DeepXDE v0.8.6

Areas of improvement

• Model.save() supports pickle
• PDE
  • supports pde as None
  • supports bcs as the empty list []
  • accepts train_distribution = 'uniform' or 'pseudo' or 'sobol'
• OpNN supports user-defined branch net
• OpDataSet supports mini-batch training

API changes

• Rename Model._print_model() to Model.print_model()

New APIs

• Add Model.state_dict()
• Add PDE.resample_train_points()

DeepXDE v0.8.5

Areas of improvement

• Add Triangle.on_boundary and Triangle.boundary_normal
• OpNN supports different activations for branch and trunk nets
• Bug fix: GeometryXTime.uniform_points

DeepXDE v0.8.4

Areas of improvement

• Support numpy inputs to define PDE
• Bug fix: Hypershpere sobol points

New APIs

• Add feed-forward neural networks (PFNN) that support sub-networks
• Add fractional PDE (FPDE and TimeFPDE) with fractional Laplacian

DeepXDE v0.8.3

Areas of improvement

• Bug fix: Cuboid.random_boundary_points()

DeepXDE v0.8.2

Areas of improvement

• OpNN can be partially trained
• DirichletBC supports functions that returns a number
• DataSet does not standardize by default
• Add Swish activation
• Bug fix in loss plot

DeepXDE v0.8.1

Areas of improvement

• Add loss: mean absolute error
• Bug fix: Geometry point sampling

DeepXDE v0.8.0

Areas of improvement

• Support layer-wise locally adaptive activation function
• Bug fix

API changes

• Rename Map.outputs_modify() to Map.apply_output_transform()

New APIs

• Add Map.apply_feature_transform()

DeepXDE v0.7.0

Areas of improvement

• Support both TensorFlow 1 and TensorFlow 2 as backend
• Check if a Polygon is a Triangle or Rectangle

DeepXDE v0.6.1

Areas of improvement

• Bug fix: mean_absolute_percentage_error allows y_true to be negative

New APIs

• Add new metric: max_absolute_percentage_error

DeepXDE v0.6.0

Areas of improvement

• Fix the bug of float64 support

API changes

• PDE, TimePDE, IDE: remove argument num_outputs, rename argument func to solution
• PeriodicBC supports first order derivative