### Install PolyChord dependency

Source: https://gvcallen.github.io/paramrf/index.html

Install the PolyChord solver and associated dependencies.

```bash
$ pip install git+https://github.com/PolyChord/PolyChordLite.git anesthetic mpi4py
```

--------------------------------

### start

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Frequency.html

The starting frequency in Hz.

```APIDOC
## property start

### Description
The starting frequency in Hz.

### Returns
- **float** - Start frequency.
```

--------------------------------

### BulkConductor surface impedance example

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.materials.BulkConductor.html

Creates a BulkConductor with copper conductivity and evaluates its surface impedance over a frequency range. The example shows the required imports and the call to properties(freq) to obtain the surface impedance zs.

```python
import pmrf as prf
from pmrf.materials import BulkConductor

copper = BulkConductor(sigma=5.8e7)
freq = prf.Frequency(start=1, stop=10, npoints=101, unit='ghz')
zs = copper.properties(freq).zs
```

--------------------------------

### Frequency constructor examples

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Frequency.html

Two examples of constructing Frequency objects: one with start/stop/npoints and unit, and one with a log sweep.

```python
>>> wr1p5band = Frequency(start=500, stop=750, npoints=401, unit='ghz')
>>> logband = Frequency(1, 1e9, 301, sweep_type='log')
```

--------------------------------

### Declaring a parameter with constraints and scaling

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.param.html

This example demonstrates how to define a model with parameters using constraints and scaling. It shows successful initialization and a ValueError when a constraint is violated.

```python
import pmrf as prf
from pmrf.constraints import Positive

class RC(prf.Module):
    R: prf.Param = prf.param(constraint=Positive())
    C: prf.Param = prf.param(constraint=Positive(), scale=1e-12)

RC(1.0, 2.0)
# RC(R=1., C=2.e-12)

RC(-1.0, 2.0)
# ValueError: out of bounds
```

--------------------------------

### Benchmark Output Example

Source: https://gvcallen.github.io/paramrf/_sources/skrf_comparison/performance.rst.txt

Example output showing the performance difference between scikit-rf and ParamRF.

```text
scikit-rf forward pass: 39.359 ms
ParamRF forward pass:    1.197 ms
Speedup:                 32.9x
```

--------------------------------

### Initialize and compute S-parameters for a MicrostripLine

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.models.components.lines.microstrip.MicrostripLine.html

This example demonstrates how to instantiate a MicrostripLine with specific physical dimensions and material properties, then compute its S-parameters over a defined frequency range.

```python
import pmrf as prf
from pmrf.models import MicrostripLine
from pmrf.materials import BulkConductor, ConstantDielectric

phys_microstrip = MicrostripLine(
    w=4e-3,
    h=2.0e-3,
    dielectric=ConstantDielectric(ep_r=4.6, tand=0.025),
    conductor=BulkConductor(sigma=5.8e7),
    length=0.5
)

freq = prf.Frequency(start=1, stop=20, npoints=101, unit='ghz')
s_phys = phys_microstrip.s(freq)
```

--------------------------------

### Using Substrate with AbstractBuilder

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.materials.Substrate.html

Shows how to define a model with a shared Substrate and inject it into multiple MicrostripLine instances. The example creates a Board class with two line widths and a shared substrate, then lists the named parameters ending with 'ep_r'.

```python
>>> import pmrf as prf
>>> from pmrf.materials import Substrate
>>> from pmrf.models import AbstractBuilder, MicrostripLine
>>> class Board(AbstractBuilder):
...     substrate: Substrate
...     w1: prf.Param
...     w2: prf.Param
...     def build(self):
...         return (MicrostripLine(w=self.w1, substrate=self.substrate, length=0.1)
...              ** MicrostripLine(w=self.w2, substrate=self.substrate, length=0.2))
>>> board = Board(substrate=Substrate(h=1.6e-3, dielectric=4.3), w1=1e-3, w2=2e-3)
>>> [name for name in board.named_params() if name.endswith("ep_r")]
['substrate.dielectric.ep_r']
```

