RL tutorial example error

[pogo85]

Problem location

https://www.mlpack.org/doc/stable/doxygen/rltutorial.html

Description of problem

When I try to run the example using the custom FFN network, it throws:
error: 'class mlpack::ann::FFN<mlpack::ann::MeanSquaredError<>, mlpack::ann::GaussianInitialization>' has no member named 'ResetNoise'
but works when using SimpleDQN class.

[zoq]

Thanks for the info, the FFN does not implement the ResetNoise method, I'll update the tutorial.

[heisenbuug]

Heyy @zoq
What's the solution here? Do we implement the method?

[gunnxx]

@heisenbuug
QLearning expects an object which has ResetNoise method (e.g. SimpleDQN, DuelingDQN) because it calls the method during training. You are passing an FFN object (which has no ResetNoise method) to QLearning. The solution is to wrap your FFN object to a DQN object.

FFN<MeanSquaredError<>, GaussianInitialization> network(
        MeanSquaredError<>(), GaussianInitialization(0, 0.001));
network.Add<Linear<>>(4, 128);
network.Add<ReLULayer<>>();
network.Add<Linear<>>(128, 128);
network.Add<ReLULayer<>>();
network.Add<Linear<>>(128, 2);
SimpleDQN<> model(network);