Transfer Learning Task

A Transfer Learning Task is a learning task that requires inferences on a testing record to be made from a target domain (with a sample space) from a different underlying distribution than the source domain (of the training records).

• AKA: Domain Adaptable Learning.
• Context:
  • Input: a Pretrained Model.
  • It can be solved by a Transfer Learning System (that implements a transfer learning algorithm).
  • It can can be easier to set up than a Supervised Learning Task, which needs extensive labeled data in the target domain.
  • It can range from being a Transfer Learning Class-Inference Task to being a Transfer Learning Number-Infernce Task.
  • ...
• Example(s):
  • An Image-based Transfer Learning Task, such as after learning on ImageNet-based Model, train a model on different labels.
  • An NLP-based Transfer Learning Task, such as through the use of word embeddings trained on an external text corpus.
  • A Zero-Shot Transfer Learning Task, such as using a model trained on images of animals to identify a type of animal it was never explicitly trained on.
  • A Few-Shot Transfer Learning Task, such as using a model trained to recognize handwriting to identify a new character after only a few examples.
  • A Pre-Trained Model Fine Tuning Task/model fine tuning.
  • …
• Counter-Example(s):
  • Single Domain Learning Task, where training and testing data come from the same distribution.
  • Unsupervised Learning Task, where the model learns without labeled examples from any domain.
• See: Metalearning, Transfer of Learning, Instance-Weighing Task, Anti-Spam Techniques, Inductive Transfer, Multi-Task Learning, Domain Adaptation.

---

---

References

2022

• (Wikipedia, 2022) ⇒ https://en.wikipedia.org/wiki/Domain_adaptation Retrieved:2022-5-11.
  • Domain adaptation is a field associated with machine learning and transfer learning. This scenario arises when we aim at learning from a source data distribution a well performing model on a different (but related) target data distribution. For instance, one of the tasks of the common spam filtering problem consists in adapting a model from one user (the source distribution) to a new user who receives significantly different emails (the target distribution). Domain adaptation has also been shown to be beneficial for learning unrelated sources. Note that, when more than one source distribution is available the problem is referred to as multi-source domain adaptation.

2015

• (Wikipedia, 2015) ⇒ http://en.wikipedia.org/wiki/inductive_transfer Retrieved:2015-5-3.
  • Inductive transfer, or transfer learning, is a research problem in machine learning that focuses on storing knowledge gained while solving one problem and applying it to a different but related problem. For example, the abilities acquired while learning to walk presumably apply when one learns to run, and knowledge gained while learning to recognize cars could apply when recognizing trucks. This area of research bears some relation to the long history of psychological literature on transfer of learning, although formal ties between the two fields are limited. Notably, scientists have developed algorithms for inductive transfer in Markov logic networks and Bayesian networks. Furthermore, researchers have applied techniques for transfer to problems in text classification, spam filtering, and urban combat simulation. There still exists much potential in this field while the "transfer" hasn't yet led to significant improvement in learning. Also, an intuitive understanding could be that "transfer means a learner can directly learn from other correlated learners". However, in this way, such a methodology in transfer learning, whose direction is illustrated by, is not a hot spot in the area yet.

2011

• (Vilalta, Giraud-Carrier, Brazdil, & Soares, 2011) ⇒ Ricardo Vilalta; Christopher Giraud-Carrier; Pavel Brazdil; Carlos Soares. (2011). “Inductive Transfer.” In: (Sammut & Webb, 2011) p.545

2010

• (Pan & Tang, 2010) ⇒ Sinno Jialin Pan, and Qiang Yang. (2010). “A Survey on Transfer Learning.” In: IEEE Trans. on Knowl. and Data Eng., 22(10). doi:10.1109/TKDE.2009.191

2007

• (Jiang & Zhai, 2007) ⇒ Jing Jiang, and ChengXiang Zhai. (2007). “Instance Weighting for Domain Adaptation in NLP.” In: Proceedings of ACL (ACL 2007).
  • QUOTE: Domain adaptation is an important problem in natural language processing (NLP) due to the lack of labeled data in novel domains. In this paper, we study the domain adaptation problem from the instance weighting perspective. We formally analyze and characterize the domain adaptation problem from a distributional view, and show that there are two distinct needs for adaptation, corresponding to the different distributions of instances and classification functions in the source and the target domains. We then propose a general instance weighting framework for domain adaptation. Our empirical results on three NLP tasks show that incorporating and exploiting more information from the target domain through instance weighting is effective.

---