资 源 简 介
MT-factor performs multi-task learning for the state factors (e.g., interacting objects) in model-based Reinforcement Learning. MT-factors shares data between similar state factors and, thus, allows the agent to make accurate plans in high-dimensional problems.
The main code of the implementation is written in R with Fitted Q-learning part written in Java. Please see the Installation Guide for how to set it up.
文 件 列 表
mt-factors
boat_results_success.eps
FQServer2.class
boat1_test_intmodel.R
setup.txt
rjava-functions.R
boat1_test_models.R
rsge-test-new.R
boat1_test_fqfail3.R
FQClient.class
boat1_run_sges.R
rsge-test.R
boat1_test_fqfail.R
FittedQ.class
boat1_run_sge.R
boat1_test_mpi3.R
Matrix.java
boat1_test_mpi4.R
boat1_test_fqfail2.R
boat1_test_qfun.R
boat1_test_sge.R
FQServer2.java
boat1_test_qiter.R
boat1_test_mpi.R
boat1_fqs_client.R
ExtraTrees.java
ExtraTrees.class
boat_results_value.pdf
fittedq-java
neg_ge_plotting.R
boat1_test_probfun.R
boat1_test_data_speed.R
boat1_model.R
FittedQ.java
BinaryTree.java
boat1_test_safeat.R
boat1_test_holdout.R
boat1_test_features.R
rjava-test.R
boat_results_value.eps
FQServer2$ClientConn.class
boat1_test_mpolicy.R
FQClient.java
route.pdf
do_plotting_ge.R
Matrix.class
BinaryTree.class
boat_plot_results.R
boat1_example.R
boat1_test_mpi2.R
boat1_test_qitermodel.R
boat1_test.R
gwf1_do_plotting.R
install.txt
