NeuralMAS

Neural Network and Multi-Agent System Hybrid Framework.

NeuralMAS is composed of a Koinè-DALI MAS, a Keras Neural Network and NeuralRedis, a module I made to manage Redis and the Dataset, so that the Dataset can be easily changed with another one.

The Neural Network

The Neural Network has been developed using Keras library, running on top of TensorFlow. See:

• https://keras.io/
• https://www.tensorflow.org/

Keras is very powerful and can run on top of Theano or CNTK too, but you can use a different technology of your choise insted of it.

The Multi-Agent System

The Multi-Agent System (MAS) has been developed using DALI, or better Koinè DALI, running on top of SICStus Prolog. See:

• https://github.com/agnsal/KOINE-DALI
• https://github.com/AAAI-DISIM-UnivAQ/DALI
• https://sicstus.sics.se/ You can choose to develop the MAS with a different technology instead of it.

The Communication

The communication channel between the Neural Network and the MAS is made using Redis. See:

• https://redis.io/

The Environment

You can use it on both Linux and Windows, and it is compatible with Docker too (so you can use it on distributed Systems too): See:

• https://github.com/agnsal/docker-PyRedis
• https://www.docker.com/

For more information, see NeuralMAS.pdf.

Instructions

1. To install Redis see: https://redis.io/
2. To install Keras and Tensorflow see: https://keras.io/ and https://www.tensorflow.org/
3. To install SISCtus Prolog see: https://sicstus.sics.se/
4. To install numpy and Keras dependencies:

  sudo pip3 install h5py
  sudo pip3 install numpy

5. To get NeuralMas:

  git clone https://github.com/agnsal/NeuralMAS.git

6. To get KOINE-DALI and its dependencies and include it in this project:

  cd NeuralMAS
  git clone https://github.com/agnsal/KOINE-DALI.git
  cd KOINE-DALI/DALI
  git clone https://github.com/AAAI-DISIM-UnivAQ/DALI.git
  mv DALI/* .
  rm -rf DALI
  cp -r KOINE-MAS-example KOINE-MAS-hospital
  rm -rf KOINE-MAS-hospital/mas
  mv ../../mas KOINE-MAS-hospital/

7. To join the trained net file pieces and decompress it:

  cd ../..
  cat LetterNet.h5* > LetterNet.h5.tar.gz
  tar -xvzf LetterNet.h5.tar.gz

8. To start the MAS:

  cd KOINE-DALI/DALI/KOINE-MAS-hospital/conf
  chmod u+rx makeconf.sh
  chmod u+rx startagent.sh
  cd ..
  sudo bash startmas.sh  # Or sudo bash startmasMary.sh if you prefer a single window MAS

The agents will say when they finish the handshake.

9. Open another terminal window to monitor Redis:

  redis-cli monitor

10. In another terminal window, to open the communication channel between the Neural Net and the MAS:

  cd NeuralMAS/KOINE-DALI
  python3 Redis2LINDA.py

11. In another terminal window, to start the Neural Net:

  cd NeuralMAS
  python3 BreastCancerNet.py

12. In another terminal window, to make some queries, about the dataset (for example):

  redis-cli
  rpush neuralNetQueryList [776715,3,1,1,1,3,2,1,1,1]
  rpush neuralNetQueryList [897471,4,8,8,5,4,5,10,4,1]

and so on. You can see the data obtained by the Net and by the MAS via the Redis monitor.

13. To stop the MAS just press a button.

Screenshots:

Meta:

Agnese Salutari – [email protected]

Distributed under the Apache 2.0 license. See LICENSE for more information.

https://github.com/agnsal/NeuralMAS