added warning on pickle

README.md

@@ -147,21 +147,19 @@ Documentation is available at readthedocs: [http://hpfrec.readthedocs.io](http:/
 
 It is also internally documented through docstrings (e.g. you can try `help(hpfrec.HPF))`, `help(hpfrec.HPF.fit)`, etc.
 
-## Saving model with pickle
+## Serializing (pickling) the model
 
-Using pickle to save an `HPF` object might fail due to problems with lambda functions. The following solves it:
+Don't use `pickle` to save an `HPF` object, as it will fail due to problems with lambda functions. Rather, use `dill` instead, which has the same syntax as pickle:
 
 ```python
-import pickle
+import dill
 from hpfrec import HPF
 
 h = HPF()
-h.step_size = None
-pickle.dump(h, open("HPF_obj.p", "wb"))
+dill.dump(h, open("HPF_obj.dill", "wb"))
+h = dill.load(open("HPF_obj.dill", "rb"))
 ```
 
-(Be aware though that afterwards it won't be possible to use `partial_fit` or `add_user` with updates to item parameters.)
-
 ## Speeding up optimization procedure
 
 For faster fitting and predictions, use SciPy and NumPy libraries compiled against MKL. In Windows, you can find Python wheels (installable with pip after downloading them) of numpy and scipy precompiled with MKL in [Christoph Gohlke's website](https://www.lfd.uci.edu/~gohlke/pythonlibs/). In Linux and Mac, these come by default in Anaconda installations (but are likely to get overwritten if you enable `conda-forge`). In some small experiments from my side, this yields a near 4x speedup compared to using free linear algebra libraries (for AMD cpu's, the speedup might not be as large).