2018/06/04 by Tariq Daouda, Jeremie Zumer, Jérémie Zumer +6 · 1 voice · 1 citation
Biochemistry, Genetics and Molecular Biology · Computer Science · Engineering · Mathematics · #Advanced Memory and Neural Computing #Advanced Optical Imaging Technologies #Artificial intelligence #Computer architecture #Computer science #Holography #Neural Networks and Reservoir Computing #Optics #Physics #acm:62-07l #acm:68T05 #acm:68T30 #cs.AI #cs.LG #msc:62-07l #msc:68T05 #msc:68T30 #q-bio.GN #q-bio.TO #stat.ML
paper · pdf · doi:10.48550/arxiv.1806.00931
published in arXiv (Cornell University) (Cornell University) · 10 pages, 7 figures, 1 table
arxiv created 2018/06/04 · openalex publication_date 2018/06/04 · arxiv updated 2018/06/05 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Representation learning is at the heart of what makes deep learning effective. In this work, we introduce a new framework for representation learning that we call "Holographic Neural Architectures" (HNAs). In the same way that an observer can experience the 3D structure of a holographed object by looking at its hologram from several angles, HNAs derive Holographic Representations from the training set. These representations can then be explored by moving along a continuous bounded single dimension. We show that HNAs can be used to make generative networks, state-of-the-art regression models and that they are inherently highly resistant to noise. Finally, we argue that because of their denoising abilities and their capacity to generalize well from very few examples, models based upon HNAs are particularly well suited for biological applications where training examples are rare or noisy.