{ "id": 215453, "title": "Confusion regarding Multiclass vs Multilabel Classification Task", "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/215453", "author_name": "Parth Dhameliya", "post_date": "2021-01-29T22:10:28.844000", "votes": 8, "comment_count": 2, "views": 0, "content": "

Difference between Multiclass and Multilabel Classification?

\n

For eg, we have a chest X-ray diseases classification problem in which we have three labels
\n[ cardiomegaly, pneumonia, fibrosis]

\n

(Multiclass Classification) Use of Softmax Activation:

\n

Suppose we have 3 images in the dataset and each image(patient) has one disease

\n

image1 -> [1,0,0]
\nimage2 -> [1,0,0]
\nimage3 -> [0,1,0]

\n

In this case, we want to classify only one disease, and the sum of all labels for each sample will be equal to one.

\n

Loss Function used here is a categorical cross-entropy loss.

\n

(Multilabel Classification) Use of Sigmoid Activation:

\n

Suppose we have 3 images in the dataset and it is possible that each image(patient) has multiple diseases

\n

image1 -> [1,1,0]
\nimage2 -> [1,0,1]
\nimage3 -> [1,1,1]

\n

In this case, we want to classify multiple diseases, and the sum of all labels for each sample will not be equal to one.

\n

Loss Function used here is a Binary cross-entropy loss.

\n

Example notebook for Multilabel Classification: https://www.kaggle.com/parthdhameliya77/nih-multi-label-chest-x-ray-classification

\n

Also, share your view on weighted loss for challenging class imbalance :
\n(RANZCR CLiP EDA - Class Imbalance, Patient Overlap)

", "messages": [ { "id": 1176909, "postDate": "2021-01-29T22:10:28.843Z", "content": "

Difference between Multiclass and Multilabel Classification?

\n

For eg, we have a chest X-ray diseases classification problem in which we have three labels
\n[ cardiomegaly, pneumonia, fibrosis]

\n

(Multiclass Classification) Use of Softmax Activation:

\n

Suppose we have 3 images in the dataset and each image(patient) has one disease

\n

image1 -> [1,0,0]
\nimage2 -> [1,0,0]
\nimage3 -> [0,1,0]

\n

In this case, we want to classify only one disease, and the sum of all labels for each sample will be equal to one.

\n

Loss Function used here is a categorical cross-entropy loss.

\n

(Multilabel Classification) Use of Sigmoid Activation:

\n

Suppose we have 3 images in the dataset and it is possible that each image(patient) has multiple diseases

\n

image1 -> [1,1,0]
\nimage2 -> [1,0,1]
\nimage3 -> [1,1,1]

\n

In this case, we want to classify multiple diseases, and the sum of all labels for each sample will not be equal to one.

\n

Loss Function used here is a Binary cross-entropy loss.

\n

Example notebook for Multilabel Classification: https://www.kaggle.com/parthdhameliya77/nih-multi-label-chest-x-ray-classification

\n

Also, share your view on weighted loss for challenging class imbalance :
\n(RANZCR CLiP EDA - Class Imbalance, Patient Overlap)

", "rawMarkdown": "Difference between Multiclass and Multilabel Classification?\n\nFor eg, we have a chest X-ray diseases classification problem in which we have three labels \n[ cardiomegaly, pneumonia, fibrosis]\n\n**(Multiclass Classification) Use of Softmax Activation:**\n\n**Suppose we have 3 images in the dataset and each image(patient) has one disease**\n\nimage1 -> [1,0,0]\nimage2 -> [1,0,0]\nimage3 -> [0,1,0]\n\nIn this case, we want to classify only one disease, and **the sum of all labels for each sample will be equal to one.**\n\nLoss Function used here is a categorical cross-entropy loss.\n\n**(Multilabel Classification) Use of Sigmoid Activation:**\n\n**Suppose we have 3 images in the dataset and it is possible that each image(patient) has multiple diseases**\n\nimage1 -> [1,1,0]\nimage2 -> [1,0,1]\nimage3 -> [1,1,1]\n\nIn this case, we want to classify multiple diseases, and **the sum of all labels for each sample will not be equal to one.**\n\nLoss Function used here is a Binary cross-entropy loss.\n\n\nExample notebook for Multilabel Classification: https://www.kaggle.com/parthdhameliya77/nih-multi-label-chest-x-ray-classification\n\nAlso, share your view on weighted loss for challenging class imbalance : \n([RANZCR CLiP EDA - Class Imbalance, Patient Overlap](https://www.kaggle.com/parthdhameliya77/ranzcr-clip-eda-class-imbalance-patient-overlap))", "votes": 8 }, { "id": 1178517, "postDate": "2021-01-30T23:53:32.703Z", "rawMarkdown": "", "votes": 1, "isDeleted": true, "replies": [ { "id": 1178667, "postDate": "2021-01-31T04:16:10.750Z", "content": "

Yes, it is multilabel, I have no confusion about it. I have posted this topic just for making the difference between multiclass vs multilabel and what activation and loss function should be used.

", "rawMarkdown": "Yes, it is multilabel, I have no confusion about it. I have posted this topic just for making the difference between multiclass vs multilabel and what activation and loss function should be used.", "votes": 1 } ] } ], "comments": [ { "id": 1178517, "author_name": "", "author_url": "", "post_date": "2021-01-30T23:53:32.703000", "content": "", "votes": 1, "replies": [ { "id": 1178667, "author_name": "Parth Dhameliya", "author_url": "", "post_date": "2021-01-31T04:16:10.750000", "content": "

Yes, it is multilabel, I have no confusion about it. I have posted this topic just for making the difference between multiclass vs multilabel and what activation and loss function should be used.

", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "1176909": "Difference between Multiclass and Multilabel Classification?\n\nFor eg, we have a chest X-ray diseases classification problem in which we have three labels \n[ cardiomegaly, pneumonia, fibrosis]\n\n**(Multiclass Classification) Use of Softmax Activation:**\n\n**Suppose we have 3 images in the dataset and each image(patient) has one disease**\n\nimage1 -> [1,0,0]\nimage2 -> [1,0,0]\nimage3 -> [0,1,0]\n\nIn this case, we want to classify only one disease, and **the sum of all labels for each sample will be equal to one.**\n\nLoss Function used here is a categorical cross-entropy loss.\n\n**(Multilabel Classification) Use of Sigmoid Activation:**\n\n**Suppose we have 3 images in the dataset and it is possible that each image(patient) has multiple diseases**\n\nimage1 -> [1,1,0]\nimage2 -> [1,0,1]\nimage3 -> [1,1,1]\n\nIn this case, we want to classify multiple diseases, and **the sum of all labels for each sample will not be equal to one.**\n\nLoss Function used here is a Binary cross-entropy loss.\n\n\nExample notebook for Multilabel Classification: https://www.kaggle.com/parthdhameliya77/nih-multi-label-chest-x-ray-classification\n\nAlso, share your view on weighted loss for challenging class imbalance : \n([RANZCR CLiP EDA - Class Imbalance, Patient Overlap](https://www.kaggle.com/parthdhameliya77/ranzcr-clip-eda-class-imbalance-patient-overlap))", "1178517": "" } }