1 code implementation • 3 Jul 2021 • Yaniv Shulman
Quantization based model compression serves as high performing and fast approach for inference that yields models which are highly compressed when compared to their full-precision floating point counterparts.
1 code implementation • 7 Dec 2020 • Yaniv Shulman
Modern neural network architectures typically have many millions of parameters and can be pruned significantly without substantial loss in effectiveness which demonstrates they are over-parameterized.
Ranked #46 on Image Classification on MNIST
no code implementations • 3 Jun 2020 • Yaniv Shulman
These structural similarity features may be used with various algorithms however in this paper the focus and the second main contribution is on integrating these features with a revisited pooling layer DiffPool arXiv:1806. 08804 to propose a pooling layer referred to as SimPool.
no code implementations • 5 Nov 2019 • Yaniv Shulman
Where dealing with temporal sequences it is fair to assume that the same kind of deformations that motivated the development of the Dynamic Time Warp algorithm could be relevant also in the calculation of the dot product ("convolution") in a 1-D convolution layer.
no code implementations • 1 Apr 2019 • Yaniv Shulman
A method for unsupervised contextual anomaly detection is proposed using a cross-linked pair of Variational Auto-Encoders for assigning a normality score to an observation.
Contextual Anomaly Detection Unsupervised Contextual Anomaly Detection