Optimization¶
Quadratic and non-convex cost functions, linear constraints and bounds, wrappers around scipy.optimize.minimize and minimize_scalar, and loss-surface evaluation for plotting.
cost_quadratic(x)
¶
cost_nonconvex(x)
¶
multi_var_cost(x)
¶
2-variable cost function: f(x, y) = (x - 1)^2 + (y - 2.5)^2
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
x
|
array-like of length 2
|
Input vector [x0, x1]. |
required |
Returns:
| Type | Description |
|---|---|
float
|
Scalar cost value. |
Source code in utils/optimization_utils.py
get_linear_constraint()
¶
Build a linear inequality constraint: x + y <= 3.
Returns:
| Type | Description |
|---|---|
LinearConstraint
|
Constraint object compatible with scipy.optimize.minimize. |
Source code in utils/optimization_utils.py
get_bounds_2d()
¶
Create box bounds for a 2D optimization problem: 0 <= x,y <= 5.
Returns:
| Type | Description |
|---|---|
Bounds
|
Bounds object compatible with scipy.optimize.minimize. |
run_minimization(func, x0, bounds=None, constraints=None, method='trust-constr')
¶
Run multivariate minimization using scipy.optimize.minimize.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
func
|
callable
|
Objective function to minimize, taking a 1D array and returning a scalar. |
required |
x0
|
array - like
|
Initial guess for the solution. |
required |
bounds
|
Bounds
|
Variable bounds. |
None
|
constraints
|
constraint or list of constraints
|
Linear or nonlinear constraints. |
None
|
method
|
str
|
Optimization algorithm. Common choices: 'SLSQP', 'L-BFGS-B', 'trust-constr'. |
'trust-constr'
|
Returns:
| Type | Description |
|---|---|
OptimizeResult
|
Result object with attributes: x (solution), fun (min value), success, message. |
Source code in utils/optimization_utils.py
run_scalar_minimization(func, bracket=(0, 5), method='Brent')
¶
Run scalar (1D) minimization using scipy.optimize.minimize_scalar.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
func
|
callable
|
Single-variable objective function. |
required |
bracket
|
tuple of (float, float)
|
Bracket interval for the search. |
(0, 5)
|
method
|
str
|
Scalar minimization method: 'Brent', 'Bounded', or 'Golden'. |
'Brent'
|
Returns:
| Type | Description |
|---|---|
OptimizeResult
|
Result with attributes: x (solution), fun (min value). |
Source code in utils/optimization_utils.py
evaluate_loss_surface(func, x_range=(0, 5), y_range=(0, 5), steps=50)
¶
Evaluate a 2D loss function over a grid for surface visualization.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
func
|
callable
|
Two-variable objective function taking a list/array [x, y]. |
required |
x_range
|
tuple of (float, float)
|
Range for the x-axis grid. |
(0, 5)
|
y_range
|
tuple of (float, float)
|
Range for the y-axis grid. |
(0, 5)
|
steps
|
int
|
Number of grid points along each axis. |
50
|
Returns:
| Type | Description |
|---|---|
tuple of (np.ndarray, np.ndarray, np.ndarray)
|
(X, Y, Z) meshgrid arrays where Z[i,j] = func([X[i,j], Y[i,j]]). |
Examples:
>>> X, Y, Z = evaluate_loss_surface(multi_var_cost, x_range=(0,3), y_range=(0,5))
>>> Z.shape
(50, 50)