Multiclass classification of dephasing channels

We address the use of neural networks (NNs) in classifying environmental parameters single-qubit dephasing channels. In particular, we investigate performance linear perceptrons and two nonlinear NN architectures. At variance with time-series-based approaches, our goal is to learn a discretized probability distribution over using tomographic data at just random instants time. consider channels originating either from classical $1/{f}^{\ensuremath{\alpha}}$ noise or interaction bath quantum oscillators. The be classified are color $\ensuremath{\alpha}$ Ohmicity parameter $s$ environment. both cases, find that NNs able exactly classify into 16 classes noiseless (a enough for color, whereas single-layer needed Ohmicity). presence noisy (e.g., coming measurements), network about $70%$ accuracy, classification turns out challenging. also more coarse-grained task train discriminate between macroclasses corresponding $\ensuremath{\alpha}\ensuremath{\lessgtr}1$ $s\ensuremath{\lessgtr}1$, obtaining up $96%$ $79%$ accuracy NNs.