Numerical Barren Plateau Detection

Code to reproduce the numerical results and graphs from my MSc thesis.

Building

• Install Rust from https://www.rust-lang.org/tools/install
• Build the tool using cargo build --release

Usage

The binary will be availabe at target/release/bpdetect and accepts the following arguments: bpdetect circuitName numQubits numLayers pauliString parameterIdx:

• circuitName must be one of: introExample, iqpExample, sim1, sim2, sim9, sim10, sim11, sim12, sim15, iqp1, iqp2, iqp3. Here, introExample and iqpExample are the example circuits we discuss in Sections 5.4.1 and 5.5.4 respectively.

• pauliString represents the measurement Hamiltonian, for example ZXIIYX. Should have length numQubits.

• parameterIdx is the parameter with regards to which the derivative is analysed. Counting starts at 0.

Reproducing Results

To reproduce the numerical results from the thesis, first run python3 experiments.py. This will perform the variance computation using a single CPU core and take about 2 hours to run depending on hardware. The execution time can be greatly sped up by utilising multiples CPU cores. For example, python3 experiments.py 10 will use 10 cores and takes about 12 minutes to run on my machine.

To produce the graphs, run python3 plot.py afterwards. Note that this requires pdflatex to be in the system path.