def test_simpleAgent(agentType, expected_monitor_results):
#start agent network server
agentNetwork = AgentNetwork(dashboard_modules=False)
#init agents by adding into the agent network
simple_agent = agentNetwork.add_agent(agentType=agentType)
monitor_agent_1 = agentNetwork.add_agent(agentType=MonitorAgent)
#shorten n wait loop time
simple_agent.init_agent_loop(0.01)
#connect agents
agentNetwork.bind_agents(simple_agent, monitor_agent_1)
# set all agents states to "Running"
agentNetwork.set_running_state()
time.sleep(test_timeout)
# test to see if monitor agents have received the correct data
assert str(
monitor_agent_1.get_attr('memory')) == str(expected_monitor_results)
# shutdown agent network
agentNetwork.shutdown()
time.sleep(3)
def test_remove_agent():
#start agent network server
agentNetwork = AgentNetwork(dashboard_modules=False)
#init agents by adding into the agent network
dummy_agent1 = agentNetwork.add_agent(agentType=AgentMET4FOF)
dummy_agent2 = agentNetwork.add_agent(agentType=AgentMET4FOF)
dummy_agent3 = agentNetwork.add_agent(agentType=AgentMET4FOF)
dummy_agent4 = agentNetwork.add_agent(agentType=AgentMET4FOF)
agentNetwork.bind_agents(dummy_agent1, dummy_agent2)
agentNetwork.bind_agents(dummy_agent1, dummy_agent3)
agentNetwork.bind_agents(dummy_agent1, dummy_agent4)
agentNetwork.bind_agents(dummy_agent4, dummy_agent2)
agentNetwork.bind_agents(dummy_agent3, dummy_agent4)
agentNetwork.bind_agents(dummy_agent2, dummy_agent4)
agentNetwork.remove_agent(dummy_agent1)
agentNetwork.remove_agent(dummy_agent2)
agentNetwork.remove_agent(dummy_agent3)
agentNetwork.remove_agent(dummy_agent4)
time.sleep(2)
assert len(agentNetwork.agents()) == 0
agentNetwork.shutdown()
def test_send_plot():
# start agent network server
agentNetwork = AgentNetwork(dashboard_modules=False)
# init agents
gen_agent = agentNetwork.add_agent(agentType=GeneratorAgent)
monitor_agent = agentNetwork.add_agent(agentType=MonitorAgent)
agentNetwork.bind_agents(gen_agent, monitor_agent)
gen_agent.dummy_send_graph()
time.sleep(3)
assert monitor_agent.get_attr('plots')['GeneratorAgent_1']
agentNetwork.shutdown()
def test_zema_emc_lda_agents():
np.random.seed(100)
#start agent network server
agentNetwork = AgentNetwork(dashboard_modules=False)
#init agents by adding into the agent network
datastream_agent = agentNetwork.add_agent(agentType=DataStreamAgent)
train_test_split_agent = agentNetwork.add_agent(
agentType=TrainTestSplitAgent)
fft_bfc_agent = agentNetwork.add_agent(agentType=FFT_BFCAgent)
pearson_fs_agent = agentNetwork.add_agent(
agentType=Pearson_FeatureSelectionAgent)
lda_agent = agentNetwork.add_agent(agentType=LDA_Agent)
evaluator_agent = agentNetwork.add_agent(agentType=EvaluatorAgent)
monitor_agent = agentNetwork.add_agent(agentType=MonitorAgent)
#init parameters
datastream_agent.init_parameters(ZEMA_DataStream())
train_test_split_agent.init_parameters(train_ratio=0.8)
#bind agents
agentNetwork.bind_agents(datastream_agent, train_test_split_agent)
agentNetwork.bind_agents(train_test_split_agent, fft_bfc_agent)
agentNetwork.bind_agents(fft_bfc_agent, pearson_fs_agent)
agentNetwork.bind_agents(pearson_fs_agent, lda_agent)
agentNetwork.bind_agents(lda_agent, evaluator_agent)
#bind to monitor agents
agentNetwork.bind_agents(fft_bfc_agent, monitor_agent)
agentNetwork.bind_agents(pearson_fs_agent, monitor_agent)
agentNetwork.bind_agents(lda_agent, monitor_agent)
agentNetwork.bind_agents(evaluator_agent, monitor_agent)
#trigger datastream to send all at once
datastream_agent.send_all_sample()
time.sleep(timeout_wait)
assert lda_agent.get_attr('lda_test_score') == 0.8204924543288324
