![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1528571%2F5884e8ca7aff6d96b5e58a9a7d7f7b88%2Fcoral.png?generation=1566528886004408&alt=media\")
Diversity helps when ensembling!! I believe most of us already have classification + regression models. Mayb only few have Ordinal regression models. For those who don't, this might be worth trying!!
\n\nExtraordinary progress has been made towards developing neural network architectures for classification tasks. However, commonly used loss functions such as the multi-category cross entropy loss are inadequate for ranking and ordinal regression problems. Hence, approaches that utilize neural networks for ordinal regression tasks transform ordinal target variables into a series of binary classification tasks but suffer from inconsistencies among the different binary classifiers. Thus, we propose a new framework (Consistent Rank Logits, CORAL) with theoretical guarantees for rank-monotonicity and consistent confidence scores. Through parameter sharing, our framework also benefits from lower training complexity and can easily be implemented to extend conventional convolutional neural network classifiers for ordinal regression tasks. Furthermore, the empirical evaluation of our method on a range of face image datasets for age prediction shows a substantial improvement compared to the current state-of-the-art ordinal regression method.
Diversity helps when ensembling!! I believe most of us already have classification + regression models. Mayb only few have Ordinal regression models. For those who don't, this might be worth trying!!
\n\nExtraordinary progress has been made towards developing neural network architectures for classification tasks. However, commonly used loss functions such as the multi-category cross entropy loss are inadequate for ranking and ordinal regression problems. Hence, approaches that utilize neural networks for ordinal regression tasks transform ordinal target variables into a series of binary classification tasks but suffer from inconsistencies among the different binary classifiers. Thus, we propose a new framework (Consistent Rank Logits, CORAL) with theoretical guarantees for rank-monotonicity and consistent confidence scores. Through parameter sharing, our framework also benefits from lower training complexity and can easily be implemented to extend conventional convolutional neural network classifiers for ordinal regression tasks. Furthermore, the empirical evaluation of our method on a range of face image datasets for age prediction shows a substantial improvement compared to the current state-of-the-art ordinal regression method.
Thank you for sharing this kind of research with the community.\nAppreciate it.
", "rawMarkdown": "Thank you for sharing this kind of research with the community.\nAppreciate it.", "votes": 1, "replies": [ { "id": 607865, "postDate": "2019-08-26T02:54:20.223Z", "content": "Thank u for ur kind words!! Will keep be motivated to share more interesting findings :)
", "rawMarkdown": "Thank u for ur kind words!! Will keep be motivated to share more interesting findings :)", "votes": 2 } ] }, { "id": 605969, "postDate": "2019-08-23T03:06:35.560Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 607615, "author_name": "Prashant Kikani", "author_url": "", "post_date": "2019-08-25T15:40:52.977000", "content": "Thank you for sharing this kind of research with the community.\nAppreciate it.
", "votes": 1, "replies": [ { "id": 607865, "author_name": "Bibek", "author_url": "", "post_date": "2019-08-26T02:54:20.223000", "content": "Thank u for ur kind words!! Will keep be motivated to share more interesting findings :)
", "votes": 2, "replies": [] } ] }, { "id": 605969, "author_name": "", "author_url": "", "post_date": "2019-08-23T03:06:35.560000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "605962": "Diversity helps when ensembling!! I believe most of us already have classification + regression models. Mayb only few have Ordinal regression models. For those who don't, this might be worth trying!!\n\n##[Rank-consistent Ordinal Regression for Neural Networks](https://github.com/Raschka-research-group/coral-cnn)\n`Extraordinary progress has been made towards developing neural network architectures for classification tasks. However, commonly used loss functions such as the multi-category cross entropy loss are inadequate for ranking and ordinal regression problems. Hence, approaches that utilize neural networks for ordinal regression tasks transform ordinal target variables into a series of binary classification tasks but suffer from inconsistencies among the different binary classifiers. Thus, we propose a new framework (Consistent Rank Logits, CORAL) with theoretical guarantees for rank-monotonicity and consistent confidence scores. Through parameter sharing, our framework also benefits from lower training complexity and can easily be implemented to extend conventional convolutional neural network classifiers for ordinal regression tasks. Furthermore, the empirical evaluation of our method on a range of face image datasets for age prediction shows a substantial improvement compared to the current state-of-the-art ordinal regression method.`\n\n\n\n", "607615": "Thank you for sharing this kind of research with the community.\nAppreciate it.", "605969": "" } }