IFT6760-B 1
PackNet: Adding Multiple Tasks to a
Single Network by Iterative Pruning
Arun Mallya, Svetlana Lazebnik
Presenters: Mostafa ElAraby, Dishank Bansal
Outline
IFT6760-B 2
Introduction
IFT6760-B 3
Key Theme
Use redundant parameters in network to learn new tasks, instead of removing them
Related Works
IFT6760-B 4
Related Works
IFT6760-B 5
Approach
IFT6760-B 6
Approach: Overhead & Inference
IFT6760-B 7
Approach: Pruning
IFT6760-B 8
Datasets
Experiments
IFT6760-B 9
Training Setting
IFT6760-B 10
Baselines
IFT6760-B 11
Learn without Forgetting (LWF)
03
Individual Networks
02
Classifier
01
Joint training
04
Fine Grained Tasks Vanilla VGG-16
IFT6760-B 12
Large Datasets Vanilla VGG-16
IFT6760-B 13
Change In Error LWF vs PackNet
IFT6760-B 14
Results on VGG-16 with BatchNorm
IFT6760-B 15
Results on ResNet-50
IFT6760-B 16
Results on DenseNet-121
IFT6760-B 17
Effect of Training Order
IFT6760-B 18
Effect of Pruning Ratios
IFT6760-B 19
Sharing Parameters (biases)
IFT6760-B 20
Effect of training all layers
IFT6760-B 21
Filter based pruning
IFT6760-B 22
Conclusion
IFT6760-B 23
Limitation of the Work
IFT6760-B 24
The number of possible tasks is limited by the capacity of the network, at some point the parameters left for training won’t be enough for upcoming tasks.
1.
Limitation of the Work
IFT6760-B 25
Might not work with efficient networks where there are no redundant parameters.
Number of task that can be added is correlated with tasks similarities, but no experiments are shown for that. For example, if all the task are completely different then a network might learn less task compared to if tasks are related because since capacity is constant, related task might make more use of previous learned weigths than unrelated tasks.
2.
3.
Limitations of the Work
IFT6760-B 26
No adaptive pruning.
4.
Improvements
IFT6760-B 27