Exponential Error Convergence in Data Classification with Optimized Random Features: Acceleration by Quantum Machine Learning. (arXiv:2106.09028v2 [quant-ph] UPDATED)

Classification is a common task in machine learning. Random features (RFs)
stand as a central technique for scalable learning algorithms based on kernel
methods, and more recently proposed optimized random features, sampled
depending on the model and the data distribution, can significantly reduce and
provably minimize the required number of features. However, existing research
on classification using optimized RFs has suffered from computational hardness
in sampling each optimized RF; moreover, it has failed to achieve the
exponentially fast error-convergence speed …

arxiv classification convergence data data classification error features learning machine machine learning quantum random