Add LRU-backed embedding networks (#1512)

* Before discussing the new way-to-go

* init LRU implementation based on Orvieto et al

* state dim fix

* rough version

* for loop running

* add aggregate_func

* fix aggregate

* start asociative scan

* Oscillator simulation for the test

* add scan

* add scan

* bidirectional running but different results loop and scan

* Doc improvements

* more efficient bidirectional

* fix bidirectional (outputs should also be flippedcd sbi

* Alarm in the building

* Isolated tests improved and running

* LRU API changes

* add whole pipeline test

* Latest doc updates

* fix m

* Refactoring; By default bidriectional; Added test cases

* start scan loop check test

* Undo unintended changes

* Final touch-up

* Update sbi/neural_nets/embedding_nets/lru.py

Co-authored-by: manuelgloeckler <38903899+manuelgloeckler@users.noreply.github.com>

* Fixed type hints

* Use Tuple instead of tuple

* add warning that scan does not support backwards pass yet

* Use lambda_abs property; Add test case

* fix test

* omit some linear weights of the GLU as in Smith, Warrington & Linderman, 2023

* accidentally swapped nonlinearity and dropout -fix

---------

Co-authored-by: Matthijs <matthijs@example.com>
Co-authored-by: Matthijspals <matthijs-pals@hotmail.com>
Co-authored-by: Matthijs Pals <34062419+Matthijspals@users.noreply.github.com>
Co-authored-by: manuelgloeckler <38903899+manuelgloeckler@users.noreply.github.com>

Author: 4 people

sbi/neural_nets/embedding_nets/__init__.py

@@ -1,6 +1,7 @@
 from sbi.neural_nets.embedding_nets.causal_cnn import CausalCNNEmbedding
 from sbi.neural_nets.embedding_nets.cnn import CNNEmbedding
 from sbi.neural_nets.embedding_nets.fully_connected import FCEmbedding
+from sbi.neural_nets.embedding_nets.lru import LRUEmbedding
 from sbi.neural_nets.embedding_nets.permutation_invariant import (
     PermutationInvariantEmbedding,
 )