HyperQuest

For Two Layerl Network

×

Instructions

  • You will be provided a specific dataset. Your goal is to train a two-layer network for classification on the dataset, and obtain as high validation accuracy (~0.5) as you can

  • In the first stage, you will choose the initial network configuration

  • In the second stage, you will run the network and check whether your configuration is good

Dataset Statistics

  • Classes: 10
  • Input data size: (3, 32, 32)
  • Examples per split: Train (8500), Val (1500)
Control Pad

3 5 10 50 150

10 50 100 200 400

0.004 0.003 0.001 0.0005 0.0001

0 0.1 0.5 1 10
validation accuracy is 0.