--------------------------------

### Import required packages for the tutorial

Source: https://gvcallen.github.io/paramrf/tutorials/1_cable_fitting.html

Imports matplotlib for plotting, skrf for touchstone reading, and pmrf (ParamRF). Ensure these packages are installed and import correctly before proceeding.

```python
import matplotlib.pyplot as plt     # for plotting
import skrf as rf                   # for touchstone reading
import pmrf as prf                  # ParamRF
```

--------------------------------

### Create CoaxialLine and compute S-parameters

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.models.components.lines.coaxial.CoaxialLine.html

Example from the class docstring: imports pmrf, CoaxialLine, BulkConductor, and ConstantDielectric; creates a CoaxialLine with d_in=0.9e-3, d_out=2.95e-3, dielectric with ep_r=1.5 and tand=0.0004, conductor with sigma=5.8e7, and length=0.5; then defines a frequency range from 1 to 20 GHz with 101 points and computes S-parameters using the s method.

```python
import pmrf as prf
from pmrf.models import CoaxialLine
from pmrf.materials import BulkConductor, ConstantDielectric

phys_cable = CoaxialLine(
    d_in=0.9e-3,
    d_out=2.95e-3,
    dielectric=ConstantDielectric(ep_r=1.5, tand=0.0004),
    conductor=BulkConductor(sigma=5.8e7),
    length=0.5
)

freq = prf.Frequency(start=1, stop=20, npoints=101, unit='ghz')
s_phys = phys_cable.s(freq)
```

--------------------------------

### Install distreqx dependency

Source: https://gvcallen.github.io/paramrf/index.html

Install the required fork of distreqx for Bayesian inference.

```bash
$ pip install git+https://github.com/gvcallen/distreqx
```

--------------------------------

### Install BlackJAX dependency

Source: https://gvcallen.github.io/paramrf/index.html

Install the BlackJAX Bayesian solver.

```bash
$ pip install git+https://github.com/handley-lab/blackjax.git@v0.1.0-beta
```

--------------------------------

### Create a StriplineLine and compute S-parameters

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.models.components.lines.stripline.StriplineLine.html

Shows how to instantiate a StriplineLine with custom geometry and materials, then compute its S-parameters over a frequency range. The example uses ConstantDielectric and BulkConductor for the dielectric and conductor, respectively.

```python
import pmrf as prf
from pmrf.models import StriplineLine
from pmrf.materials import BulkConductor, ConstantDielectric

line = StriplineLine(
    w=2.655e-3,
    b=3.2e-3,
    t=35e-6,
    dielectric=ConstantDielectric(ep_r=2.2, tand=0.001),
    conductor=BulkConductor(sigma=5.8e7),
    length=0.1,
)

freq = prf.Frequency(start=1, stop=20, npoints=101, unit='ghz')
s = line.s(freq)
```

--------------------------------

### Define a custom model expansion

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Model.html

This example demonstrates how to implement an expand method for a custom 2-port model. It maps external ports to internal components and defines the netlist connections.

```python
>>> def expand(self):
...     # 1. Grab internal components
...     L, C = self.inductor, self.capacitor
...
...     # 2. Map our external ports to the internal components
...     port_mapping = [
...         (L, 0),  # External port 0 maps to Inductor port 0
...         (C, 1)   # External port 1 maps to Capacitor port 1
...     ]
...
...     # 3. Define the internal connections (the netlist)
...     # Connect Inductor port 1 to Capacitor port 0
...     internal_connections = [
...         [(L, 1), (C, 0)]
...     ]
...
...     return port_mapping, internal_connections
```

--------------------------------

### Model Setup for Derivatives and Sweeps

Source: https://gvcallen.github.io/paramrf/_sources/examples/derivatives_and_sweeps.rst.txt

Initializes a low-pass filter model with unconstrained parameters for use in derivative and sweep operations.

```python
import pmrf as prf
from pmrf.models import ShuntCapacitor, Inductor, Cascade

c1 = ShuntCapacitor(C=prf.Unconstrained(1.0e-12), name='c1')
l1 = Inductor(L=prf.Unconstrained(1.0e-9), name='l1')
c2 = ShuntCapacitor(C=prf.Unconstrained(1.0e-12), name='c2')

lpf = Cascade([c1, l1, c2])
```

