Virtual brains w/ JAX
Project description
Neural fields with Jax
basic plan
- implement shtns like api + custom kernels for shtlc
- focus on kernel design parity w/ sparse matrix
- epi2d, heun, obs2d, scan loop in time
- explore batching in time for hybrid
testing different methods, Jax allows
- numpyro: HMC & VI
- batched eval: efficient parameters sweeps for SBI
- run notebooks as tests with jupyter execute via dockerfile + gh actions
setup
Use Dockerfile if possible
pipenv
pipenv is used to manage the dependencies, so you can install deps in a virtualenv and run Jupyter notebook like so:
pip install -U pipenv
pipenv install -d
pipenv run jupyter notebook
installing shtns on mac
In the Python environment,
brew install fftw
git clone https://bitbucket.org/nschaeff/shtns
./configure --enable-python --disable-simd --prefix=/opt/homebrew
make -j && make install && python setup.py install
Pipenv
Some packages that may be used are not in the Pipfile because pipenv doesn't manage to install them.
- numba
- tvb-data & tvb-library
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
vbjax-0.0.1.tar.gz
(9.5 kB
view hashes)
Built Distribution
vbjax-0.0.1-py2.py3-none-any.whl
(10.8 kB
view hashes)
Close
Hashes for vbjax-0.0.1-py2.py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | afd83e13a8acbb4a2a7247bd73c07f7f21630e9afc9fdfdb3c10de6a9946d503 |
|
MD5 | c919e59f1a98226942936bdda80dbc9a |
|
BLAKE2b-256 | 44b4386214529e6c3e81fac44de48a04ceebea6d8c4c45d58983c66e2cba320c |