- ES Español (Latinoamérica)

- EN English

5.10. Computational and Data-Driven Physics (CDP)
This Knowledge Area focuses on the intersection of numerical analysis, high-performance computing, and modern data science. It equips physicists with the tools to simulate systems that lack analytical solutions and to extract fundamental laws from massive datasets using machine learning and statistical inference.
5.10.1. CDP/Numerical Methods for Differential Equations in Physics (Core Tier1: 1 hr, Core Tier2: 1 hr) ↑ Back to top
Discrete approximations for solving the fundamental equations of motion and field theories.
Topics:
Core
- Runge-Kutta methods (RK2, RK4) for initial value problems
- Finite Difference Time Domain (FDTD) for Maxwell's equations
- Boundary value problems: Shooting methods and relaxation techniques
- Symplectic integrators: Verlet and leapfrog algorithms for Hamiltonian systems
- Stability analysis and the Courant-Friedrichs-Lewy (CFL) condition
- Finite Element Method (FEM) for complex geometries in elasticity and fluid flow
Learning Outcomes:
Core:
- Implement a velocity Verlet algorithm to simulate a 1D harmonic oscillator [Usage]
- Evaluate the scaling of discretization errors for different numerical integration orders [Assessment]
- Solve the 2D Laplace equation for a given set of boundary potentials using Jacobi iteration [Usage]
5.10.2. CDP/Monte Carlo Methods and Simulations (Core Tier1: 1 hr, Core Tier2: 1 hr) ↑ Back to top
Stochastic sampling techniques for multi-dimensional integration and statistical physics.
Topics:
Core
- Pseudo-random and quasi-random number generation algorithms
- The Metropolis-Hastings algorithm for Markov Chain Monte Carlo (MCMC)
- Monte Carlo simulation of the Ising model and phase transitions
- Importance sampling and variance reduction techniques
- Kinetic Monte Carlo (KMC) for time-dependent stochastic processes
- Statistical error estimation and autocorrelation time in MCMC chains
Learning Outcomes:
Core:
- Estimate the value of \(\pi\) using a hit-or-miss Monte Carlo integration method [Usage]
- Analyze the effect of critical slowing down near the Curie temperature in an Ising simulation [Assessment]
- Develop an importance sampling scheme to integrate a peaked multidimensional function [Usage]
5.10.3. CDP/High-Performance Computing (HPC) for Large-Scale Simulations (Core Tier1: 1 hr, Core Tier2: 1 hr) ↑ Back to top
Architectures and programming models for massive parallelization of physics code.
Topics:
Core
- Shared-memory programming using OpenMP directives
- Distributed-memory systems and the Message Passing Interface (MPI)
- GPU acceleration and GPGPU programming models (CUDA, OpenCL)
- Scalability analysis: Amdahl's Law and Gustafson's Law
- Domain decomposition methods for spatial grid parallelization
- Optimization for cache locality and memory bandwidth limits
Learning Outcomes:
Core:
- Parallelize a large-scale vector addition using OpenMP thread synchronization [Usage]
- Measure the speedup and parallel efficiency of a code across multiple compute nodes [Assessment]
- Implement non-blocking MPI communication for a ghost-cell exchange in a fluid simulation [Usage]
5.10.4. CDP/Machine Learning for Physical Discovery (Core Tier1: 1 hr, Core Tier2: 1 hr) ↑ Back to top
Applying statistical learning to identify patterns and surrogate models in data.
Topics:
Core
- Supervised learning: Linear regression, Lasso, and Ridge for physics models
- Unsupervised learning: Clustering and PCA for detecting phase transitions
- Physical feature engineering and dimensional reduction
- Gaussian Process Regression for uncertainty-aware surrogate modeling
- Symbolic regression for discovering analytical physical laws from data
- Kernel methods for approximating quantum many-body states
Learning Outcomes:
Core:
- Train a regression model to predict the potential energy surface of a molecule [Usage]
- Evaluate model overfitting using cross-validation on a spectroscopic dataset [Assessment]
- Apply Principal Component Analysis (PCA) to visualize high-dimensional spin configurations [Usage]
5.10.5. CDP/Data Analysis and Uncertainty Quantification (Core Tier1: 1 hr, Core Tier2: 1 hr) ↑ Back to top
Statistical frameworks for validating models and quantifying experimental errors.
Topics:
Core
- Bayesian inference: Priors, likelihoods, and posterior distributions
- Maximum Likelihood Estimation (MLE) for parameter fitting
- Linear and non-linear propagation of experimental uncertainties
- Resampling techniques: Bootstrapping and Jackknife methods
- Model selection criteria: AIC, BIC, and Evidence ratios
- Global sensitivity analysis and Sobol indices
Learning Outcomes:
Core:
- Perform a Bayesian parameter fit for a radioactive decay dataset [Usage]
- Calculate confidence intervals for a physical constant derived from noisy measurements [Assessment]
- Compute the Fisher Information matrix to determine the precision limits of a parameter estimate [Usage]
5.10.6. CDP/Modeling Complex Adaptive Systems (Core Tier1: 1 hr, Core Tier2: 1 hr) ↑ Back to top
Dynamics of systems with emergent behavior and non-linear interactions.
Topics:
Core
- Agent-based modeling (ABM) for social and biological systems
- Network science: Scale-free and small-world network dynamics
- Self-organized criticality (SOC) and power-law distributions
- Cellular Automata: Game of Life and fluid flow models (LBM)
- Collective synchronization and the Kuramoto model
Elective
- Evolutionary game theory and ecological stability
Learning Outcomes:
Core:
- Simulate 2D fluid flow around an obstacle using a Lattice Boltzmann method [Usage]
- Analyze the robustness of a scale-free network against targeted attacks [Assessment]
- Model a sandpile to demonstrate the principles of self-organized criticality [Usage]
5.10.7. CDP/Neural Networks for Quantum Systems (Core Tier1: 1 hr, Core Tier2: 1 hr) ↑ Back to top
Representing and solving many-body quantum problems using artificial neural networks.
Topics:
Core
- Restricted Boltzmann Machines (RBM) as Neural Quantum States (NQS)
- Variational Monte Carlo using neural network trial wavefunctions
- Neural network-based quantum state tomography
- CNNs for identifying topological phases in spin lattices
- Autoencoders for compressing many-body wavefunction data
Elective
- Generative models (GANs, VAEs) for sampling quantum configurations
Learning Outcomes:
Core:
- Train an RBM to approximate the ground state of a transverse-field Ising chain [Usage]
- Compare the accuracy of Neural Quantum States against traditional Hartree-Fock methods [Assessment]
- Reconstruct a density matrix from experimental measurement data using a neural network [Usage]
5.10.8. CDP/Physics-Informed Neural Networks (PINNs) (Core Tier1: 1 hr, Core Tier2: 1 hr) ↑ Back to top
Integrating physical laws directly into the loss functions of deep learning architectures.
Topics:
Core
- Incorporating PDEs as residuals in neural network loss functions
- Leveraging automatic differentiation for exact derivative constraints
- Data-efficient learning: Solving inverse problems with sparse data
- Application of PINNs to the Navier-Stokes equations
- Bayesian PINNs for uncertainty quantification in PDE solutions
Elective
- Introduction to Neural Operators (DeepONet) for parametric PDEs
Learning Outcomes:
Core:
- Solve the 1D Burgers' equation using a PINN architecture [Usage]
- Compare the computational cost and accuracy of a PINN versus a traditional Finite Element solver [Assessment]
- Use a PINN to estimate the viscosity coefficient of a fluid from flow visualization data [Usage]