felixlinker / ib_fpsrl Goto Github PK
View Code? Open in Web Editor NEWPrototype to benchmark the fuzzy particle swarm reinforcement learning (Hein, Hentschel, Runkler, Udluft 2017) algorithm on the industrial benchmark
Prototype to benchmark the fuzzy particle swarm reinforcement learning (Hein, Hentschel, Runkler, Udluft 2017) algorithm on the industrial benchmark
When applying a partial policy, sometimes one element of the activation vector becomes zero which leads to a division by zero error. The paper mentions that this is not possible so I guess the issue comes down to badly selected bounds for the policy vector.
output = np.sum(actions) / np.sum(activations) # shape ()
Traceback (most recent call last):
File "ib_policy.py", line 229, in <module>
generate_policy(*parse_cfg_args(load_policy_cfg))
File "ib_policy.py", line 217, in generate_policy
evaluation.evaluate_policy(cfg, policy_weights)
File "/home/fl34gufe/Documents/IB_FPSRL/eval_policy.py", line 75, in evaluate_policy
random_cost = np.mean(evaluater(random_policies))
File "/home/fl34gufe/Documents/IB_FPSRL/ib_policy.py", line 124, in __call__
costs = self.cost_function(batch)
File "/home/fl34gufe/Documents/IB_FPSRL/eval_policy.py", line 32, in __call__
dtype=float
File "/home/fl34gufe/Documents/IB_FPSRL/eval_policy.py", line 31, in <lambda>
map(lambda t: self.batch_applier(*t), zip(setpoints, action_batch)),
File "/home/fl34gufe/Documents/IB_FPSRL/eval_policy.py", line 18, in __call__
_, (fuel, consumption) = list(map(env.__call__, trajectory))[-1]
File "/home/fl34gufe/Documents/IB_FPSRL/gen_dataset.py", line 78, in __call__
self.benchmark.step(action_vector)
File "/home/fl34gufe/Documents/IB_FPSRL/industrialbenchmark/industrial_benchmark_python/IDS.py", line 112, in step
self.updateGS()
File "/home/fl34gufe/Documents/IB_FPSRL/industrialbenchmark/industrial_benchmark_python/IDS.py", line 237, in updateGS
reward, domain, phi_idx, system_response = self.gsEnvironment.state_transition(self.gsEnvironment._dynamics.Domain(domain), phi_idx, self.gsEnvironment._dynamics.System_Response(system_response), effective_shift)
File "/home/fl34gufe/Documents/IB_FPSRL/industrialbenchmark/industrial_benchmark_python/goldstone/environment.py", line 39, in state_transition
domain, phi_idx, system_response = self._dynamics.state_transition(domain, phi_idx, system_response, position)
File "/home/fl34gufe/Documents/IB_FPSRL/industrialbenchmark/industrial_benchmark_python/goldstone/dynamics.py", line 62, in state_transition
domain = self._compute_domain(old_domain, position)
File "/home/fl34gufe/Documents/IB_FPSRL/industrialbenchmark/industrial_benchmark_python/goldstone/dynamics.py", line 88, in _compute_domain
return self.Domain(sign(position))
File "/u/fl34gufe/.local/share/virtualenvs/IB_FPSRL-bu2U3Xcm/lib64/python3.6/enum.py", line 291, in __call__
return cls.__new__(cls, value)
File "/u/fl34gufe/.local/share/virtualenvs/IB_FPSRL-bu2U3Xcm/lib64/python3.6/enum.py", line 533, in __new__
return cls._missing_(value)
File "/u/fl34gufe/.local/share/virtualenvs/IB_FPSRL-bu2U3Xcm/lib64/python3.6/enum.py", line 546, in _missing_
raise ValueError("%r is not a valid %s" % (value, cls.__name__))
ValueError: nan is not a valid Domain```
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.