| --- |
| license: mit |
| --- |
| |
| This is a dataset repository for the paper: |
|
|
| Neuro-Symbolic Predicate Invention: Learning Relational Concepts from Visual Scenes |
|
|
| Authors: Jingyuan Sha, Hikaru Shindo, Kristian Kersting, Devendra Singh Dhami. |
|
|
| --- |
|
|
| The source code is on the github, please click [here](https://github.com/ml-research/NeSy-PI/tree/main). |
|
|
| --- |
|
|
| ### Dataset Category |
|
|
| #### Simple Pattern |
| - close |
| - diagonal |
| - three same |
| - two pairs |
|
|
| The result of `Simple Patterns` is shown as follows: |
|
|
| | Pattern | Recall | Precision | |
| |-----------|--------|-----------| |
| | Close | 0.91 | 0.99 | |
| | Diagonal | 1.00 | 1.00 | |
| | Same | 0.95 | 1.00 | |
| | Two Pairs | 0.96 | 1.00 | |
|
|
| --- |
|
|
| #### Complex Pattern (to be uploaded) |
| - Check Mark |
| - Parallel |
| - Perpendicular |
| - Square |
|
|
| #### Alphabet Pattern (to be uploaded) |
| From letter A-Z (upper case), a-z (lower case). |
|
|
|
|
|
|
| #### If you have interest about this work, please consider cite: |
|
|
| ``` |
| @article{cite-key, |
| author = {Sha, Jingyuan and Shindo, Hikaru and Kersting, Kristian and Dhami, Devendra Singh}, |
| journal = {Neurosymbolic Artificial Intelligence}, |
| number = {Preprint}, |
| pages = {1--26}, |
| title = {Neuro-symbolic Predicate Invention: Learning relational concepts from visual scenes}, |
| volume = {Preprint}, |
| year = {2024}} |
| ``` |
|
|
|
|
