Sparsity-preserving differentially private training

Large embedding models have emerged as a fundamental tool for various applications in recommendation systems [1, 2] and natural language processing [3, 4, 5]. Such models enable the integration of non-numerical data into deep learning models by mapping categorical or string-valued input attributes with large vocabularies to fixed-length representation vectors using embedding layers. These models are widely deployed …

algorithms and natural language processing applications categorical data deep learning differential privacy embedding embedding models google google research huang integration language language processing mapping natural natural language natural language processing numerical processing recommendation recommendation systems research research scientist sparsity string systems tool training