Unbounded Capacity Associative Memory for Real-valued Pattern Storage and Recall

被引：0

作者：

Salem, Fathi M. ^{[1
,2
]}

机构：

[1] Neural AI Syst Inst, Neurai Syst, Okemos, MI 48864 USA

[2] Michigan State Univ, Dept ECE, E Lansing, MI 48824 USA

来源：

2021 IEEE INTERNATIONAL MIDWEST SYMPOSIUM ON CIRCUITS AND SYSTEMS (MWSCAS) | 2021年

关键词：

Real-valued Associate Memory; Deep Learning; Neural Networks; Gradient Systems; (Stochastic) Gradient Descent; Non-convex Optimization;

D O I：

10.1109/MWSCAS47672.2021.9531698

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

We describe an unbounded capacity Associative Memory which effectively stores and retrieves unrestricted real-valued data/patterns with fidelity. This Associative Memory arises from a gradient system of a (differentiable scalar) energy function, and thus can directly be incorporated within existing layers of computational Deep Learning (DL) frameworks. The design effort also describes two options for key pattern retrieval.

引用

页码：1066 / 1069

页数：4

共 50 条

[41] Comments on″Capacity Analysis of the Asymptotically Stable Multi-Valued Exponential Bidirectinal Associative Memory″
CHEN LeiYANG GengXU Bihuan College of ComputerNanjing University of Posts and TelecommunicationsNanjing China Institute of Computer TechnologyNanjing University of Posts and TelecommunicationsNanjing China College of ScienceNanjing University of Posts and TelecommunicationsNanjing China
南京邮电大学学报(自然科学版), 2011, 31 (03) : 90 - 93
[42] Associative memory using nonlinear line attractor network for multi-valued pattern association
Seow, MJ
Asari, VK
ADVANCES IN NEURAL NETWORKS - ISNN 2005, PT 1, PROCEEDINGS, 2005, 3496 : 485 - 490
[43] Storage capacity diverges with synaptic efficiency in an associative memory model with synaptic delay and pruning
Miyoshi, S
Okada, M
IEEE TRANSACTIONS ON NEURAL NETWORKS, 2004, 15 (05): : 1215 - 1227
[44] Optimal matrix compression yields storage capacity 1 for binary Willshaw associative memory
Knoblauch, A
ARTIFICAIL NEURAL NETWORKS AND NEURAL INFORMATION PROCESSING - ICAN/ICONIP 2003, 2003, 2714 : 325 - 332
[45] INFORMATION-STORAGE CAPACITY OF AN OPTICAL HOLOGRAPHIC ASSOCIATIVE MEMORY UTILIZING PHASE CONJUGATION
OWECHKO, Y
SOFFER, BH
MAROM, E
DUNNING, GJ
JOURNAL OF THE OPTICAL SOCIETY OF AMERICA A-OPTICS IMAGE SCIENCE AND VISION, 1986, 3 (13): : P31 - P32
[46] Low-Complexity Continuous-Flow Memory-Based FFT Architectures for Real-Valued Signals
Hazarika, Jinti
Khan, Mohd Tasleem
Ahamed, Shaik Rafi
2019 32ND INTERNATIONAL CONFERENCE ON VLSI DESIGN AND 2019 18TH INTERNATIONAL CONFERENCE ON EMBEDDED SYSTEMS (VLSID), 2019, : 46 - 51
[47] A Novel Address Scheme for Continuous-Flow Parallel Memory-Based Real-Valued FFT Processor
Yuan, Min
Ma, Zhenguo
Yu, Feng
Xing, Qianjian
ELECTRONICS, 2019, 8 (09)
[48] Pattern discovery on character sets and real-valued data: Linear bound on irredundant motifs and an efficient polynomial time algorithm
Parida, L
Rigoutsos, I
Floratos, A
Platt, D
Gao, Y
PROCEEDINGS OF THE ELEVENTH ANNUAL ACM-SIAM SYMPOSIUM ON DISCRETE ALGORITHMS, 2000, : 297 - 308
[49] Implementation of Hopfield Associative Memory with Evolutionary Algorithm and MC-Adaptation Rule for Pattern Storage
Kumar, Somesh
Goel, Rajkumar
Singh, Manu Pratap
PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON SOFT COMPUTING FOR PROBLEM SOLVING (SOCPROS 2011), VOL 2, 2012, 131 : 213 - +
[50] AN AFTER-SHANNON MEASURE OF THE STORAGE CAPACITY OF AN ASSOCIATIVE NOISE-LIKE CODING MEMORY
BOTTINI, S
BIOLOGICAL CYBERNETICS, 1988, 59 (03) : 151 - 159

← 1 2 3 4 5 →