This is a repository of the experiment code supporting the paper Reinforcement Learning for Architecture Search by Network Transformation.
The discovered top networks along with their weights are provided under the folder model. For checking these networks, please run under the folder of code:
$ python3 main.py "../model/top_net1" --test "cifar10"
and you will get
test acc: 0.9325921535491943
Similarly run
$ python3 main.py "../model/top_net2" --test "cifar10"
and you will get
test acc: 0.9430088400840759