--------------------------------

### Example for converter

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.field.html

Shows how to use pmrf.field with a converter to convert a value to a JAX array on initialization. The converter is applied after __init__ or __post_init__, and the example asserts the field becomes a jax.Array.

```python
class MyModule(eqx.Module):
    foo: Array = eqx.field(converter=jax.numpy.asarray)

mymodule = MyModule(1.0)
assert isinstance(mymodule.foo, jax.Array)
```

--------------------------------

### from_f class method example

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Frequency.html

Shows how to create a Frequency object from a frequency vector using the from_f class method.

```python
>>> f = np.linspace(75,100,101)
>>> rf.Frequency.from_f(f, unit='GHz')
```

--------------------------------

### Example for static

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.field.html

Shows how to use pmrf.field with static=True to make a field part of the PyTree structure rather than a leaf, so it does not interact with JAX transforms. The example flattens the module and asserts that only the non-static field is a leaf.

```python
class MyModule(eqx.Module):
    normal_field: int
    static_field: int = eqx.field(static=True)

mymodule = MyModule("normal", "static")
leaves, treedef = jax.tree_util.tree_flatten(mymodule)
assert leaves == ["normal"]
assert "static" in str(treedef)
```

--------------------------------

### Create a PhaseLine instance

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.models.components.lines.ideal.PhaseLine.html

Demonstrates how to instantiate a PhaseLine with a 90-degree electrical length and 50-ohm characteristic impedance at 1 GHz, then calculate its S-parameters over a frequency range.

```python
import pmrf as prf
from pmrf.models import PhaseLine

# Create an ideal 90-degree (quarter-wave) 50-ohm line at 1 GHz
quarter_wave = PhaseLine(
    z0=50.0,
    theta=90.0,
    f0=1e9
)

freq = prf.Frequency(start=0.5, stop=1.5, npoints=101, unit='ghz')
s = quarter_wave.s(freq)
```

--------------------------------

### Manipulate module parameters with at()

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Module.html

Use the at() method to create a lens for surgical inspection or replacement of module parameters. This example demonstrates getting a parameter value and setting a new parameter object.

```python
>>> import pmrf as prf
>>> from pmrf.models import Resistor
>>> module = Resistor(R=50.0, name="res")
>>> module.at("res.R").get()
50.0
>>> updated = module.at("res.R").set(prf.Unconstrained(100.0))
```

--------------------------------

### Initialize and use DatasheetLine

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.models.components.lines.empirical.DatasheetLine.html

Demonstrates how to instantiate a DatasheetLine with specific impedance, velocity factor, and loss coefficients, then calculate scattering parameters over a frequency range.

```python
import pmrf as prf
from pmrf.models import DatasheetLine

cable = DatasheetLine(
    zn=50.0,
    vf=0.69,  # Velocity factor (e.g., solid PTFE)
    k1=0.2,   # Skin effect loss factor
    k2=0.01,  # Dielectric loss factor
    length=1.0
)

freq = prf.Frequency(start=0.1, stop=10, npoints=201, unit='ghz')
s = cable.s(freq)
```

--------------------------------

### run(logposterior_fn, y0, args, key, init_samples=None, max_steps=None, **kwargs)

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.infer.AbstractJointSampler.html

Execute the sampling algorithm. This method takes a log-posterior function, initial parameters, arguments, a random key, and optional initial samples and maximum steps, and returns a SampleResult.

```APIDOC
## run(logposterior_fn, y0, args, key, init_samples=None, max_steps=None, **kwargs)

### Description
Execute the sampling algorithm.

### Method
run

### Signature
`abstractmethod run(logposterior_fn: Callable[[PyTree, Any], Any], y0: PyTree, args: PyTree[Any], key: Array, init_samples: PyTree | None = None, max_steps: int | None = None, **kwargs) -> SampleResult`

### Parameters
- **logposterior_fn** (callable) - Required - A function taking the parameters and args as input and returning the log-posterior probability.
- **y0** (PyTree) - Required - The initial parameters, either for shape reference or as a starting point.
- **args** (Any) - Required - Args to pass to fn.
- **key** (Array) - Required - A random JAX key.
- **init_samples** (PyTree, optional) - Optional - An optional batched PyTree the same structure as y0 with initial samples to warm-start the algorithm.
- **max_steps** (int, optional) - Optional - The maximum number of sampling steps to take. If None, implies there should be no limit.
- **kwargs** - Optional - Runtime arguments forward to the solver backend.

### Returns
- **results** (SampleResult) - An instance of `pmrf.infer.SampleResult`.
```

