hdim-opt: High-Dimensional Optimization Toolkit
hdim_opt is a lightweight, comprehensive Python toolkit to streamline numerical optimization, sampling, and analysis of high-dimensional function landscapes and datasets. Home of the QUASAR evolutionary algorithm and hyperellipsoid density sampling.
All core functions, listed below, are single-line executable and depend on three essential parameters: [obj_function, bounds, n_samples]:
Sampling
- uniform: Generate uniform QMC sample sequences (via Scipy.stats.qmc).
- hyperellipsoid: Generate hyperellipsoidal sample sequence; may accelerate optimization.
- isotropize: Isotropize the input data via zero-phase component analysis (ZCA).
- encode_bipolar: Bipolar-logarithmic transform, when negative values and exponents are present.
- lorentzian: Fit a Lorentzian/Cauchy kernel density estimation (KDE) to the data.
Optimization
- quasar: Optimization using the QUASAR evolutionary algorithm.
- minimize: Optimization using gradient-based minimization (via SciPy.minimize).
- symbolic: Symbolic regression to approximate the input data or function (via gplearn).
- stepAIC: Stepwise feature selection for linear or logistic regression (R's MASS:stepAIC).
Analysis
- sensitivity: Sensitivity analysis to quantify each dimension's influence (via SALib).
- hyperslice: Create a hyperslice of the function's underlying solution space.
- waveform: Decompose any 2D waveform.
- analyze: Analyze any input dataset.
Installation
Install hdim_opt directly from PyPI:
pip install hdim_opt
Example Usage
import hdim_opt as h
### Parameter Space
n_samples = 2**10
n_dimensions = 10
bounds = [(-100,100)] * n_dimensions # Parameter bounds
obj_func = h.test_functions.rastrigin # Test function
### Sampling
uniform_samples = h.uniform(n_samples, bounds, method='sobol') # Uniform sampling
ellipsoid_samples = h.hyperellipsoid(n_samples, bounds, verbose=True) # Hyperellipsoid sampling
iso_samples, iso_params = h.isotropize(ellipsoid_samples) # Isotropize data (ZCA)
bipolar_log_samples = h.encode_bipolar(iso_samples, [b[0] for b in bounds]) # Bipolar-logarithm transform
kde = h.lorentzian(iso_samples, 1.0, iso_samples, verbose=True) # Lorentzian multivariate KDE
### Optimization
solution, fitness = h.quasar(obj_func, bounds, init=ellipsoid_samples) # Evolutionary optimization
local_sol, local_fit = h.minimize(obj_func, bounds, init=solution) # Gradient-based optimization
all_expr, best_expr = h.symbolic(obj_func, bounds) # Symbolic regression
opt_features, opt_model = h.stepAIC(iso_samples, solution) # R's stepAIC for linear/logistic regression
### Analysis
Si, S2 = h.sensitivity(obj_func, bounds) # Sensitivity analysis
slice_data, stats = h.hyperslice(obj_func, bounds, slice_dims=(0,1)) # Estimate/hyperslice the solution space
signal_results = h.waveform(uniform_samples[:,0], slice_data.iloc[:,1]) # Analyze 2D waveform
h.analyze(slice_data) # Analyze any numerical dataset
QUASAR Optimization
QUASAR (Quasi-Adaptive Search with Asymptotic Reinitialization) is a quantum-inspired evolutionary algorithm, highly efficient for minimizing high-dimensional, non-differentiable, and non-parametric objective functions.
- Benefit: Significant improvements in convergence speed and solution quality for high-dimensional spaces compared to standard optimization algorithms like Differential Evolution and L-SHADE. (Reference: [https://arxiv.org/abs/2511.13843]).
Hyperellipsoid Sampling
HDS (Hyperellipsoid Density Sampling) is a non-uniform Quasi-Monte Carlo sampling method, specifically designed to exploit promising regions of the parameter space.
- Benefit: Provides control over high-dimensional sample distributions. Results in higher average solution quality when initializing optimization. (Reference: [https://arxiv.org/abs/2511.07836]).
Issues & Support
Bug reports are highly appreciated and will be handled in a timely manner.
1 maintainer
Metadata
1 maintainer
Release files for hdim-opt 1.6.2
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| hdim_opt-1.6.2.tar.gz | 45.0 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| hdim_opt-1.6.2-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 91.1 kB
Release files / hdim_opt-1.6.2.tar.gz
| Download URL | hdim_opt-1.6.2.tar.gz |
|---|---|
| Size | 45.0 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
4486571d3f59fe9eab38f938f50f128aa4c89d5164941e7ec4d09049e012b7d4
|
|
BLAKE2b-256 checksum How to use checksums |
597002ce6b40887a883e7ecc881f9ba641abd7214867c32181f2abd09c2d5604
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.1
|
Release files / hdim_opt-1.6.2-py3-none-any.whl
| Download URL | hdim_opt-1.6.2-py3-none-any.whl |
|---|---|
| Size | 46.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
c33d0fd550164fe454ab5d8a3bc6487bb65c076dfb442040d97b3b366c5a1922
|
|
BLAKE2b-256 checksum How to use checksums |
dcbce445e7d416aeaeeff83c327c1d43bba2defce10ed2345c8eb664f86fc6f0
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.1
|
