Description

hi, thanks for developing this outstanding tools, and could use my own dataset instead of using the default dataset? If it is ok, how I replace the dataset?

Other Comments

Description

Add a new evaluation method (Stratified Evaluation) proposed here and update the documents accordingly (add new tutorial,...).

Related Issues

Checklist:

• [ ] I have added tests.
• [*] I have updated the documentation accordingly.
• [*] I have updated README.md (if you are adding a new model).
• [ ] I have updated examples/README.md (if you are adding a new example).
• [ ] I have updated datasets/README.md (if you are adding a new dataset).

Description

Great library! Is there a mechanism to remove already seen/liked/interacted-with items upon prediction for some model? MS Recommenders uses this.

Other Comments

Description

Hi，

• Why are the NDCG values calculated by cornac.metrics.NDCG mostly at [0.01,0.1], while the results calculated by many recommendation system papers are mostly at [0.1,1]?
• Are the two methods of calculating NDCG different?
• What should I do if I want to reproduce the results of the paper?

BTW, cornac is awesome!

Other Comments

Description

Hi,

When I used cornac with implicit datset, some evaluations make me puzzled(the "MAE" and "RMSE" were always 0, is it right or a bug?). Note that the implicit dataset in my experiments only have the positive feedback(the ratings is 1) while the negative feedbacks have beed removed.

when the epochs=20, the results were shown as follow:

when the epochs=40, the results were shown as follow:

when the epochs=60, the results were shown as follow:

the codes have been used were shown as follow:

Thanks for your reading and your excellent jobs in this tool!

In which platform does it happen?

How do we replicate the issue?

Expected behavior (i.e. solution)

Other Comments

Description

I just added a new tutorial to represent propensity-based stratified evaluation.

Related Issues

Checklist:

• [ ] I have added tests.
• [ ] I have updated the documentation accordingly.
• [ ] I have updated README.md (if you are adding a new model).
• [ ] I have updated examples/README.md (if you are adding a new example).
• [ ] I have updated datasets/README.md (if you are adding a new dataset).

Description

Hello, I use Ubunto version 20.04, and installed cornac using the pip3 option (under venv) and conda (under conda venv), and in both case, I am receiving this error when I want to run cornac, first_example.py:

Traceback (most recent call last):
  File "/home/yas/PycharmProjects/cornac/examples/first_example.py", line 17, in <module>
    import cornac
  File "/home/yas/PycharmProjects/cornac/cornac/__init__.py", line 16, in <module>
    from . import data
  File "/home/yas/PycharmProjects/cornac/cornac/data/__init__.py", line 18, in <module>
    from .text import TextModality, ReviewModality
  File "/home/yas/PycharmProjects/cornac/cornac/data/text.py", line 27, in <module>
    from ..utils import normalize
  File "/home/yas/PycharmProjects/cornac/cornac/utils/__init__.py", line 16, in <module>
    from .common import validate_format
  File "/home/yas/PycharmProjects/cornac/cornac/utils/common.py", line 21, in <module>
    from .fast_sparse_funcs import (
ModuleNotFoundError: No module named 'cornac.utils.fast_sparse_funcs'

Let me add that I use PyCharm. Any idea why the issue emerged and how it could be handled?

Fit using modalities

Hello,

First of all, I can't thank you enough for your efforts with Cornac, it is a really powerful tool. I'm trying to implement PCRL using Cornac. The experiments work as shown in examples. But I can't fit the model. There seems to be a lack of documentation on this area, I can't find how to do it in any page. Here's the issue:

I hope you can help me.

my very best,

Miguel Ângelo Rebelo

Description

Thank you for putting up cornac, it is a very useful resource for pushing research.

I just wanted to ask if there is a way to see the per-user evaluation score for RS models? Let's assume, we do not want to average over user-obtained NDCGs but we want to compute a PDF on the results. How can we get user-related evaluation scores in order to aggregate the results in a specific manner (beyond simple average)? Thank you for your input.

