7. Experiments and Ablation Studies We train the model on Google Co Laboratory Pro. This gave us a faster GPU (T4, P100/V100) with 25 GB GPU memory. The training was stopped after 10 epochs so as to compare performances. Batch sizes 4 and 8 were used to meet the allocated memory limits. The learning rate was 0.0001. The learning rate exponentially decayed for successive epochs. The filter weights were randomly initialized. The loss optimizer was ADAM.
Is this you? As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.
Claim your profile