--------------------------------

### Compute derivative of a model's response with respect to parameters

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.derivative.html

Shows how to use pmrf.derivative to compute the sensitivity of a component's response (S21 magnitude) with respect to its parameters. The example creates a ShuntCapacitor model, defines an eval_fn that returns the S-matrix magnitude at a specific frequency, and then calls derivative to get the derivative. The derivative preserves the structural layout of the model, allowing access via .at('c1.C').get(). The output is -1.060e-01.

```python
>>> import pmrf as prf
>>> from pmrf.models import ShuntCapacitor
>>>
>>> freq = prf.Frequency(2.4, 2.4, 1, 'GHz')
>>> cap = ShuntCapacitor(C=prf.Unconstrained(1.0e-12), name='c1')
>>>
>>> def eval_s21(model):
...     return model.s_mag(freq)[0, 1, 0]
...
>>> (d_cap,) = derivative(eval_s21, cap)
>>>
>>> # The structural layout of the model is preserved in the derivative
>>> print(f"{d_cap.at('c1.C').get():.3e}")
-1.060e-01
```

--------------------------------

### sample(_key: Key[jaxlib._jax.Array, '']) -> Array

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.distributions.Normal.html

Draws a sample from the distribution.

```APIDOC
## sample(_key: Key[jaxlib._jax.Array, '']) -> Array

### Description
Draws a sample from the distribution. See Distribution.sample.
```

--------------------------------

### Install ParamRF via pip

Source: https://gvcallen.github.io/paramrf/index.html

Install the ParamRF library using pip. Requires Python 3.11 or higher.

```bash
$ pip install paramrf
```

--------------------------------

### Sampling from a Transformed distribution

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.distributions.Transformed.html

Demonstrates how to initialize a Transformed distribution using a base distribution and a bijector, then generate samples.

```python
dist = distrax.Normal(loc=0., scale=1.) bij = distrax.ScalarAffine(shift=jnp.asarray([3., 3., 3.])) transformed_dist = distrax.Transformed(distribution=dist, bijector=bij) samples = transformed_dist.sample(jax.random.key(0)) print(samples) # [2.7941577, 2.7941577, 2.7941577]
```

--------------------------------

### start_scaled

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Frequency.html

The starting frequency in the specified unit.

```APIDOC
## property start_scaled

### Description
The starting frequency in the specified unit.

### Returns
- **float** - Start frequency in the specified unit.
```

--------------------------------

### span

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Frequency.html

The frequency span (stop - start) in Hz.

```APIDOC
## property span

### Description
The frequency span (stop - start) in Hz.

### Returns
- **float** - Span in Hz.
```

--------------------------------

### span_scaled

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Frequency.html

The frequency span (stop - start) in the specified unit.

```APIDOC
## property span_scaled

### Description
The frequency span (stop - start) in the specified unit.

### Returns
- **float** - Span in units.
```

--------------------------------

### DatasheetLine

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.models.components.lines.empirical.DatasheetLine.html

Class definition and constructor for DatasheetLine, including parameters and an example usage.

```APIDOC
## DatasheetLine

### Description
Line defined by nominal impedance, velocity factor, and loss coefficients.

### Class
`pmrf.models.components.lines.empirical.DatasheetLine`

### Constructor
`DatasheetLine(_length_, _zn=50.0_, _vf=1.0_, _k1=0.0_, _k2=0.0_, _loss_coeffs_normalized=False_, *, _name=None_, _metadata=None_)`

### Parameters
- **length** (Any) - Required - Length of the line.
- **zn** (Param, default=50.0) - Nominal characteristic impedance.
- **vf** (Param, default=1.0) - Velocity factor (ratio of propagation speed to the speed of light).
- **k1** (Param, default=0.0) - Skin effect loss factor.
- **k2** (Param, default=0.0) - Dielectric loss factor.
- **loss_coeffs_normalized** (bool, default=False) - If true, use `k1` and `k2` directly. Otherwise, normalize them to the 100 MHz reference convention.
- **name** (str | None, optional) - Name of the line.
- **metadata** (Any, optional) - Additional metadata.

### Example
```python
import pmrf as prf
from pmrf.models import DatasheetLine

