Fig. 3: Learning curve and sample variance.
From: Machine learning of high dimensional data on a noisy quantum processor
Learning curve for an SVM trained using noiseless circuit encoding on 17 qubits vs. RBF kernel \(k({{{{\bf{x}}}}}_{i},{{{{\bf{x}}}}}_{j})=\exp (-\gamma | | {{{{\bf{x}}}}}_{i}-{{{{\bf{x}}}}}_{j}| {| }^{2})\) with γ = 0.012 optimized via adaptive grid search over [10−5, 10−1]. Points reflect train/test accuracy for a classifier trained on a stratified 10-fold split resulting in a size-x balanced subset of preprocessed supernova data points. Error bars indicate standard deviation over 10 trials of downsampling, and the dashed line indicates the size m = 210 of the training set chosen for this experiment.
