-
Polynomial Approximation of Activation Function
The dataset used in the paper is a polynomial approximation of the activation function ReLU(x) = max(0, x) over the range [-8, +8] by a polynomial of degree 2. -
Low-Latency CryptoNets (LoLa) for Private Inference
The CalTech-101 dataset is used to evaluate the performance of the proposed Low-Latency CryptoNets (LoLa) solution for private inference. -
nGraph-HE2
nGraph-HE2: A High-Throughput Framework for Neural Network Inference on Encrypted Data