public bool doesSlimRnnSolveTask(SlimRnn slimRnn, bool mustHalt, double tLim)
{
// NOTE< we ignore must halt because we hardcoded to check for termination >
instrumentationNewSolveTrial();
// simulate slimRnn in environment and correctness demonstration
// we don't have to show that the old solver doesn't solve the task because we did the step already in searchForCandidateProgram()
// we just have to show that the new solver can solve the new task and all old tasks
IList <SlimRnnNeuronWithWeight> traceResult;
double requiredTime;
bool wasTerminated;
{ // test new task
slimRnn.world = new TaskSlimRnnWorld(currentTask);
currentTask.resetState();
slimRnn.t_lim = tLim;
slimRnn.spread(out traceResult, out requiredTime, out wasTerminated);
bool thisTaskSolvedAndTerminated = wasTerminated && currentTask.wasTaskSolved;
if (!thisTaskSolvedAndTerminated)
{
return(false);
}
}
// test all old tasks
foreach (ITask iOldTask in allExistingTasks)
{
slimRnn.world = new TaskSlimRnnWorld(iOldTask);
iOldTask.resetState();
slimRnn.t_lim = tLim;
slimRnn.spread(out traceResult, out requiredTime, out wasTerminated);
bool thisTaskSolvedAndTerminated = wasTerminated && iOldTask.wasTaskSolved;
if (!thisTaskSolvedAndTerminated)
{
return(false);
}
}
return(true);
}
public bool doesSlimRnnSolveTask(SlimRnn slimRnn, bool mustHalt, double tLim)
{
IList <SlimRnnNeuronWithWeight> trace;
slimRnn.t_lim = tLim;
double time;
bool wasTerminated;
slimRnn.spread(out trace, out time, out wasTerminated);
return(true);
}
// tests if the network successfully calculates a complex formula
public static void testCalculation()
{
SlimRnn rnn = new SlimRnn();
rnn.neurons.Add(SlimRnnNeuron.makeInputNeuron()); // [0] input
rnn.neurons.Add(SlimRnnNeuron.makeInputNeuron()); // [1] input
rnn.neurons.Add(SlimRnnNeuron.makeInputNeuron()); // [2] input
rnn.neurons.Add(new SlimRnnNeuron(SlimRnnNeuron.EnumType.ADDITIVE)); // [3]
rnn.neurons.Add(new SlimRnnNeuron(SlimRnnNeuron.EnumType.MULTIPLICATIVE)); // [4] output
rnn.neurons.Add(new SlimRnnNeuron(SlimRnnNeuron.EnumType.ADDITIVE)); // [5] termination
rnn.outputNeuronsStartIndex = 4;
rnn.numberOfOutputNeurons = 1;
rnn.terminatingNeuronIndex = 5;
rnn.terminatingNeuronThreshold = 0.5f;
rnn.numberOfInputNeurons = 3;
rnn.t_lim = 5; // are actually less steps but fine for testing
rnn.world = new TestWorld();
rnn.neurons[0].outNeuronsWithWeights.Add(new SlimRnnNeuronWithWeight(rnn.neurons[0], rnn.neurons[3], 0.5f, 0));
rnn.neurons[1].outNeuronsWithWeights.Add(new SlimRnnNeuronWithWeight(rnn.neurons[1], rnn.neurons[3], 0.81f, 0));
rnn.neurons[2].outNeuronsWithWeights.Add(new SlimRnnNeuronWithWeight(rnn.neurons[2], rnn.neurons[4], 1.5f, 0));
rnn.neurons[3].outNeuronsWithWeights.Add(new SlimRnnNeuronWithWeight(rnn.neurons[3], rnn.neurons[4], 0.5f, 0));
rnn.neurons[4].outNeuronsWithWeights.Add(new SlimRnnNeuronWithWeight(rnn.neurons[4], rnn.neurons[5], 5000.0f, 0)); // terminate if output is set
rnn.initializeNeurons();
IList <SlimRnnNeuronWithWeight> trace;
double time;
bool wasTerminated;
rnn.spread(out trace, out time, out wasTerminated);
Debug.Assert(wasTerminated); // must terminate
}