Novel Magnetic RAM Architecture Paves Way for AI in IoT Edge Devices

Bridging AI and IoT: The Challenge of Edge Computing

As artificial intelligence (AI) and the Internet of Things (IoT) continue to advance rapidly, researchers face a significant challenge: implementing AI capabilities, particularly artificial neural networks (ANNs), on small IoT edge devices with limited resources 1

1

2

2

. These devices typically have constraints in power, processing speed, and circuit space, making it difficult to run computationally intensive AI algorithms efficiently.

Innovative Solution: Ternarized Gradient Binarized Neural Network (TGBNN)

To address this challenge, Professor Takayuki Kawahara and Yuya Fujiwara from the Tokyo University of Science have developed a novel training algorithm called ternarized gradient binarized neural network (TGBNN) 1

1

2

2

. This algorithm builds upon binarized neural networks (BNNs), which use only -1 and +1 for weights and activation values, reducing the smallest unit of information to one bit.

The TGBNN algorithm introduces three key innovations:

1. Employing ternary gradients during training while maintaining binary weights and activations
2. Enhancing the Straight Through Estimator (STE) to improve gradient backpropagation control
3. Adopting a probabilistic approach for parameter updates based on MRAM cell behavior

Cutting-Edge Computing-in-Memory (CiM) Architecture

The researchers implemented the TGBNN algorithm in a novel computing-in-memory (CiM) architecture, designed specifically for IoT devices 1

1

2

2

. This approach performs calculations directly in memory, saving circuit space and power. The team developed a new XNOR logic gate as the building block for a Magnetic Random Access Memory (MRAM) array, using a magnetic tunnel junction to store information.

Advanced MRAM Cell Manipulation

To change the stored value of individual MRAM cells, the researchers utilized two mechanisms 1

1

2

2

:

1. Spin-orbit torque: The force generated when an electron spin current is injected into a material
2. Voltage-controlled magnetic anisotropy: Manipulation of the energy barrier between different magnetic states in a material

These methods allowed the team to reduce the size of the product-of-sum calculation circuit to half that of conventional units.

Promising Performance Results

The researchers tested their MRAM-based CiM system for BNNs using the MNIST handwriting dataset 1

1

2

2

. The results were impressive:

• Achieved over 88% accuracy using Error-Correcting Output Codes (ECOC)-based learning
• Matched the accuracy of regular BNNs with the same structure
• Demonstrated faster convergence during training

Implications for IoT and AI Integration

This breakthrough could lead to more powerful IoT devices with enhanced AI capabilities 1

1

2

2

. Potential applications include:

• Wearable health monitoring devices: Improved efficiency, smaller size, and reliability without constant cloud connectivity
• Smart homes: More complex tasks and responsive operations
• Energy efficiency: Reduced power consumption across various use cases, contributing to sustainability goals

The innovative MRAM-based architecture and TGBNN algorithm represent a significant step towards implementing efficient neural networks on edge IoT devices, potentially revolutionizing the integration of AI and IoT technologies.