Cleyber B. dos Reis1, Victor Hugo L.Lopes2, Antonio Oliveira-JR2
1Universidade Federal de Goiás - Goiânia, GO - Brazil
2Instituto Federal de Goiás - Inhumas, GO - Brazil
cleyber.bezerra@discente.ufg.br, victor.lopes@ifg.edu.br, antonio@inf.ufg.br
The repository contains all the development related to the master's dissertation entitled "Split Learning as an enabler of wireless networks for future generations". Being developed through the Academic Master's Course in Computer Science - Goiânia, by the Institute of Informatics (INF) of the Federal University of Goiás (UFG).
Split Learning (SL) is an innovative and effective approach to address security and privacy concerns in the training of deep neural networks (DNN). This technique combines the protection of raw data with the division of the model between client devices and a central server, significantly reducing the risks of data leaks and cyberattacks. In addition, it enables the training of deep neural networks on devices with limited computational resources.
However, splitting the model leads to an increase in the communication flow between the distributed devices and the central server, which can not only generate a communication overhead in environments with computational constraints, but also negatively impact the training accuracy, compromising the final results of the model.
This paper covers the inference problem of improving accuracy. Through a case study of integrating (ns3-ai) with distributed Split Learning to train a Convolutional Neural Network (CNN) and MNIST dataset. The NS3-LENA simulator used the characteristics of a B5G network environment with mobile devices (UE) and a gNB (5G access module).
In this integrated scenario, network experiments were simulated with distance variations between 200 mt. With powers of 13 and 26 dBm and loss exponents of 2, 3 and 4 dB. Based on the network output results, with regard to latency, a policy was defined that values above 10 seconds are considered timeouts and are not included in machine learning experiments. With the objective of training and testing the split learning model, the impacts of changes in the network simulation on training accuracy were observed.
The methods defined as baselines for our proposal are: (1) the use of the synchronous algorithm in the training of the Split Learning Vanilla model and (2) the definition of the training based on events provided by the simulation in NS3, following the latency policy established in 15 distinct seeds for each exponent. In the simulation, six (06) mobile devices and the base station are considered, with the policy defined for network latency stipulating nodes with latency below 23 seconds.
Demonstrations of results within the scope of network simulation, in two different scenarios, graphically presenting: latencies, througput, packet loss rates and energy consumption.
Fig. 1. Latency.
Fig. 2. Throughput.
Figures 1 and 2 show the latency and throughput results in the network simulation.
Fig. 3. Packet Loss.
Fig. 4. Energy Consumed .
Figures 3 and 4 show the results of packet loss and energy consumption in the network simulation.
Fig. 5. Accuracy.
Figure 5 shows the accuracy result in training.
-
GNU (>=8.0.0)
command to know the version of KERNEL, GCC and GNU Binutils in the terminal
cat /proc/version
-
GCC (>=11.4.0)
commands to know the GCC version in the terminal.
gcc --version ls -l /usr/bin/gcc* -
CMAKE (>=3.24)
command to know the version of CMAKE in the terminal.
cmake --version
-
python (>=3.11.5)
commands to know the version of PYTHON in the terminal.
python --version python3 --version
INSTALL GCC AND MAKE
sudo apt update
sudo apt install build-essential
sudo apt install gcc-10 g++-10 sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-10 100
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-10 100 gcc --version
make --versionINSTALl THE CMAKE
download the file in the desired version: https://cmake.org/download/
extract it, access the folder and run the command:
./bootstrap && make && sudo make installINSTALL THE GIT
sudo apt-get install git -yINSTALL THE NS-3-DEV
git clone https://gitlab.com/nsnam/ns-3-dev.git
git checkout -b ns-3.43-release ns-3.43CONFIGURE AND COMPILE NS-3
In the ns-3-dev folder use the commands:
./ns3 configure --enable-examples --enable-tests
./ns3 buildINSTALL THE NR-LENA
sudo apt-get install libc6-dev
sudo apt-get install sqlite sqlite3 libsqlite3-dev
sudo apt-get install libeigen3-dev cd contrib
git clone https://gitlab.com/cttc-lena/nr.git
cd nr
git checkout -b 5g-lena-v3.1.y origin/5g-lena-v3.1.yINSTALL THE NS3-IA
sudo apt install python3-pip
pip install tensorflow==2.17.0
pip install cloudpickle==1.2.0
pip install pyzmq
pip install protobuf==3.20.3
pip install tensorflow-estimator==2.15.0
pip install tensorboard==2.17.0
pip install numpy==1.18.1
pip install Keras==3.2.0
pip install Keras-Applications==1.0.8
pip install Keras-Preprocessing==1.1.2
pip install matplotlib==3.3.2
pip install psutil==5.7.2 cd contrib/
git clone https://github.com/hust-diangroup/ns3-ai.git
cd ns3-ai/py_interface
pip3 install . --userENABLE MODULES
./ns3 configure --enable-modules=nr,internet-apps,flow-monitor,config-store,buildings,applications,network,core,wifi,energy,spectrum,propagation,mobility,antennaDOWNLOAD PROJECT
inside the ns-3-dev/scratch folder use the commands:
git clone https://github.com/cleyber-bezerra/SplitLearning-B5G.gitEXECUTE PROJECT
inside the Split Learning folder run the script file.
./simulator_ns3.sh