Light

mathphyssim / fermi_rl_paper Goto Github PK

The repo for the FERMI FEL paper using model-based and model-free reinforcement learning methods to solve a particle accelerator operation problem.

License: MIT License

TeX 42.42% Python 57.58%

accelerators acclerator-physics ae-dyna applied-reinforcement-learning fel model-based-reinforcement-learning model-free-rl naf reinforcement-learning reinforcement-learning-algorithms trpo

fermi_rl_paper's Introduction

👋 Hi, I’m Simon
👀 I’m interested in state of the art methods for societal benefit, meaning there should be a real-world application.
🌱 I’m currently applying and learning many aspects of science as stochastic series, deep reinforcement learning and many more...leading a reinforcement learning team at the University in Salzburg. Before I spend several years at CERN.
💞️ I’m collaboration and look to collaborate on industry applications, accelerator physics, medical research and fundamendal research in general.
📫 How to reach me simon.hirlaender(@)plus.ac.at

fermi_rl_paper's People

Stargazers

Watchers

Forkers

leandergrech archcosmos

fermi_rl_paper's Issues

Problem with NormaliseEnv

NormalizeEnv.scale_state_env() in [FERMI_RL_PAPER/naf2_new.py] produces only NaNs and Infs

Example:
env = gym.make('MountainCarContinuous-v0')
nenv = NormalizeEnv(env)
print(nenv.scale_state_env(env.observation_space.low))

Output:
[-inf -inf]
FERMI_RL_PAPER/naf2_new.py:123: RuntimeWarning: divide by zero encountered in true_divide return (2 * ob - (self.obs_high + self.obs_low)) / scale

TypeError when used in Colab

FERMI_RL_Paper/naf2_new.py

Lines 188 to 189 in 0e5c130

    
           V = tf.scalar_mul(-1, self.fc(h, 1, activation=tf.nn.leaky_relu, 
        
                                         kernel_initializer=self.kernel_initializer, name='V'))

The above line causes:
TypeError: Input 'y' of 'Mul' Op has type float64 that does not match type int32 of argument 'x'.

Target policy smoothing noise out of predict function

FERMI_RL_Paper/naf2_new.py

Lines 564 to 569 in cd2545c

    
           action = self.noise_function(np.squeeze(model.get_action([state])),self.idx_episode) 
        
           if self.q_smoothing_clip is None: 
        
               return_value = np.clip(action, -1, 1) 
        
           else: 
        
               return_value = np.clip(action + np.clip(self.sigma * np.random.randn(self.action_size), 
        
                                                       -self.q_smoothing_clip, self.q_smoothing_clip), -1, 1)

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.

	V = tf.scalar_mul(-1, self.fc(h, 1, activation=tf.nn.leaky_relu,
	kernel_initializer=self.kernel_initializer, name='V'))

	action = self.noise_function(np.squeeze(model.get_action([state])),self.idx_episode)
	if self.q_smoothing_clip is None:
	return_value = np.clip(action, -1, 1)
	else:
	return_value = np.clip(action + np.clip(self.sigma * np.random.randn(self.action_size),
	-self.q_smoothing_clip, self.q_smoothing_clip), -1, 1)