modify_order_features函数

介绍

modify_order_features函数是gempy.core.model.ImplicitCoKriging类中的一个方法。该函数实现了通过对预测网格的顺序特征进行修改来改进预测结果的功能。该函数是修改函数序列特征的主要接口。

参数

• grid: gempy.core.grid.Grid类型，预测网格对象。
• grads: numpy.ndarray类型，在当前位置处计算出的梯度值。
• values: numpy.ndarray类型，当前位置处的模型值。
• residuals: numpy.ndarray类型，模型残差。

返回值

返回值为修改后的特征函数序列，也是numpy.ndarray类型。

使用方法

1. 从gempy.core.model.ImplicitCoKriging中创建对象。

model = ImplicitCoKriging()

2. 通过set_parameters方法设置模型参数。

model.set_parameters(nn=3, output='continuous', drift_terms=1, c_kwargs={"n_categories":10})

3. 确定预测网格和参数。

resolution = 50
extent = np.array([[-3_000, 3_000], [-3_000, 3_000], [-3_000, 3_000]])
grid = Grid(extent=extent, resolution=resolution)

4. 通过modify_order_features方法修改预测网格的顺序特征。

grads = model.compute_model_gradient_at(grid)
mod_vals, _ = model.block_forward_pass(grid)
residuals = model.residuals(grid, mod_vals)

.modified_features = modify_order_features(
    grid=grid,
    grads=grads,
    values=mod_vals,
    residuals=residuals
)

示例

以下代码演示了如何使用modify_order_features函数修改预测网格的顺序特征。

import gempy as gp

# Create model object
model = gp.create_model('test')

# Add surfaces
gp.add_surfaces(
    model,
    ['rock1', 'rock2', 'rock3', 'rock4'],
    colors=['red', 'green', 'blue', 'yellow'],
)

# Add series
gp.add_series(
    model,
    ['fault', 'sediments'],
    order_series=['fault', 'sediments'],
    add_to_model=True
)

# Add fault
gp.add_faults(model, ['fault1'], series_name='fault', fault_boundaries=[-200, 0], fault_dip=45)

# Add interfaces
gp.set_interpolator(model, theano_optimizer='fast_compile', verbose=['bla'])
gp.set_c_o_type(model)
gp.add_data(model, -1000, 1000, 0, grade=[0], series='fault')

gp.add_data(model, 0, 0, 0, grade=[1], series='sediments')
gp.add_data(model, -500, 500, -2000, grade=[1], series='sediments')
gp.add_data(model, -500, 500, -1000, grade=[2], series='sediments')
gp.add_data(model, -1000, 1000, -1500, grade=[3], series='sediments')

model.surfaces.df['order_surface'] = [-2, -1, 1, 2]
gp.map_series_to_surfaces(model)
gp.set_frequencies(model)

# Define grid and compute model
resolution = 50
extent = np.array([[-3_000, 3_000], [-3_000, 3_000], [-3_000, 3_000]])
grid = gp.Grid(extent=extent, resolution=resolution)

model.set_parameters(nn=3, output='continuous', drift_terms=1, c_kwargs={"n_categories":10})

grads = model.compute_model_gradient_at(grid)
mod_vals, _ = model.block_forward_pass(grid)
residuals = model.residuals(grid, mod_vals)

.modified_features = gp.modify_order_features(
    grid=grid,
    grads=grads,
    values=mod_vals,
    residuals=residuals
)

# Plot results
gp.plot_3d(model)