I am trying to run TimeSeriesKMeans from TSlearn via reticulate but apparently fail to properly convert my R data.frame to a numpy array. Full code and error below. Please have a look what I do wrong. Explanation appreciated because my knowledge of Python internals is quite poor.
library(reticulate)
library(magrittr)
reticulate::py_install(c("numpy", "pandas", "tslearn"))
#> Using virtual environment '/home/rstudio/.virtualenvs/r-reticulate' ...
#> + /home/rstudio/.virtualenvs/r-reticulate/bin/python -m pip install --upgrade --no-user numpy pandas tslearn
# Simulate rhythmic time series
time <- seq(1, 21, 2)
time_series <-
lapply(seq(0, 1, .2), function(x) x*sin(2*pi*time/24)) %>%
unlist %>%
matrix(., byrow=TRUE, ncol=length(time)) %>%
data.frame %>% setNames(paste0("time", time))
# Convert time_series to numpy array
np <- reticulate::import("numpy")
ts_np <- r_to_py(np$array(time_series))
# Run https://tslearn.readthedocs.io/en/latest/gen_modules/clustering/tslearn.clustering.TimeSeriesKMeans.html via reticulate
ts <- reticulate::import("tslearn.clustering")
tsk <- ts$TimeSeriesKMeans(n_clusters=2, metric="dtw")
# Problematic line
try(tsk$fit(ts_np))
#> Error in py_call_impl(callable, call_args$unnamed, call_args$named) :
#> TypeError: 'float' object cannot be interpreted as an integer
#> Run `reticulate::py_last_error()` for details.
reticulate::py_list_packages()
#> package version requirement
#> 1 contourpy 1.1.0 contourpy==1.1.0
#> 2 cycler 0.11.0 cycler==0.11.0
#> 3 dtaidistance 2.3.10 dtaidistance==2.3.10
#> 4 fonttools 4.42.0 fonttools==4.42.0
#> 5 h5py 3.9.0 h5py==3.9.0
#> 6 importlib-metadata 6.8.0 importlib-metadata==6.8.0
#> 7 importlib-resources 6.0.1 importlib-resources==6.0.1
#> 8 joblib 1.3.2 joblib==1.3.2
#> 9 kiwisolver 1.4.4 kiwisolver==1.4.4
#> 10 llvmlite 0.41.0 llvmlite==0.41.0
#> 11 matplotlib 3.7.2 matplotlib==3.7.2
#> 12 natsort 8.4.0 natsort==8.4.0
#> 13 networkx 3.1 networkx==3.1
#> 14 numba 0.58.0 numba==0.58.0
#> 15 numpy 1.24.4 numpy==1.24.4
#> 16 packaging 23.1 packaging==23.1
#> 17 pandas 2.0.3 pandas==2.0.3
#> 18 patsy 0.5.3 patsy==0.5.3
#> 19 Pillow 10.0.0 Pillow==10.0.0
#> 20 pkg_resources 0.0.0 pkg_resources==0.0.0
#> 21 polars 0.18.15 polars==0.18.15
#> 22 progressbar 2.5 progressbar==2.5
#> 23 pyarrow 12.0.1 pyarrow==12.0.1
#> 24 pyparsing 3.0.9 pyparsing==3.0.9
#> 25 python-dateutil 2.8.2 python-dateutil==2.8.2
#> 26 pytz 2023.3 pytz==2023.3
#> 27 scikit-fuzzy 0.4.2 scikit-fuzzy==0.4.2
#> 28 scikit-learn 1.3.1 scikit-learn==1.3.1
#> 29 scipy 1.10.1 scipy==1.10.1
#> 30 seaborn 0.12.2 seaborn==0.12.2
#> 31 six 1.16.0 six==1.16.0
#> 32 statsmodels 0.14.0 statsmodels==0.14.0
#> 33 threadpoolctl 3.2.0 threadpoolctl==3.2.0
#> 34 tslearn 0.6.2 tslearn==0.6.2
#> 35 tzdata 2023.3 tzdata==2023.3
#> 36 zipp 3.16.2 zipp==3.16.2
reticulate::py_version()
#> [1] '3.8'
sessionInfo()
#> R version 4.2.1 (2022-06-23)
#> Platform: x86_64-pc-linux-gnu (64-bit)
#> Running under: Ubuntu 20.04.4 LTS
#>
#> Matrix products: default
#> BLAS: /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3
#> LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/liblapack.so.3
#>
#> locale:
#> [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
#> [3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
#> [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
#> [7] LC_PAPER=en_US.UTF-8 LC_NAME=C
#> [9] LC_ADDRESS=C LC_TELEPHONE=C
#> [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
#>
#> attached base packages:
#> [1] stats graphics grDevices utils datasets methods base
#>
#> other attached packages:
#> [1] magrittr_2.0.3 reticulate_1.31
#>
#> loaded via a namespace (and not attached):
#> [1] Rcpp_1.0.9 rstudioapi_0.13 knitr_1.39 lattice_0.20-45
#> [5] R.cache_0.16.0 rlang_1.0.4 fastmap_1.1.0 fansi_1.0.3
#> [9] stringr_1.4.0 styler_1.10.2 highr_0.9 tools_4.2.1
#> [13] grid_4.2.1 xfun_0.32 png_0.1-7 R.oo_1.25.0
#> [17] utf8_1.2.2 cli_3.3.0 withr_2.5.0 htmltools_0.5.3
#> [21] yaml_2.3.5 digest_0.6.29 lifecycle_1.0.1 Matrix_1.4-1
#> [25] purrr_0.3.4 vctrs_0.4.1 R.utils_2.12.0 fs_1.5.2
#> [29] glue_1.6.2 evaluate_0.16 rmarkdown_2.14 reprex_2.0.1
#> [33] stringi_1.7.8 pillar_1.8.0 compiler_4.2.1 R.methodsS3_1.8.2
#> [37] jsonlite_1.8.0
Created on 2023-09-22 by the reprex package (v2.0.1)
The py_last_error:
> reticulate::py_last_error()
── Python Exception Message ───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
Traceback (most recent call last):
File "/home/rstudio/.virtualenvs/r-reticulate/lib/python3.8/site-packages/tslearn/clustering/kmeans.py", line 821, in fit
self._fit_one_init(X_, x_squared_norms, rs)
File "/home/rstudio/.virtualenvs/r-reticulate/lib/python3.8/site-packages/tslearn/clustering/kmeans.py", line 675, in _fit_one_init
self.cluster_centers_ = _k_init_metric(
File "/home/rstudio/.virtualenvs/r-reticulate/lib/python3.8/site-packages/tslearn/clustering/kmeans.py", line 98, in _k_init_metric
centers = numpy.empty((n_clusters, n_timestamps, n_features), dtype=X.dtype)
TypeError: 'float' object cannot be interpreted as an integer
── R Traceback ────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
▆
1. ├─base::try(tsk$fit(ts_np))
2. │ └─base::tryCatch(...)
3. │ └─base (local) tryCatchList(expr, classes, parentenv, handlers)
4. │ └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
5. │ └─base (local) doTryCatch(return(expr), name, parentenv, handler)
6. └─tsk$fit(ts_np)
7. └─reticulate:::py_call_impl(callable, call_args$unnamed, call_args$named)
Thanks to rpolicastro via Slack.
The answer is that one has to make the cluster number explicitely integer:
tsk <- ts$TimeSeriesKMeans(n_clusters=2L, metric="dtw")
So 2L rather than 2 which turns it integer, not float/numeric.