agentNetwork.shutdown()
def on_received_message(self, message):
if message['senderType'] == "Predictor":
data = message['data']
y_pred = data['y_pred']
y_true = data['y_true']
res = self.compute_accuracy(y_pred=y_pred, y_true=y_true)
self.send_output(res)
if __name__ == '__main__':
# start agent network server
agentNetwork = AgentNetwork()
# init agents
gen_agent = agentNetwork.add_agent(agentType=DataStreamAgent)
trainer_agent = agentNetwork.add_agent(agentType=Trainer)
predictor_agent = agentNetwork.add_agent(agentType=Predictor)
evaluator_agent = agentNetwork.add_agent(agentType=Evaluator)
monitor_agent_1 = agentNetwork.add_agent(agentType=MonitorAgent)
monitor_agent_2 = agentNetwork.add_agent(agentType=MonitorAgent)
gen_agent.init_parameters(stream=SineGenerator(),
pretrain_size=1000,
batch_size=1)
trainer_agent.init_parameters(ml_model=HoeffdingTree())
# connect agents : We can connect multiple agents to any particular agent
# However the agent needs to implement handling multiple input types
agentNetwork.bind_agents(gen_agent, trainer_agent)
agentNetwork.bind_agents(gen_agent, predictor_agent)
agentNetwork.bind_agents(trainer_agent, predictor_agent)
# classifier accuracy - user defined
def compute_accuracy(self, y_pred, y_true):
res = y_pred == y_true
num_accurate = [1 if y == True else 0 for y in res]
accuracy = np.sum(num_accurate) / len(num_accurate) * 100
return accuracy
if __name__ == '__main__':
#start agent network
agentNetwork = AgentNetwork()
#add agents
data_stream_agent_1 = agentNetwork.add_agent(agentType=DataStreamAgent)
ml_agent_hoeffdingTree = agentNetwork.add_agent(agentType=ML_Model)
ml_agent_neuralNets = agentNetwork.add_agent(agentType=ML_Model)
monitor_agent_1 = agentNetwork.add_agent(agentType=MonitorAgent)
#init parameters
data_stream_agent_1.init_parameters(stream=WaveformGenerator(),
pretrain_size=1000,
batch_size=100)
ml_agent_hoeffdingTree.init_parameters(ml_model=HoeffdingTree())
ml_agent_neuralNets.init_parameters(ml_model=NaiveBayes())
#connect agents
agentNetwork.bind_agents(data_stream_agent_1, ml_agent_hoeffdingTree)
agentNetwork.bind_agents(data_stream_agent_1, ml_agent_neuralNets)
agentNetwork.bind_agents(ml_agent_hoeffdingTree, monitor_agent_1)
class SineGeneratorAgent(AgentMET4FOF):
def init_parameters(self):
self.stream = SineGenerator()
def agent_loop(self):
if self.current_state == "Running":
sine_data = self.stream.next_sample() #dictionary
self.send_output(sine_data['x'])
if __name__ == '__main__':
# start agent network server
agentNetwork = AgentNetwork()
# init agents
gen_agent = agentNetwork.add_agent(agentType=SineGeneratorAgent)
math_agent = agentNetwork.add_agent(agentType=MathAgent)
multi_math_agent = agentNetwork.add_agent(agentType=MultiMathAgent)
monitor_agent = agentNetwork.add_agent(agentType=MonitorAgent)
# connect agents : We can connect multiple agents to any particular agent
agentNetwork.bind_agents(gen_agent, math_agent)
agentNetwork.bind_agents(gen_agent, multi_math_agent)
# connect
agentNetwork.bind_agents(gen_agent, monitor_agent)
agentNetwork.bind_agents(math_agent, monitor_agent)
agentNetwork.bind_agents(multi_math_agent, monitor_agent)
# set all agents states to "Running"
agentNetwork.set_running_state()
class SineGeneratorAgent(AgentMET4FOF):
def init_parameters(self):
self.stream = SineGenerator()
def agent_loop(self):
if self.current_state == "Running":
sine_data = self.stream.next_sample() #dictionary
self.send_output(sine_data['x'])
if __name__ == '__main__':
#start agent network server
agentNetwork = AgentNetwork(dashboard_modules=False)
#init agents by adding into the agent network
