**3.3 Cortical columns computing systems**

Our research builds on these two bodies of prior works to explore the potential of creating brain-inspired and brain-like computing fabric, which we call *Cortical Columns Computing Systems (C3S)*. **Figure 4** compares our C3S approach against conventional computers and biological neocortex. Our goal is to leverage the best attributes of the other two paradigms to create a new computing paradigm. **Figure 4** shows the linkage of comparable levels of abstractions across these three distinct computing system paradigms. Our research spans all four levels of abstractions including sensory processing applications, a processor-level microarchitecture model, RTL-level functional units, and gate-level building blocks. In order to target and impact mass market computing, our research must be able to leverage current digital CMOS technology and design tools. One of our goals is to develop an end-to-end framework for designing and implementing Cortical Columns Computing Systems (C3S). This will require the development of novel design exploration and design synthesis tools. Our target applications include diverse sensory signal processing units capable of energy-efficient and edge-native AI inference and online continuous learning.

As illustrated in **Figure 5**, this framework consists of three key components: (1) a microarchitecture model that can facilitate RTL implementation of Cortical Columns (CCs) and Reference Frames (RFs) employed in C3S designs; (2) a suite of highly optimized functional units and building blocks implemented in System Verilog to support efficient application-specific implementations of C3S designs; and (3) a software tool suite consisting of a PyTorch simulator for rapid design space exploration of C3S designs and a design synthesis flow for mapping PyTorch functional models to corresponding C3S hardware. Specific applications of current interest include visual object recognition, anomaly detection on time-series signals (e.g., ECG), edge-native always-on keyword spotting, and multi-modal human activity recognition (HAR) [56].


#### **Figure 4.**

*Comparison of cortical columns computing systems (C3S) against conventional computers and biological computing systems in terms of levels of computing abstraction. Our research strategy is highlighted in the red box and spans the four abstraction layers: Applications and system architecture, processor-level microarchitecture, RTL-level functional units, and gate-level building blocks.*

#### **Figure 5.**

*Proposed framework for implementing C3S designs consists of three key components: Microarchitecture model [11], functional building blocks [12] and design exploration tools. Some applications of interest are visual object recognition, anomaly detection [13], keyword spotting, and human activity recognition [56]. These framework components currently support TNN design and implementation and key relevant publications are cited here. Our ongoing research aims to extend this framework to support more general C3S designs.*