cable = DatasheetLine(
    zn=50.0,
    vf=0.69,  # Velocity factor (e.g., solid PTFE)
    k1=0.2,   # Skin effect loss factor
    k2=0.01,  # Dielectric loss factor
    length=1.0
)

freq = prf.Frequency(start=0.1, stop=10, npoints=201, unit='ghz')
s = cable.s(freq)
```

### Mathematical Formulation
With normalized coefficients \(k_{1,norm}\) and \(k_{2,norm}\):

\[
\alpha_c = k_{1,norm} \cdot \frac{\ln(10)}{20} \cdot \sqrt{\omega}
\]

\[
\alpha_d = k_{2,norm} \cdot \frac{\ln(10)}{20} \cdot \omega
\]

Per-unit-length parameters:

\[
R = 2 z_n \alpha_c
\]

\[
L = \frac{z_n}{v_f c}
\]

\[
G = \frac{2 \alpha_d}{z_n}
\]

\[
C = \frac{1}{z_n v_f c}
\]
```

--------------------------------

### Benchmark Execution Entry Point

Source: https://gvcallen.github.io/paramrf/_sources/skrf_comparison/performance.rst.txt

Main execution block to run the defined benchmarks.

```python
if __name__ == "__main__":
    run_forward_pass_benchmark()
    print()
    run_optimization_benchmark()
```

--------------------------------

### __getitem__

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.NetworkCollection.html

Gets a network by index or name. If slice, returns a new NetworkCollection.

```APIDOC
## __getitem__

### Description
Gets a network by index or name. If slice, returns a new NetworkCollection.

### Method
Operator overload

### Signature
`__getitem__(key: int | str | slice) -> Network | NetworkCollection`

### Parameters
- **key** (int, str, or slice) - If int, returns the network at that index. If str, returns the network with that name. If slice, returns a new NetworkCollection.

### Return Type
Network or NetworkCollection
```

--------------------------------

### run(loglikelihood_fn, prior_transform_fn, u0, args, key, init_cube_samples=None, max_steps=None, **kwargs)

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.infer.AbstractHypercubeSampler.html

Executes the sampling algorithm for the AbstractHypercubeSampler class.

```APIDOC
## run(loglikelihood_fn, prior_transform_fn, u0, args, key, init_cube_samples=None, max_steps=None, **kwargs)

### Description
Execute the sampling algorithm.

### Method
run

### Parameters
- **loglikelihood_fn** (callable) - Required - A function taking the physical parameters and args as input and returning the log-likelihood.
- **prior_transform_fn** (callable) - Required - A function taking the hypercube parameters and args as input and returning the physical parameters.
- **u0** (PyTree) - Required - The initial parameters in the unit hypercube, either for shape reference or as a starting point.
- **args** (Any) - Required - Args to pass to fn.
- **key** (Array) - Required - A random JAX key.
- **init_cube_samples** (PyTree, optional) - Optional - An optional batched PyTree the same structure as u0 with initial hypercube samples to warm-start the algorithm.
- **max_steps** (int, optional) - Optional - The maximum number of sampling steps to take. If None, implies there should be no limit.
- **kwargs** - Optional - Runtime arguments forward to the solver backend.

### Returns
An instance of `pmrf.infer.SampleResult`.
```

--------------------------------

### Optimization Convergence Warning

Source: https://gvcallen.github.io/paramrf/tutorials/1_cable_fitting.html

Example of a UserWarning output indicating that the optimization process failed to converge.

```text
/home/gvcallen/Masters/Python/paramrf/pmrf/optimize/base.py:204: UserWarning: Optimization failed to converge. Trying increasing the maximum number of iterations or loosening the solver tolerances.
  warnings.warn("Optimization failed to converge. Trying increasing the maximum number of iterations or loosening the solver tolerances.")
```

