networks

A rich collection of neural network architectures for use in Approximators.

The module features inference networks (IN), summary networks (SN), as well as general purpose networks.

Modules

diffusion_model

Classes

ConsistencyModel(*args, **kwargs)	(IN) Implements a Consistency Model with Consistency Training (CT) as described in [1-2].
CouplingFlow(*args, **kwargs)	(IN) Implements a coupling flow as a sequence of dual couplings with permutations and activation normalization.
DeepSet(*args, **kwargs)	(SN) Implements a deep set encoder introduced in [1] for learning permutation-invariant representations of set-based data, as generated by exchangeable models.
DiffusionModel(*args, **kwargs)	Score-based diffusion model for simulation-based inference as described in [1]:
FlowMatching(*args, **kwargs)	(IN) Implements Optimal Transport Flow Matching, originally introduced as Rectified Flow, with ideas incorporated from [1-5].
FusionNetwork(*args, **kwargs)	(SN) Wraps multiple summary networks (backbones) to learn summary statistics from multi-modal data.
FusionTransformer(*args, **kwargs)	(SN) Implements a more flexible version of the TimeSeriesTransformer that applies a series of self-attention layers followed by cross-attention between the representation and a learnable template summarized via a recurrent net.
InferenceNetwork(*args, **kwargs)
MLP(*args, **kwargs)	Implements a simple configurable MLP with optional residual connections and dropout.
PointInferenceNetwork(*args, **kwargs)	Implements point estimation for user specified scoring rules by a shared feed forward architecture with separate heads for each scoring rule.
Sequential(*args, **kwargs)	A custom sequential model for managing a sequence of Keras layers.
SetTransformer(*args, **kwargs)	(SN) Implements the set transformer architecture from [1] which ultimately represents a learnable permutation-invariant function.
StableConsistencyModel(*args, **kwargs)	(IN) Implements an sCM (simple, stable, and scalable Consistency Model) with continuous-time Consistency Training (CT) as described in [1].
SummaryNetwork(*args, **kwargs)
TimeMLP(*args, **kwargs)	Implements a time-conditioned multi-layer perceptron (MLP).
TimeSeriesNetwork(*args, **kwargs)	(SN) Implements a LSTNet Architecture as described in [1]
TimeSeriesTransformer(*args, **kwargs)	(SN) Creates a regular transformer coupled with Time2Vec embeddings of time used to flexibly compress time series.