Lifelong Learning with Dynamically Expandable Networks
- 울산과기대,한글 정리 ppt 및 후기s
한글 정리
영문 정리
정의 : able to learn continuously over time.
원리 : techniques like transfer learning are used
- where the model is trained on previous data
- and some features are used from that model to learn new data.
효과
- This is usually done to reduce the time required to train models from scratch.
- It is also used when the new data is sparse.
1. 구현 방법(간단버젼) : constantly fine-tuning the model based on newer data.
문제점 # 1
- However, if the new task is very different from the old tasks, the model will not be able to perform well on that new task, as features from the old task are not useful,
- e.g. if a model that is trained on a million images of animals, it will probably not work very well if it is fine-tuned on images of cars.
문제점 # 2
- after fine-tuning, the model may begin to perform the original task poorly (in this example, predicting animals).
- For example, the stripes on a zebra has a vastly different meaning than a striped T-shirt or a fence. Fine-tuning such a model will degrade its performance recognizing zebras.
2. 구현 방법 - Dynamically Expandable Networks
- 원리
- Train a model,
- and if it cannot predict very well, increase its capacity to learn.
- If a new task arrives that is vastly different from an existing task, extract whatever useful information you can from the old model and train a new model.
2.1 Techniques
- Selective retraining — Find the neurons that are relevant to the new task and retain them.
- Dynamic Network Expansion — If the model is unable to learn from step 1 (i.e. the loss is above a threshold value), increase the capacity of the model by adding more neurons.
- Network Split/Duplication — If some new models’ units have begun to change drastically, duplicate those weights, and retrain those duplicates, while keeping the old weights fixed.
In the above, figure t denotes task number. Thus, t-1 denotes the previous task, and t denotes the current task.