--------------------------------

### Download and load measurement data from .s2p file

Source: https://gvcallen.github.io/paramrf/tutorials/1_cable_fitting.html

Downloads a sample .s2p file from the ParamRF repository, saves it locally, loads it into a scikit-rf Network with frequency in GHz, filters frequencies below 10 GHz, and plots the S-parameter data. The original file is available at https://www.minicircuits.com/WebStore/dashboard.html?model=CBN-1.5FT-SMSM%2B.

```python
import os
import urllib.request
import skrf as rf
import matplotlib.pyplot as plt

base_url = "https://raw.githubusercontent.com/gvcallen/paramrf/main/docs/tutorials/data/"
file_name = "CBN-1.5FT-SMSM.s2p"
full_url = base_url + file_name
local_dir = "data"
local_filename = os.path.join(local_dir, file_name)

os.makedirs(local_dir, exist_ok=True)
if not os.path.exists(local_filename):
    print(f"Downloading {file_name} to {local_dir}/...")
    urllib.request.urlretrieve(full_url, local_filename)

# Load the measured data into skrf using the new local path
measured = rf.Network(local_filename, f_unit='GHz')
measured = measured[measured.frequency.f_scaled < 10]
fig, ax = plt.subplots(figsize=(4, 3))
measured.plot_s_db(m=0, n=0, ax=ax)
```

--------------------------------

### Load constructor

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.models.components.ideal.Load.html

Creates an ideal N-port load model with a specified reflection coefficient and reference impedance.

```APIDOC
## Load(_gamma: Any, _z0: Any = 50.0, _nports: int = 1, *, _name: str | None = None, _metadata: Any = None)

### Description
Initializes an ideal N-port load model defined by its reflection coefficient gamma, referenced to impedance z0.

### Parameters
- **gamma** (Any) - Required - The reflection coefficient.
- **z0** (Any) - Optional - The reference impedance at which gamma is defined. Default is 50.0.
- **nports** (int) - Optional - Number of ports. Default is 1.
- **name** (str | None) - Optional - Name of the model.
- **metadata** (Any) - Optional - Additional metadata.

### Attributes
- **gamma** (Param) - The reflection coefficient.
- **nports** (int) - Number of ports.
- **z0** (Param) - The reference impedance at which gamma is defined.
```

--------------------------------

### Optimize with Multiple Goals

Source: https://gvcallen.github.io/paramrf/_sources/examples/model_optimization.rst.txt

Demonstrates multi-objective optimization by defining both a passband matching goal and a stopband rejection goal.

```python
stopband = prf.Frequency(1, 2, 101, 'GHz')

result = minimize([
    (Goal('s11_db', '<', -20), passband),
    (Goal('s21_db', '<', -40), stopband),
], lpf, solver=ScipyMinimize(method='L-BFGS-B'))
```

--------------------------------

### Construct the Circuit model

Source: https://gvcallen.github.io/paramrf/_sources/examples/circuit_models.rst.txt

Initialize the Circuit model using the previously defined connections list.

```python
from pmrf.models import Circuit
pi_clc = Circuit(connections)
```

--------------------------------

### Define a Capacitor Model using ABCD parameters

Source: https://gvcallen.github.io/paramrf/_sources/examples/custom_models.rst.txt

This example demonstrates defining a custom Capacitor model by inheriting from pmrf.Model and implementing the ABCD parameter method 'a'.

```python
import pmrf as prf
import jax.numpy as jnp

class Capacitor(prf.Model):
    C: prf.Param

    def a(self, freq: prf.Frequency) -> jnp.ndarray:
        w, C = freq.w, self.C
        ones, zeros = jnp.ones_like(w), jnp.zeros_like(w)

        return jnp.array([
            [ones,  1.0 / (1j * w * C)],
            [zeros, ones]
        ]).transpose(2, 0, 1)
```

--------------------------------

### Declare a parameter with constraint and scale in ParamRF

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.parameters.param.html

Example of using `prf.param` to declare parameters with constraints and scale in a custom `prf.Module`. Shows that out-of-bounds values raise a `ValueError`.

```python
import pmrf as prf
from pmrf.constraints import Positive

class RC(prf.Module):
    R: prf.Param = prf.param(constraint=Positive())
    C: prf.Param = prf.param(constraint=Positive(), scale=1e-12)

RC(1.0, 2.0)
# RC(R=1., C=2.e-12)

RC(-1.0, 2.0)
# ValueError: out of bounds