Description

Related Issues

Checklist:

• [ ] I have added tests.
• [x] I have updated the documentation accordingly.
• [x] I have updated README.md (if you are adding a new model).
• [x] I have updated examples/README.md (if you are adding a new example).
• [ ] I have updated datasets/README.md (if you are adding a new dataset).

Hello guys, great work, way better than Microsoft! I have a small issue, please read below:

Description

If I set exclude_unknowns=False then all the ratings/rankings fail with a similar error: e.g. IndexError('index 1840 is out of bounds for axis 0 with size 1840')

In which platform does it happen?

Linux 5.11.0-36-generic

How do we replicate the issue?

This is easy, just make sure to have non-zeros for: Number of unknown users = 5 Number of unknown items = 4 after when doing RatioSplit

Expected behavior (i.e. solution)

I would still expect to get the scores even when there are non-matching indices between train_set and test_set

Description

The WMF model cannot be used in Experiment, due to a depricated tensorflow method. The new method must be 'tf.random.set_seed()'. The error output is like:

/usr/local/lib/python3.8/dist-packages/cornac/models/wmf/recom_wmf.py in _fit_cf(self) 165 graph = tf.Graph() 166 with graph.as_default(): --> 167 tf.set_random_seed(self.seed) 168 model = Model( 169 n_users=n_users,

AttributeError: module 'tensorflow' has no attribute 'set_random_seed'

In which platform does it happen?

Found while using in google colab.

How do we replicate the issue?

Try to use the WMF model in an Experiment

solution

In line 167 of recom_wmf.py change 'tf.set_random_seed()' to 'tf.random.set_seed()'.

I just installed cornac from github, liked described in the readme file. But now if I try to execute one of the examples, I always get the error: ModuleNotFoundError: No module named 'cornac.utils.fast_sparse_funcs' why does the import of cornac fail if everything is installed? On Colab within a Jupyter Notebook everything works fine, except the WMF model, because of a deprecated method. Because of this bug I`d like to use cornac on my laptop instead of running it from Colab, to fix this bug.

Description

Dear cornac developers,

I am trying to use cornac within the server of my university. However, I get a compatibility issue when I import; cornac requires GNU libc version of 2.27, and ours is 2.25.

I wonder if you can suggest any work around. E.g., is there an older cornac version that requires an earlier version of libc?

Thank you in advance!

Description

Hi,

We introduced two novel collaborative filtering techniques for recommendation systems in cases of various cold-start situations and incomplete datasets.

• Hakan Yilmazer, Selma Ayşe Özel, ImposeSVD: Incrementing PureSVD For Top-N Recommendations for Cold-Start Problems and Sparse Datasets, The Computer Journal, 2022;, bxac106, https://doi.org/10.1093/comjnl/bxac106.

The first model z-scoREC establishes an asymmetric weight matrix between items without using item meta-data and eradicates the disadvantages of neighborhood approaches by automatic determination of threshold values.

The implementation of z-scoREC model is implemented in this PR.

Related Issues

Checklist:

• [ ] I have added tests.
• [x] I have updated the documentation accordingly.
• [x] I have updated README.md (if you are adding a new model).
• [x] I have updated examples/README.md (if you are adding a new example).

Description

I check the recommender.rank() and bivaecf.score() and find that the interactive items in trainset are not removed during the evaluation phase.

When recommending top-n items with highest score from whole item space, whether it's necessary to remove the interactive item in trainset or give these imteractive item a smallest value (or zero)?

I give these interactive items zero value by adding follow codes in ***bivaecf.score()***,

if item_idx is None:
    ...
    train_mat = self.train_set.csr_matrix
    csr_row = train_mat.getrow(user_idx)
    pos_items = [item_idx for (item_idx, rating) in zip(csr_row.indices, csr_row.data) if rating >= 4.0]
    known_item_scores[pos_items] = 0.
    return known_item_scores

From the results on ml-100k, there was a significant improvement for each metrics,

| Model | [email protected] | [email protected] |[email protected] | | ------------- | ------------- | ------------- | ------------- | | BiVAECF | 0.2378 | 0.0739 | 0.4288 | | BiVAECF with setting zero value | 0.3392 | 0.1048 | 0.5233 |

Other Comments