18085011. ASSIGNING DNN WEIGHTS TO A 3D CROSSBAR ARRAY simplified abstract (International Business Machines Corporation)

ASSIGNING DNN WEIGHTS TO A 3D CROSSBAR ARRAY

Organization Name

International Business Machines Corporation

Inventor(s)

Geoffrey Burr of Cupertino CA (US)

HsinYu Tsai of Cupertino CA (US)

Irem Boybat Kara of Adliswil (CH)

Martin Michael Frank of Dobbs Ferry NY (US)

ASSIGNING DNN WEIGHTS TO A 3D CROSSBAR ARRAY - A simplified explanation of the abstract

This abstract first appeared for US patent application 18085011 titled 'ASSIGNING DNN WEIGHTS TO A 3D CROSSBAR ARRAY

The patent application discusses a system, method, and computer program product for efficiently assigning deep neural network (DNN) weight matrices to a Compute-in-Memory (CiM) accelerator system, specifically focusing on allocation strategies for assigning DNN model weight-layers to two-dimensional (2D) tiers of three-dimensional (3D) crossbar array tiles.

• Efficient allocation strategies are optimized to minimize contention, latency, and dead-time, while maximizing accelerator throughput.
• Strategies include assigning DNN weight matrices to 2D tiers of a 3D crossbar array tile to maximize throughput and minimize completion latency for finite-batch-size workflows.
• Allocation strategies also aim to minimize dead-time latency before the next batch member can be input in infinite-batch-size or continuous workflow scenarios.

Potential Applications: - Artificial intelligence - Machine learning - Neuromorphic computing

Problems Solved: - Contention, latency, and dead-time in assigning DNN weight matrices to CiM accelerators - Maximizing accelerator throughput while minimizing completion latency

Benefits: - Improved efficiency in deep neural network processing - Enhanced performance of CiM accelerator systems

Commercial Applications: Title: "Efficient Allocation Strategies for Deep Neural Network Weight Matrices in CiM Accelerators" This technology could be used in industries such as: - Healthcare for medical image analysis - Finance for fraud detection - Autonomous vehicles for real-time decision making

Questions about the technology: 1. How does this technology improve the efficiency of deep neural network processing? 2. What are the potential implications of using CiM accelerators in various industries?

Original Abstract Submitted

A system, method and computer program product for assigning deep neural network (DNN) weight matrices to a Compute-in-Memory (CiM) accelerator system, and particularly, efficient allocation strategies for assigning DNN model weight-layers to two-dimensional (2D) tiers of three-dimensional (3D) crossbar array tiles. Such efficient allocation strategies for assigning DNN model weight-layers to tiers and tiles of a CiM accelerator are optimized to minimize contention, latency and dead-time, and to maximize accelerator throughput. In one scenario, efficient allocation strategies include assigning DNN weight matrices to the 2D tiers of a 3D crossbar array tile to maximize throughput and minimize completion latency for a finite-batch-size example of an incoming workflow. In a further scenario, efficient allocation strategies assign DNN weight matrices to the 2D tiers of a 3D crossbar array tile to minimize dead-time-latency-before-next-batch-member-can-be-input in an infinite-batch-size or a continuous workflow scenario.