tohtsky/myfm

A Python/C++ implementation of Bayesian Factorization Machines

myfm's Issues

Inference at test time?

How should the FM be used to make predictions? For example, say I train this model on 1000 user movie pairs. I want to make a prediction for an unseen user, which is a vector where the values are predicted ratings for all possible movies. However, in the examples it looks like the same users get used for training and testing. ie. for user A the model trains on 80% of the known movie ratings and then tries to predict the remaining 20%. How should we call the model when we want to predict 80% of ratings for an unseen user ie. one not in the training set?

In other words I would like to take a vector of length n where I have m known ratings and infer the remaining n-m? Would I have to include the m known ratings in the training set?

Could we reconstruct the matrix (same dimensions) with the feature embedding ?

We factorize a matrix into factors, could we reconstruct a matrix based on factors.

Differences in ml100k, ml1m and documentation.

It looks as though both notebooks are loosely following the docs but all three are different. Is one example more up to date than the others or are they all intentionally different? Which notebook should readers step through?

Problem with mapper dependency.

The mapper library doesn't seem to have any type DefaultMapper...

ImportError                               Traceback (most recent call last)
<ipython-input-9-46b0ef8584af> in <module>
      6 import pandas as pd
      7 from scipy import sparse as sps
----> 8 from mapper import DefaultMapper
      9 # read movielens 1m data.
     10 from myfm.utils.benchmark_data import MovieLens1MDataManager

ImportError: cannot import name 'DefaultMapper' from 'mapper'

multicore training?

Is it possible to multicore training as in the prediction?

TypeError: 'CategoryValueToSparseEncoder' object is not subscriptable

Hi, when I run 'python ml-100k-regression.py 1', the bug traceback is as follows:
df_train.shape = (80000, 4), df_test.shape = (20000, 4)
Traceback (most recent call last):
File "ml-100k-regression.py", line 233, in
target.append(RelationBlock(user_map, augment_user_id(unique_users)))
File "ml-100k-regression.py", line 189, in augment_user_id
col.append(movie_to_internal[mid])
TypeError: 'CategoryValueToSparseEncoder' object is not subscriptable

Can you give me some instructions on the bug?

Bug in myfm.MyFMOrderedProbit() model

Hello Tomoki Ohtsuki,
I was following your ml-100k-extended exemplary notebook, but had problems running the myfm.MyFMOrderedProbit() model with use_date=False. The fitting works fine, but the problem arises during prediction. I tried to attach a screenshot of the error I get. I hope it worked. If not, the error I am getting is "ValueError: Relation blocks have inconsistent mapper size with case_size". In your notebook, if the use_date=False is set, then X is set to None and X_rel=test_blocks. The error message is based on this set None value. However, in the myfm.MyFMRegressor() model everything works as expected.

Thanks in advance for your help!