BAND-torch: Behavior-aligned neural dynamics model

The code in this repository extends lfads-torch by Sedler et. al: A modular and extensible implementation of latent factor analysis via dynamical systems

BAND is a latent dynamics model weakly supervised with behavior. Using a well-established latent dynamics model (LFADS) as a baseline, we constructed a model that not only explains neural variability but also aligns the latent space with the behavioral output (BAND schema, green is new). While LFADS is capable of inferring inputs through the controller RNN, this is only the case when these inputs cause a significant change in future dynamics and affect the neural reconstruction. To ensure that behavior-related inputs can be captured even if they cause a small, transient change in neural dynamics, we utilize an additional behavior decoder (Fig.~\ref{fig:fig1}b, green).

Latent factor analysis via dynamical systems (LFADS) is a variational sequential autoencoder that achieves state-of-the-art performance in denoising high-dimensional neural spiking activity for downstream applications in science and engineering [1, 2, 3, 4]. Recently introduced variants have continued to demonstrate the applicability of the architecture to a wide variety of problems in neuroscience [5, 6, 7, 8]. Since the development of the original implementation of LFADS, new technologies have emerged that use dynamic computation graphs [9], minimize boilerplate code [10], compose model configuration files [11], and simplify large-scale training [12]. Building on these modern Python libraries, we introduce band-torch — a new open-source implementation of LFADS designed to be easier to understand, configure, and extend.

Installation

To create an environment and install the dependencies of the project, run the following commands:

git clone https://github.com/NinelK/BAND-torch.git
cd BAND-torch
micromamba create --name band-torch python=3.9
micromamba activate band-torch
pip install -e .
pre-commit install

Notes on fixing problems

To fix /lib64/libstdc++.so.6: version CXXABI_1.3.9'error, add path to this library in your env, e.g.:export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/envs/band-torch/lib/`

To fix ImportError: cannot import name 'packaging' from 'pkg_resources': downgrade setuptools <70: micromamba install setuptools==69.5.1

To fix AttributeError: module 'numpy' has no attribute 'bool8': downgrade numpy < 2 pip install numpy==1.26.0 scikit-learn==1.3.0 basically, import sklearn won't work without these downgrades. Until sklearn can't be imported, BAND will throw confusing hydra-related errors.

Backward compatibility for loading older models with newer PyTorch: export TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD=1

Monitor the progress with:

1. myjobs
2. checking tail of the log files in ~/slurm_logs/; try tail -f ~/slurm_logs/<yourlog> to watch it live. more useful commands here