Dataset supporting publication "Realisation of early-exit dynamic neural networks on reconfigurable hardware"

This dataset supports the publication " Realisation of Early-Exit Dynamic Neural Networks on Reconfigurable Hardware " to be published in the IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems.

This dataset contains:
- 'Fig5a.csv': Data supporting Fig. 5 (a): Experimental results comparing average Execution Time per sample. Average values are calculated based on each exit point trigger rate (ai).
- 'Fig5b.csv': Data supporting Fig. 5 (b): Experimental results comparing average Energy Consumption per sample. Average values are calculated based on each exit point trigger rate (ai).
- 'Fig6a.csv': Data supporting Fig. 6 (a): Comparison of the pipeline and parallel designs over average execution time across early-exit LeNet-5, AlexNet, VGG19 and ResNet32.
- 'Fig6b.csv': Data supporting Fig. 6 (b): Comparison of the pipeline and parallel designs over average energy consumption across early-exit LeNet-5, AlexNet, VGG19 and ResNet32.
- 'Fig6c.csv': Data supporting Fig. 6 (c): Comparison of the pipeline and parallel designs over average data movement across early-exit LeNet-5, AlexNet, VGG19 and ResNet32.
- 'Table2.csv': Data supporting TABLE II: Execution Time (ms).
- 'Table3.csv': Data supporting TABLE III: Energy Consumption (mJ), with (w/ EE) and without (w/oEE) early exits.
- 'Table4.csv': Data supporting TABLE IV: Performance Comparison With Exisiting Implementations.
- 'Fig7a.csv': Data supporting Fig. 7 (a): On an 3 point early-exit Resnet-32 shows each exit’s trigger rate for different Confidence Thresholds.
- 'Fig7b.csv': Data supporting Fig. 7 (b): On an 3 point early-exit Resnet-32 the percentage difference of parallel over pipeline approaches over Energy and Time for different Confidence Thresholds.

Related projects: Engineering and Physical Sciences Research Council (EPSRC) under EP/S030069/1

Licence: CC BY 4.0