The Universal Activation Function (UAF) is a unique approach designed to enhance machine learning models, as examined in the notable research by Yuen et al. in their paper published in Scientific Reports. This article will guide you step-by-step on how to set up and run UAF in both TensorFlow and PyTorch environments using Docker.

Prerequisites

• Docker: Ensure you have Docker installed on your machine. You can download it from Docker Installation Guide.
• Git: You’ll need Git to clone the repository. Make sure it’s installed on your system.

Getting the Code

To get started, you’ll need to pull the source code repository. Execute the following command in your terminal:

git clone https://github.com/SensorOrgNet/Universal_Activation_Function.git

Running the TensorFlow 2 Version

Follow these steps to run the UAF using TensorFlow:

1. Install the CUDA 11.2 container with the command below:

docker run --name UAF --gpus all -v home/username/UAF:workspace -w workspace -it nvcr.io/nvidia/cuda:11.2.0-cudnn8-devel-ubuntu20.04 bash

2. Once inside the container, update the package manager and install Python:

apt update
apt install python3-pip

3. Install TensorFlow and run the MLP with UAF for the MNIST dataset:

pip3 install tensorflow==2.7.0
cd Universal_Activation_Function
tensorflow python3 mnist_UAF.py

Running the PyTorch Version

Similarly, here’s how to execute the UAF in a PyTorch setting:

1. Install the CUDA 11.3 container:

docker run --name UAF --gpus all -v home/username/UAF:workspace -w workspace -it nvcr.io/nvidia/cuda:11.3.0-cudnn8-devel-ubuntu20.04 bash

2. Update packages and install Python:

apt update
apt install python3-pip

3. Install the necessary PyTorch packages:

pip3 install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio==0.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html
pip3 install torch-scatter -f https://data.pyg.org/whl/torch-1.10.0+cu113.html
pip3 install torch-sparse -f https://data.pyg.org/whl/torch-1.10.0+cu113.html
pip3 install torch-cluster -f https://data.pyg.org/whl/torch-1.10.0+cu113.html
pip3 install torch-spline-conv -f https://data.pyg.org/whl/torch-1.10.0+cu113.html
pip3 install torch-geometric

4. Run the CNN with UAF for the MNIST dataset:

cd Universal_Activation_Function
pytorch python3 mnist_UAF.py

5. Execute the GCN2 with UAF for the CORA dataset:

cd Universal_Activation_Function
pytorch python3 gcn2_cora_UAF.py 0

6. Run the PNA with UAF for the ZNC dataset:

cd Universal_Activation_Function
pytorch python3 pna_UAF.py 0

Understanding the Code through Analogy

Think of the Universal Activation Function like a Swiss Army knife for your machine learning models. Just as a Swiss Army knife has various tools that can be used for different purposes – a screwdriver for fixing something, scissors for cutting, or a bottle opener for enjoying a drink – the UAF can adapt to various tasks in ML. Each part of the UAF acts like a specific tool that can be utilized depending on what challenge your model faces, whether it’s prediction, classification, or feature transformation.

Troubleshooting

If you encounter any issues during the process, here are some troubleshooting tips:

• Docker Issues: Make sure Docker is running and that you have allocated enough resources (CPU, memory) to your containers.
• CUDA Compatibility Errors: Double-check that your CUDA installation matches the version required by TensorFlow or PyTorch.
• Package Installation Problems: If you face difficulties with pip installations, ensure that your internet connection is stable during the package downloads and installations.

For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

At fxis.ai, we believe that such advancements are crucial for the future of AI, as they enable more comprehensive and effective solutions. Our team is continually exploring new methodologies to push the envelope in artificial intelligence, ensuring that our clients benefit from the latest technological innovations.