```

--------------------------------

### Create ConstantDielectric and evaluate properties

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.materials.ConstantDielectric.html

Shows how to instantiate a ConstantDielectric with ep_r=4.3 and tand=0.02, create a frequency axis from 1 to 10 GHz, and compute the complex permittivity.

```python
import pmrf as prf
from pmrf.materials import ConstantDielectric

fr4 = ConstantDielectric(ep_r=4.3, tand=0.02)
freq = prf.Frequency(start=1, stop=10, npoints=101, unit='ghz')
eps = fr4.properties(freq).ep_r
```

--------------------------------

### Frequency class properties

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.Frequency.html

Properties of the Frequency class that return frequency-related values. These are read-only attributes that provide the start, stop, step, scaled values, and angular frequency vector.

```APIDOC
## Frequency class properties

### Description
These properties return frequency-related values for the Frequency class. They are read-only attributes that provide the start, stop, step, scaled values, and angular frequency vector.

### Properties

#### start
- **Type**: float
- **Description**: Scaled start frequency.

#### step
- **Type**: float
- **Description**: The frequency step size in Hz for evenly spaced sweeps.

#### step_scaled
- **Type**: float
- **Description**: The frequency step size in the specified unit for evenly spaced sweeps.

#### stop
- **Type**: float
- **Description**: The stop frequency in Hz.

#### stop_scaled
- **Type**: float
- **Description**: The stop frequency in the specified unit.

#### w
- **Type**: jnp.ndarray
- **Description**: The angular frequency vector in radians/s. Angular frequency is defined as omega = 2 * pi * f.

### Attributes

#### unit
- **Type**: str
- **Description**: The frequency unit. Possible options are "Hz", "kHz", "MHz", "GHz" or "THz".

### Returns
Each property returns the value as described above.
```

--------------------------------

### GradientDescent.__init__

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.optimize.GradientDescent.html

Constructor for the GradientDescent optimizer. Initializes the learning rate and termination tolerances.

```APIDOC
## GradientDescent.__init__

### Description
Initializes a GradientDescent optimizer with the given learning rate and termination tolerances.

### Method
Constructor

### Parameters
- **learning_rate** (float) - Required - Step size for the gradient descent updates.
- **fatol** (float, default=1e-7) - Optional - Absolute tolerance of the function value for termination.
- **step_atol** (float, default=1e-6) - Optional - Absolute tolerance of the gradients and step sizes for termination.
- **step_rtol** (float, default=1e-6) - Optional - Relative tolerance of the gradients and step sizes for termination.
- **norm** (Callable[[PyTree], Scalar], default=optx.max_norm) - Optional - Norm function used to evaluate the error.
```

--------------------------------

### run(loglikelihood_fn, logprior_fn, y0, args, key, init_samples=None, max_steps=None, **kwargs)

Source: https://gvcallen.github.io/paramrf/api/generated/pmrf.infer.AbstractSplitSampler.html

Executes the sampling algorithm for samplers that require separate likelihood and prior densities.

```APIDOC
## run(loglikelihood_fn, logprior_fn, y0, args, key, init_samples=None, max_steps=None, **kwargs)

### Description
Execute the sampling algorithm. This method is an abstract interface for samplers that require separate likelihood and prior densities, such as modern Nested Sampling implementations.

### Method
run

### Parameters
- **loglikelihood_fn** (callable) - Required - A function taking the parameters and args as input and returning the log-likelihood.
- **logprior_fn** (callable) - Required - A function taking the parameters and args as input and returning the log prior probability.
- **y0** (PyTree) - Required - The initial parameters, either for shape reference or as a starting point.
- **args** (Any) - Required - Args to pass to fn.
- **key** (Array) - Required - A random JAX key.
- **init_samples** (PyTree, optional) - Optional - An optional batched PyTree the same structure as y0 with initial samples to warm-start the algorithm.
- **max_steps** (int, optional) - Optional - The maximum number of sampling steps to take. If None, implies there should be no limit.
- **kwargs** - Optional - Runtime arguments forward to the solver backend.

### Returns
An instance of `pmrf.infer.SampleResult`.
```