Assume that you have an existing trained reinforcement learning agent. For this example, load the trained agent from Train DDPG Agent to Control Double Integrator System.

load('DoubleIntegDDPG.mat','agent') 

Obtain the critic representation from the agent.

critic = getCritic(agent);

Obtain the learnable parameters from the critic.

params = getLearnableParameters(critic);

Modify the parameter values. For this example, simply multiply all of the parameters by 2.

modifiedParams = cellfun(@(x) x*2,params,'UniformOutput',false);

Set the parameter values of the critic to the new modified values.

critic = setLearnableParameters(critic,modifiedParams);

Set the critic in the agent to the new modified critic.

agent = setCritic(agent,critic);

Assume that you have an existing reinforcement learning agent, agent. For this example, load the trained agent from Train AC Agent to Balance Cart-Pole System:

load('MATLABCartpoleAC.mat')

Further, assume that this agent has a critic representation that contains the following shallow neural network structure with two layers:

oldCriticNetwork = [
        imageInputLayer([4 1 1],'Normalization','none','Name','state')
        fullyConnectedLayer(1,'Name','CriticFC')];

Create the new network with an additional 5-neurons fully connected hidden layer:

newCriticNetwork = [
        oldCriticNetwork(1)
        fullyConnectedLayer(5,'Name','hidden');
        oldCriticNetwork(2)];

Create the critic using the new network with the additional fully connected layer.

critic = rlValueRepresentation(newCriticNetwork,getObservationInfo(agent),'Observation',{'state'});

Set the critic representation of the agent to the new critic.

newAgent = setCritic(agent,critic);