Comparison of different algorithms for classification of transition edge sensor signals for photon number state detection. With the development of a variety of techniques in the field of A.I. the goal is to quantify the advantages of modern classification techniques in the context of photon detection.

The different algorithms are compared in a single notebook available in :

src/Metrics.ipynb

The following methods are evaluated :

• Maximum Value
• Area
• Principal Component Analysis (PCA)
• Kernel Principal Component Analysis (K-PCA)
• t-Distributed Stochastic Neighbor Embedding (t-SNE)
• Non-Negative Matrix Factorization (NMF)
• Isomap
• Autoencoder

The different algorithms are compared in a single notebook available in :

src/AutoencoderAPI.ipynb

An autoencoder is a type of neural network trained to reproduce the signal it receives as input. Half the network (encoder) can be used to asign an arbitrary number of parameters to an input signal. This way, the encoder acts as a dimensionality reduction process.

Neural networks allow for a wide variety of architectures and therefore their evaluation is less straighforward. The AutoencoderAPI was designed to explore a variety of algorithms and create a framework to compare their performance.

• Pytorch dataset classs structure for batching and true random.
• Test GPU functionalities of pytorch
• Modify Log structure to reduce the storing and change it to computing when loading results.

We thank the Ministère de l'Économie et de l’Innovation du Québec and the Natural Sciences and Engineering Research Council of Canada for their financial support.

We acknowledge the help of NIST and Guillaume Thekkadath who provided the datasets used in this work.

[1] T. Gerrits, B. Calkins, N. Tomlin, A. E. Lita, A. Migdall, R. Mirin, and S. W. Nam, “Extending single-photon optimized superconducting transition edge sensors beyond the single-photon counting regime,” Optics Express, vol. 20, no. 21, pp. 23 798–23 810, Oct. 2012.

[2] G. S. Thekkadath, “Preparing and characterizing quantum states of light using photon-number-resolving detectors.”

[3] Y. Ichinohe et al., ‘Application of Deep Learning to the Evaluation of Goodness in the Waveform Processing of Transition-Edge Sensor Calorimeters’, J Low Temp Phys, vol. 209, no. 5, pp. 1008–1016, Dec. 2022, doi: 10.1007/s10909-022-02719-7.