gen_agent = agentNetwork.add_agent(agentType=SineGeneratorAgent)
_logger_agent = agentNetwork.add_agent(agentType=_Logger)
monitor_agent = agentNetwork.add_agent(agentType=MonitorAgent)
gen_agent.set_logger(_logger_agent)
#connect agents by either way:
# 1) by agent network.bind_agents(source,target)
agentNetwork.bind_agents(gen_agent, monitor_agent)
# 2) by the agent.bind_output()
gen_agent.bind_output(monitor_agent)
# set all agents states to "Running"
agentNetwork.set_running_state()
#Each agent has internal current_state which can be used as a switch by the AgentNetwork
class SineGeneratorAgent(AgentMET4FOF):
def init_parameters(self):
self.stream = SineGenerator()
def agent_loop(self):
if self.current_state == "Running":
sine_data = self.stream.next_sample() #dictionary
self.send_output(sine_data['x'])
if __name__ == '__main__':
#start agent network server
agentNetwork = AgentNetwork()
#init agents by adding into the agent network
gen_agent = agentNetwork.add_agent(agentType=SineGeneratorAgent)
monitor_agent = agentNetwork.add_agent(agentType=MonitorAgent)
#connect agents by either way:
# 1) by agent network.bind_agents(source,target)
agentNetwork.bind_agents(gen_agent, monitor_agent)
# 2) by the agent.bind_output()
gen_agent.bind_output(monitor_agent)
# set all agents states to "Running"
agentNetwork.set_running_state()
def on_received_message(self, message):
"""
Checks for message['channel'] and handles them separately
Acceptable channels are "cosine" and "sine"
"""
if message['channel'] == "cosine":
minus_data = minus(message['data'], self.minus_param)
self.send_output({'cosine_minus': minus_data})
elif message['channel'] == 'sine':
plus_data = plus(message['data'], self.plus_param)
self.send_output({'sine_plus': plus_data})
if __name__ == '__main__':
# start agent network server
agentNetwork = AgentNetwork()
# init agents
gen_agent = agentNetwork.add_agent(agentType=MultiGeneratorAgent)
multi_math_agent = agentNetwork.add_agent(agentType=MultiOutputMathAgent)
monitor_agent = agentNetwork.add_agent(agentType=MonitorAgent)
# connect agents : We can connect multiple agents to any particular agent
# However the agent needs to implement handling multiple inputs
agentNetwork.bind_agents(gen_agent, multi_math_agent)
agentNetwork.bind_agents(gen_agent, monitor_agent)
agentNetwork.bind_agents(multi_math_agent, monitor_agent)
# set all agents states to "Running"
agentNetwork.set_running_state()
from examples.ZEMA_EMC.zema_agents import TrainTestSplitAgent, FFT_BFCAgent, Pearson_FeatureSelectionAgent, LDA_Agent, Regression_Agent, EvaluatorAgent
import examples.ZEMA_EMC.zema_agents
import numpy as np
np.random.seed(100)
if __name__ == '__main__':
#start agent network server
agentNetwork = AgentNetwork(dashboard_modules=[
examples.ZEMA_EMC.zema_datastream, examples.ZEMA_EMC.zema_agents
],
log_filename=False)
#init agents by adding into the agent network
datastream_agent = agentNetwork.add_agent(agentType=DataStreamAgent)
train_test_split_agent = agentNetwork.add_agent(
agentType=TrainTestSplitAgent)
fft_bfc_agent = agentNetwork.add_agent(agentType=FFT_BFCAgent)
pearson_fs_agent = agentNetwork.add_agent(
agentType=Pearson_FeatureSelectionAgent)
lda_agent = agentNetwork.add_agent(agentType=LDA_Agent)
bayesianRidge_agent = agentNetwork.add_agent(agentType=Regression_Agent,
name="BayesianRidge_Agent")
randomForest_agent = agentNetwork.add_agent(agentType=Regression_Agent,
name="RandomForest_Agent")
evaluator_agent = agentNetwork.add_agent(agentType=EvaluatorAgent)
monitor_agent = agentNetwork.add_agent(agentType=MonitorAgent)
#init parameters