Tensor_Decomposition

Small collection of example scripts demonstrating tensor decomposition approach.

---

Step 1: Install Docker

Docker provides the virtual environment that runs the code.

1. Download Docker Desktop: Go to the Docker Desktop Official Website and download the installer for your operating system (Windows, Mac, or Linux).
2. Install and Run: Run the installer and follow the standard on-screen prompts.
3. Crucial: Once installation is complete, launch the Docker Desktop application and keep it running in the background.

---

Step 2: Open Your Terminal & Project Directory

1. Unzip and place this repository onto your machine.
2. Open your command line interface:
   • Windows: Open Command Prompt or PowerShell.
   • macOS / Linux: Open Terminal.
3. Navigate (cd) into your extracted project folder. For example:

   cd path/to/your/extracted-folder/

Step 3: Build the Docker environment

You fill find Dockerfile in that repo. Build the environment (might take a few minutes) with:

docker build -t nspod .

Step 4: Run the docker environment

Run the simulations and save outputs according to your machine.

• Windows command prompt :

docker run --rm -v "%cd%:/app" nspod --run "Single_wave"

• Windows power shell :

docker run --rm -v  "${PWD}:/app" nspod --run "Single_wave"

• mac/Linux terminal:

docker run --rm -v "$(pwd):/app" nspod --run "Single_wave"

The options for arguments set after --run

• Single_wave,
• Crossing_StraightCubic_waves,
• Crossing_sine_StraightCubic_waves,
• Wildlandfire_1d

Optimization and algorithm parameters are set inside each script for the corresponding example. Edit the scripts to change settings. The wildland fire data is available upon request.

Step 5: Saved files

• .npy files will be saved in the data/ folder and .png files will be stored in plots/ folder.

⚠️ System Requirements & Platform Note

"It Works on My Machine" (and on our compute cluster)

This codebase intermediately performs low-level Singular Value Decomposition (SVD) matrix operations.

If you are trying to run the Docker container via Docker Desktop on an Apple Silicon Mac (M1/M2/M3/M4), you will likely hit a low-level crash stemming from the linear algebra engine: ** On entry to SLASCL parameter number 4 had an illegal value

• Why? This is an active bug inside x86_64 virtualization/emulation layers (like QEMU) when translating single-precision AVX vector instructions to ARM64 architecture.
• The Reality: Docker cannot magically make an emulated Intel Linux virtual machine look like physical Mac hardware.

How to actually run it:

1. Natively on Host Machine: Run the code natively on your machine inside the provided Conda environment. It executes on native macOS (using Apple MPS/CPU accelerators) and native Linux clusters (using Nvidia CUDA).
2. Native ARM64 Container: If you absolutely must use Docker on a Mac, build the image specifically targeting your architecture (Not recommended though):

   docker build --platform linux/arm64 -t nspod .
   docker run --rm --platform linux/arm64 -v "$(pwd):/app" nspod --run "Single_wave"