MNIST Add was first introduced in [1] and the inputs of this task are pairs of MNIST images and the outputs are their sums. The dataset looks like this:
.. image:: ../img/image_1.png
.. image:: ../img/image_1.jpg
:width: 350px
:align: center
@@ -13,7 +13,7 @@ The ``gt_pseudo_label`` is only used to test the performance of the machine lear
In the Abductive Learning framework, the inference process is as follows:
.. image:: ../img/image_2.png
.. image:: ../img/image_2.jpg
:width: 700px
[1] Robin Manhaeve, Sebastijan Dumancic, Angelika Kimmig, Thomas Demeester, and Luc De Raedt. Deepproblog: Neural probabilistic logic programming. In Advances in Neural Information Processing Systems 31 (NeurIPS), pages 3749-3759.2018.
Gao Enhao
