Skip to content

WIP: Compute projection instead of dictionary #108

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
jakirkham wants to merge 1 commit into
base: master
Choose a base branch
from
Open

WIP: Compute projection instead of dictionary #108

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
110 changes: 15 additions & 95 deletions nanshe_ipython.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -241,7 +241,7 @@
"source": [
"from nanshe_workflow.par import halo_block_parallel\n",
"\n",
"from nanshe_workflow.imp2 import extract_f0, wavelet_transform, renormalized_images, normalize_data\n",
"from nanshe_workflow.imp2 import extract_f0, wavelet_transform, normalize_data\n",
"\n",
"from nanshe_workflow.par import halo_block_generate_dictionary_parallel\n",
"from nanshe_workflow.imp import block_postprocess_data_parallel\n",
Expand Down Expand Up @@ -993,11 +993,10 @@
"cell_type": "markdown",
"metadata": {},
"source": [
"### Normalize Data\n",
"### Project\n",
"\n",
"* `block_frames` (`int`): number of frames to work with in each full frame block (run in parallel).\n",
"* `block_space` (`int`): extent of each spatial dimension for each block (run in parallel).\n",
"* `norm_frames` (`int`): number of frames for use during normalization of each full frame block (run in parallel)."
"* `proj_type` (`str`): type of projection to take."
]
},
{
Expand All @@ -1007,12 +1006,11 @@
"outputs": [],
"source": [
"block_frames = 40\n",
"block_space = 300\n",
"norm_frames = 100\n",
"proj_type = \"max\"\n",
"\n",
"\n",
"with get_executor(client) as executor:\n",
" dask_io_remove(data_basename + postfix_norm + zarr_ext, executor)\n",
" dask_io_remove(data_basename + postfix_dict + zarr_ext, executor)\n",
"\n",
"\n",
" with open_zarr(data_basename + postfix_wt + zarr_ext, \"r\") as f:\n",
Expand All @@ -1027,106 +1025,28 @@
" da_imgs_flt.dtype.itemsize >= 4):\n",
" da_imgs_flt = da_imgs_flt.astype(np.float32)\n",
"\n",
" da_imgs_flt_mins = da_imgs_flt.min(\n",
" axis=tuple(irange(1, da_imgs_flt.ndim)),\n",
" keepdims=True\n",
" )\n",
"\n",
" da_imgs_flt_shift = da_imgs_flt - da_imgs_flt_mins\n",
"\n",
" da_result = renormalized_images(da_imgs_flt_shift)\n",
" da_result = da_imgs\n",
" if proj_type == \"max\":\n",
" da_result = da_result.max(axis=0, keepdims=True)\n",
" elif proj_type == \"std\":\n",
" da_result = da_result.std(axis=0, keepdims=True)\n",
"\n",
" # Store denoised data\n",
" dask_store_zarr(data_basename + postfix_norm + zarr_ext, [\"images\"], [da_result], executor)\n",
"\n",
"\n",
" zip_zarr(data_basename + postfix_norm + zarr_ext, executor)\n",
"\n",
"\n",
"if __IPYTHON__:\n",
" result_image_stack = LazyZarrDataset(data_basename + postfix_norm + zarr_ext, \"images\")\n",
"\n",
" mplsv = plt.figure(FigureClass=MPLViewer)\n",
" mplsv.set_images(\n",
" result_image_stack,\n",
" vmin=par_compute_min_projection(num_frames=norm_frames)(result_image_stack).min(),\n",
" vmax=par_compute_max_projection(num_frames=norm_frames)(result_image_stack).max()\n",
" )"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Dictionary Learning\n",
"\n",
"* `n_components` (`int`): number of basis images in the dictionary.\n",
"* `batchsize` (`int`): minibatch size to use.\n",
"* `iters` (`int`): number of iterations to run before getting dictionary.\n",
"* `lambda1` (`float`): weight for L<sup>1</sup> sparisty enforcement on sparse code.\n",
"* `lambda2` (`float`): weight for L<sup>2</sup> sparisty enforcement on sparse code.\n",
"\n",
"<br>\n",
"* `block_frames` (`int`): number of frames to work with in each full frame block (run in parallel).\n",
"* `norm_frames` (`int`): number of frames for use during normalization of each full frame block (run in parallel)."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"n_components = 50\n",
"batchsize = 256\n",
"iters = 100\n",
"lambda1 = 0.2\n",
"lambda2 = 0.0\n",
"\n",
"block_frames = 51\n",
"norm_frames = 100\n",
" dask_store_zarr(data_basename + postfix_dict + zarr_ext, [\"images\"], [da_result], executor)\n",
"\n",
"\n",
"with get_executor(client) as executor:\n",
" dask_io_remove(data_basename + postfix_dict + zarr_ext, executor)\n",
"\n",
"\n",
"result = LazyZarrDataset(data_basename + postfix_norm + zarr_ext, \"images\")\n",
"block_shape = (block_frames,) + result.shape[1:]\n",
"with open_zarr(data_basename + postfix_dict + zarr_ext, \"w\") as f2:\n",
" new_result = f2.create_dataset(\"images\", shape=(n_components,) + result.shape[1:], dtype=result.dtype, chunks=True)\n",
"\n",
" result = par_generate_dictionary(block_shape)(\n",
" result,\n",
" n_components=n_components,\n",
" out=new_result,\n",
" **{\"sklearn.decomposition.dict_learning_online\" : {\n",
" \"n_jobs\" : 1,\n",
" \"n_iter\" : iters,\n",
" \"batch_size\" : batchsize,\n",
" \"alpha\" : lambda1\n",
" }\n",
" }\n",
" )\n",
"\n",
" result_j = f2.create_dataset(\"images_j\", shape=new_result.shape, dtype=numpy.uint16, chunks=True)\n",
" par_norm_layer(num_frames=norm_frames)(result, out=result_j)\n",
"\n",
"\n",
"with get_executor(client) as executor:\n",
" zip_zarr(data_basename + postfix_dict + zarr_ext, executor)\n",
"\n",
"\n",
"if __IPYTHON__:\n",
" result_image_stack = LazyZarrDataset(data_basename + postfix_dict + zarr_ext, \"images\")\n",
" result_image_stack = LazyZarrDataset(data_basename + postfix_dict + zarr_ext, \"images\")[...][...]\n",
"\n",
" mplsv = plt.figure(FigureClass=MPLViewer)\n",
" mplsv.set_images(\n",
" result_image_stack,\n",
" vmin=par_compute_min_projection(num_frames=norm_frames)(result_image_stack).min(),\n",
" vmax=par_compute_max_projection(num_frames=norm_frames)(result_image_stack).max()\n",
" )\n",
" mplsv.time_nav.stime.label.set_text(\"Basis Image\")"
" vmin=result_image_stack.min(),\n",
" vmax=result_image_stack.max()\n",
" )"
]
},
{
Expand Down