DeepXDE v1.7.0

Areas of improvement

• dde.icbc.PointSetBC supports mini-batch
• Bug fix

New APIs

• Backend paddle: Support dde.nn.DeepONet and dde.nn.DeepONetCartesianProd

API changes

• Change dde.callback.PDEResidualResampler to dde.callback.PDEPointResampler

DeepXDE v1.6.2

Areas of improvement

• Set Hammersley as the default point sampling for PINN
• Improve point sampling of dde.geometry.GeometryXTime.random_points
• dde.callback.ModelCheckpoint supports monitoring test loss
• PyTorch backend: dde.nn.PODMIONet and dde.nn.MIONetCartesianProd support multiple merge operations

DeepXDE v1.6.1

Areas of improvement

• Backend TensorFlow 1.x dde.nn.DeepONet supports customized branch
• Fix Triangle.on_boundary for float32
• Bug fix: a few issues of float64
• Many documentation improvements

New APIs

• Backend PyTorch adds dde.nn.PODMIONet
• Backend PyTorch adds dde.nn.MIONetCartesianProd

DeepXDE v1.6.0

Backend JAX supports PDE inverse problems.🎉🎉🎉

Areas of improvement

• Backend PyTorch supports l2 regularization
• Improve XLA jit

New APIs

• Add dde.data.PDEOperatorCartesianProd
• Backend TensorFlow supports dde.nn.PFNN

API changes

• Model.train: epochs is replaced by iterations
• Merge PIDeepONet into DeepONet

DeepXDE v1.5.1

Areas of improvement

• Bug fix: dde.icbc.NeumannBC for dde.geometry.Polygon

New APIs

• Add new geometry dde.geometry.PointCloud

DeepXDE v1.5.0

Areas of improvement

• Support new loss "mean l2 relative error"
• Backend PyTorch supports dde.nn.DeepONetCartesianProd and dde.nn.PODDeepONet
• Backend PyTorch supports "step" LR decay
• dde.callbacks.OperatorPredictor supports backends tensorflow.compat.v1, tensorflow, pytorch, paddle
• Fix dde.config.set_random_seed

New APIs

• Add dde.config.enable_xla_jit and dde.config.disable_xla_jit

API changes

• Change DDEBACKEND to DDE_BACKEND

DeepXDE v1.4.0

DeepXDE supports a new backend PaddlePaddle.🎉🎉🎉

Areas of improvement

• Backend PyTorch supports dde.nn.PFNN
• Backend TensorFlow 1.x enables XLA for GPU

DeepXDE v1.3.1

This is a bugfix release.

Areas of improvement

• Bug fix: NeumannBC and RobinBC

DeepXDE v1.3.0

Backend JAX supports ODE forward problems.🎉🎉🎉

Areas of improvement

• Backend TensorFlow supports model.save and model.restore.
• Backend TensorFlow uses @tf.function(jit_compile=True) for faster speed.

DeepXDE v1.2.0

DeepXDE supports physics-informed DeepONet.🎉🎉🎉

Areas of improvement

• Fix the issue when setting random seed via dde.set_random_seed()

New APIs

• Support physics-informed DeepONet
  • Add function spaces: dde.data.PowerSeries, dde.data.Chebyshev, dde.data.GRF, dde.data.GRF_KL, dde.data.GRF2D
  • Add dde.data.PDEOperator
  • Add dde.nn.PIDeepONet