U mdT@sddlmZddlmZmZddlZddlmZddlm Z m Z ddl m Z m Z mZddlmZdd lmZe ee d dddeeeeeeeefeeefeeefe eeeeeeed ddZdS))version)OptionalUnionN)AnnData) _doc_params AnyRandom)_choose_representation doc_use_rep doc_n_pcs)settings)logging)r use_rep F euclidean)metric) adatan_pcsr perplexityearly_exaggeration learning_rate random_state use_fast_tsnen_jobscopyrreturnc  Csddl} td} | r|n|}t|||d} |dkr>tjn|}t||tjdk|||| d}| dkrt | j }|t dkrd |d <nt d | j d t|d kr|st td|rt td|r0z4ddlm}|f|}td|| d}Wn(tk r.d}t tdYnX|dkrdddlm}|f|}td|| }||jd<ddd||||| |dDi|jd<tjd| dd| r|SdS) a t-SNE [Maaten08]_ [Amir13]_ [Pedregosa11]_. t-distributed stochastic neighborhood embedding (tSNE) [Maaten08]_ has been proposed for visualizating single-cell data by [Amir13]_. Here, by default, we use the implementation of *scikit-learn* [Pedregosa11]_. You can achieve a huge speedup and better convergence if you install `Multicore-tSNE `__ by [Ulyanov16]_, which will be automatically detected by Scanpy. Parameters ---------- adata Annotated data matrix. {doc_n_pcs} {use_rep} perplexity The perplexity is related to the number of nearest neighbors that is used in other manifold learning algorithms. Larger datasets usually require a larger perplexity. Consider selecting a value between 5 and 50. The choice is not extremely critical since t-SNE is quite insensitive to this parameter. metric Distance metric calculate neighbors on. early_exaggeration Controls how tight natural clusters in the original space are in the embedded space and how much space will be between them. For larger values, the space between natural clusters will be larger in the embedded space. Again, the choice of this parameter is not very critical. If the cost function increases during initial optimization, the early exaggeration factor or the learning rate might be too high. learning_rate Note that the R-package "Rtsne" uses a default of 200. The learning rate can be a critical parameter. It should be between 100 and 1000. If the cost function increases during initial optimization, the early exaggeration factor or the learning rate might be too high. If the cost function gets stuck in a bad local minimum increasing the learning rate helps sometimes. random_state Change this to use different intial states for the optimization. If `None`, the initial state is not reproducible. n_jobs Number of jobs for parallel computation. `None` means using :attr:`scanpy._settings.ScanpyConfig.n_jobs`. copy Return a copy instead of writing to `adata`. Returns ------- Depending on `copy`, returns or updates `adata` with the following fields. **X_tsne** : `np.ndarray` (`adata.obs`, dtype `float`) tSNE coordinates of data. rNzcomputing tSNE)rr)rrverboserrrrrz0.24.0TZsquare_distanceszQResults for non-euclidean metrics changed in sklearn 0.24.0, while you are using .aIn previous versions of scanpy, calling tsne with n_jobs > 1 would use MulticoreTSNE. Now this uses the scikit-learn version of TSNE by default. If you'd like the old behaviour (which is deprecated), pass 'use_fast_tsne=True'. Note, MulticoreTSNE is not actually faster anymore.ztArgument `use_fast_tsne` is deprecated, and support for MulticoreTSNE will be dropped in a future version of scanpy.) MulticoreTSNEz7 using the 'MulticoreTSNE' package by Ulyanov (2017)Zfloat64Fz?Could not import 'MulticoreTSNE'. Falling back to scikit-learn.)TSNEz using sklearn.manifold.TSNEX_tsneparamscSsi|]\}}|dk r||qS)N).0kvr&r&K/home/sam/Atlas/atlas_env/lib/python3.8/site-packages/scanpy/tools/_tsne.py sztsne..)rrrrrrtsnez finishedz1added 'X_tsne', tSNE coordinates (adata.obsm))timedeep)sklearnlogginforr r rdict verbosityrparse __version__warningswarn UserWarning FutureWarningr"Z fit_transformZastype ImportErrorZsklearn.manifoldr#ZobsmitemsZuns)rrrrrrrrrrrr/startXZparams_sklearnZsklearn_versionr#r,r$r&r&r*r, sE                 r,) NNrrrrFNF) packagingrtypingrrr6Zanndatar_utilsrrZ tools._utilsr r r Z _settingsr r r0intstrfloatboolr,r&r&r&r*sB