While traditional machine learning can effectively tackle a wide range o...
The rapid expansion of foundation pre-trained models and their fine-tune...
The Click-Through Rate (CTR) prediction task is critical in industrial
r...
Figuring out which Pre-Trained Model (PTM) from a model zoo fits the tar...
Class-Incremental Learning (CIL) or continual learning is a desired
capa...
Learning new classes without forgetting is crucial for real-world
applic...
Multivariate time series data comprises various channels of variables. T...
Class-incremental learning (CIL) aims to adapt to emerging new classes
w...
Deep models, e.g., CNNs and Vision Transformers, have achieved impressiv...
When there are unlabeled Out-Of-Distribution (OOD) data from other class...
Traditional self-supervised contrastive learning methods learn embedding...
Real-world applications require the classification model to adapt to new...
The knowledge of a well-trained deep neural network (a.k.a. the "teacher...
Knowledge distillation (KD) has shown its effectiveness in improving a
s...
The ability to learn new concepts continually is necessary in this
ever-...
Rich semantics inside an image result in its ambiguous relationship with...
New classes arise frequently in our ever-changing world, e.g., emerging
...
Novel classes frequently arise in our dynamically changing world, e.g., ...
Traditional machine learning systems are deployed under the closed-world...
Traditional learning systems are trained in closed-world for a fixed num...
Few-shot learning (FSL) aims to train a strong classifier using limited
...
Model-agnostic meta-learning (MAML) is arguably the most popular
meta-le...
One single instance could possess multiple portraits and reveal diverse
...
Although vital to computer vision systems, few-shot action recognition i...
The support/query (S/Q) training protocol is widely used in meta-learnin...
Neural networks trained with class-imbalanced data are known to perform
...
Traditional classifiers are deployed under closed-set setting, with both...
While deep learning demonstrates its strong ability to handle independen...
Meta-learning becomes a practical approach towards few-shot image
classi...
Few-shot classification algorithms can alleviate the data scarceness iss...
Recent years have witnessed an abundance of new publications and approac...
We investigate learning a ConvNet classifier with class-imbalanced data....
Visual recognition in real-world requires handling long-tailed and even
...
Learning with limited data is a key challenge for visual recognition.
Fe...