CHANGELOG.md

Changelog

v0.28.5

• Fixes some issues with threading.

v0.28.4

• Improves speed of SMC^2 when we require an increase in state particles by checking acceptance rate threshold in kernel instead.

v0.28.3

• Check average acceptance rate for MH kernel instead.

v0.28.2

• Bug fix for matrix shaped parameters.

v0.28.1

• Use torch.Tensor.copy_ instead of fill_.

v0.28.0

• Uses stoch-proc version 0.3.0, which means that whenever the context is updated in an inference algorithm we rebuild the model.
• Speed improvement for SISR.

v0.27.0

• Major change in backend by interchaning batch and sample shape in inference algorithms. This enables much more natural handling of parameters.
• Adds support for finding mode of distribution by means of `functorch

v0.26.1

• Adds the Gaussian Particle Filter
• Reworks correct/predict logic by enforcing that correct only takes a prediction as input
• Simplifies proposal logic
• Uses latest version of stoch-proc

v0.26.0

• Reworks how we define parameters by initializing all parameters to their "correct" shapes from the get-go by introducing a shape object stored on InferenceContext.

v0.25.1

• Reworks ParameterContext logic by removing the requirements for the context being on the stack in order to register parameters.
• Adds fixed-lag smoothing.
• Renames ParameterContext to InferenceContext

v0.24.11

• Adds support for using QMC points in SMC2

v0.24.10

• Utilizes int instead of long

v0.24.5

• Adds a proposal for locally linearized observation dynamics.

v0.24.4

• Adds support for performing variational inference using particle filter.

v0.24.3

• Improves plotting functionality by handling multi-dimensional parameters.

v0.24.2

• Adds support for applying function on context and return a copy of that context.