Now you are in the subtree of TECHNOLOGY and MARKETS public knowledge tree. 

Basic literature for BNN-MRAM analog accelerator - april 2021

motivation and demand for AI Hardware

Binary Neural Networks and Deep Learning

2016 Matthieu Courbariaux, Yoshua Bengio Binarized Neural Networks: Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1

2015 Review - Deep learning Yann LeCun, Yoshua Bengio, Geoffrey Hinton

Error Tolerance

2019 T. Hirtzlin Outstanding Bit Error Tolerance of Resistive RAM-Based Binarized Neural Networks

Basics of STT-MRAM and other non-volatile MRAMs

https://www.mram-info.com/stt-mram

latency, endurance, retention
c

2018 JP Wang INVITED: A Pathway to Enable Exponential Scaling for the Beyond-CMOS Era

2019 - MRAM Technology and Business

2018 Beyond CMOS computing with spin and polarization Sasikanth Manipatruni, Dmitri E. Nikonov    and Ian A. Young

Business Opportunity at the EDGE and IoT

GS series - 2018 How edge computing will augment the cloud & unlock real-time, big data applications

2018 McKinsey New demand, new markets: What edge computing means for hardware companies

Mapping the 100bn IT spending opportunity

  • software for the edge devices

tinyML as edge ecosystem as well as the related libraries

In-memory computing architecture

2020 Bavikadi A Review of In-Memory Computing Architectures for Machine Learning Applications

2019 Bing Li An Overview of In-memory Processing with Emerging Non-volatile Memory for Data-intensive Applications

Some, not necessarily analog type, examples of stt-MRAM based accelerators

Ameya D. Patil 2018(?) An MRAM-based Deep In-Memory Architecture for Deep Neural Networks IBM

Angizi 2019 MRIMA: An MRAM-based In-Memory Accelerator

Yu Pan 2018 A Multilevel Cell STT-MRAM-Based Computing In-Memory Accelerator for Binary Convolutional Neural Network

some examples of commercial products

Mythic - the product - M1108 ANALOG MATRIX PROCESSOR

Syntiant - the product - Always-On Sensor & Speech Recognition Processor -NDP100

Collections of further links ... a bit chaotic,it needs re-organising

STT-MRAM - computing
Crossbar circuit for MRAM
Learning with CIM-stt-MRAM