (ensemble)
2 Design and Optimization (airplane, Traffic)
3 Control
4 Understanding mechanisms (interpretable)
5 Generalize
D. Roy (University of Amsterdam) ABM June 2, 2024 9 / 16
Why Model?
Everybody models - all of you forecast the future in some way,
predicting the way a friend reacts, planning your future1
1
Epstein, “Why Model?”
D. Roy (University of Amsterdam) ABM June 2, 2024 10 / 16
Why Model?
1 Explain (very distinct from predict)
2 Guide data collection
3 Illuminate core dynamics
4 Suggest dynamical analogies
5 Discover new questions
6 Promote a scientific habit of mind
7 Bound outcomes to plausible ranges
8 Illuminate core uncertainties.
9 Offer crisis options in near-real time
10 Demonstrate tradeoffs / suggest efficiencies
11 Train practitioners
12 Discipline the policy dialogue
13 Educate the general public
14 Reveal the apparently simple (complex) to be complex (simple)
D. Roy (University of Amsterdam) ABM June 2, 2024 11 / 16
Why Model?
Explain (very distinct from predict)
Plate tectonics describe the mechanism of earthquakes, but doesn’t
allow us to predict.
generative explanation, where large scale regularities such as wealth
distributions, spatial settlement patterns, or epidemic
dynamics—emerge in populations of heterogeneous software individuals
(agents) interacting locally under plausible behavioral rules.
D. Roy (University of Amsterdam) ABM June 2, 2024 12 / 16
Why Model?
Guide Data Collection
Inductivism:: ’Science proceeds from observation, and then models are
constructed to ’account for’ the data.’ First collects lots of data and
then runs regressions on it or machine learning.
A model is simply a theory - the model may predict a particular
dynamic that you would need to measure to corroborate. This happens
a lot in Physics - Maxwell’s equations, Higg’s Boson, etc.
D. Roy (University of Amsterdam) ABM June 2, 2024 13 / 16
Why Model?
Illuminate Core Dynamics: All the Best Models are Wrong
A lot of ABMs do this, this is really about furthering scientific
understanding. Building the simplest theory you can do describe the
processes and dynamics of a system without it necessarily being
predictive.
The Lotka-Volterra ecosystem model, Hooke’s Law, or the
Kermack-McKendrick epidemic equations. They continue to form the
conceptual foundations of their respective fields
D. Roy (University of Amsterdam) ABM June 2, 2024 14 / 16
Why Model?
Suggest Analogies
It is a startling and wonderful fact that a huge variety of seemingly
unrelated processes have formally identical models (i.e., they can all be
seen as interpretations of the same underlying formalism).
Do revolutions, or religions, or the adoption of innovations unfold like
epidemics? Is it useful to think of these processes as formal analogues?
If so, then a powerful pre-existing theory can be brought to bear on the
unexplored field, perhaps leading to rapid advance. This type of
thinking/science is endemic in Complexity Science.
D. Roy (University of Amsterdam) ABM June 2, 2024 15 / 16
Why Model?
To me, however, the most important con-
tribution of the modeling enterprise—as dis-
tinct from any particular model, or model-
ing technique—is that it enforces a scientific
habit of mind, which I would characterize as
one of militant ignorance—an iron commit-
ment to ”I don’t know.”
— [Joshua M. Epsteina ]
a
Epstein, “Why Model?”
D. Roy (University of Amsterdam) ABM June 2, 2024 16 / 16