Update to version 0.2.0, update references, unpin polyagamma

.github/workflows/ci.yml

@@ -18,7 +18,7 @@ jobs:
     runs-on: ubuntu-20.04
     strategy:
       matrix:
-        python-version: [3.8]
+        python-version: [3.9]
     steps:
     - uses: actions/checkout@v2
     - name: Set up Python ${{ matrix.python-version }}
@@ -38,7 +38,7 @@ jobs:
     needs: lint
     strategy:
       matrix:
-        python-version: [3.8]
+        python-version: [3.9]
     steps:
       - uses: actions/checkout@v2
       - name: Set up Python ${{ matrix.python-version }}
@@ -51,7 +51,7 @@ jobs:
           sudo apt-get update
           sudo apt-get install gcc-8 g++-8 ninja-build
           python -m pip install --upgrade pip wheel setuptools
-          pip install torch==1.11.0+cpu -f https://download.pytorch.org/whl/torch_stable.html
+          pip install torch==2.3.1+cpu -f https://download.pytorch.org/whl/torch_stable.html
           pip install .[test]
           pip install coveralls
           pip freeze

docs/source/getting_started.rst

@@ -142,7 +142,7 @@ Martin Jankowiak: martin@basis.ai
 References
 ----------
 
-* Jankowiak, M., 2022. `Bayesian Variable Selection in a Million Dimensions <https://arxiv.org/abs/2208.01180>`__ arXiv preprint arXiv:2208.01180.
+* Jankowiak, M., 2023. `Bayesian Variable Selection in a Million Dimensions <https://proceedings.mlr.press/v206/jankowiak23a.html>`__ AISTATS 2023.
 
 * Zanella, G. and Roberts, G., 2019. `Scalable importance tempering and Bayesian variable selection <https://rss.onlinelibrary.wiley.com/doi/abs/10.1111/rssb.12316>`__. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 81(3), pp.489-517.
 
@@ -153,12 +153,17 @@ If you use millipede please consider citing:
 
 ::
 
-@article{jankowiak2022bayesian,
-      title={Bayesian Variable Selection in a Million Dimensions},
-      author={Martin Jankowiak},
-      journal={arXiv preprint arXiv:{2208.01180},
-      year={2022},
-      eprint={2208.01180},
-      archivePrefix={arXiv},
-      primaryClass={stat.ME}
+@InProceedings{pmlr-v206-jankowiak23a,
+  title = 	  {Bayesian Variable Selection in a Million Dimensions},
+  author =    {Jankowiak, Martin},
+  booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics},
+  pages = 	  {253--282},
+  year = 	  {2023},
+  volume = 	  {206},
+  series = 	  {Proceedings of Machine Learning Research},
+  month = 	  {25--27 Apr},
+  publisher = {PMLR},
+  pdf = 	  {https://proceedings.mlr.press/v206/jankowiak23a/jankowiak23a.pdf},
+  url = 	  {https://proceedings.mlr.press/v206/jankowiak23a.html},
 }
+

millipede/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.1.0"
+__version__ = "0.2.0"
 
 from millipede.binomial import CountLikelihoodSampler
 from millipede.normal import NormalLikelihoodSampler

millipede/util.py

@@ -20,7 +20,7 @@ def safe_cholesky(A, epsilon=1.0e-8):
     if A.shape == (1, 1):
         return A.sqrt()
     try:
-        return torch.linalg.cholesky(A)
+        return torch.linalg.cholesky(A, upper=False)
     except RuntimeError as e:
         Aprime = A.clone()
         jitter_prev = 0.0
@@ -29,7 +29,7 @@ def safe_cholesky(A, epsilon=1.0e-8):
             Aprime.diagonal(dim1=-2, dim2=-1).add_(jitter_new - jitter_prev)
             jitter_prev = jitter_new
             try:
-                return torch.linalg.cholesky(Aprime)
+                return torch.linalg.cholesky(Aprime, upper=False)
             except RuntimeError:
                 continue
         raise e

setup.py

@@ -30,7 +30,7 @@
     install_requires=[
         "torch>=1.11",
         "pandas",
-        "polyagamma==1.3.2",
+        "polyagamma",
         "tqdm",
     ],
     extras_require={

tests/test_count_samplers.py

@@ -15,7 +15,7 @@
 
 @pytest.mark.parametrize("subset_size", [12, None])
 @pytest.mark.parametrize("variable_S", [False, True])
-def test_binomial(subset_size, variable_S, streaming=False, N=512, P=16, T=2000, T_burnin=200, intercept=0.17, seed=1):
+def test_binomial(subset_size, variable_S, streaming=False, N=512, P=16, T=2000, T_burnin=500, intercept=0.17, seed=1):
     torch.manual_seed(seed)
     X = torch.randn(N, P).double()
     X_assumed = torch.randn(N, 2).double()