Neuromorphic computing systems execute machine learning tasks designed with spiking neural networks. These systems are embracing non-volatile memory to implement high-density and low-energy synaptic storage. Elevated voltages and currents needed to operate non-volatile memories cause aging of CMOS-based transistors in each neuron and synapse circuit in the hardware, drifting the transistor’s parameters from their nominal values. If these circuits are used continuously for too long, the parameter drifts cannot be reversed, resulting in permanent degradation of circuit performance over time, eventually leading to hardware faults. Aggressive device scaling increases power density and temperature, which further accelerates the aging, challenging the reliable operation of neuromorphic systems. Existing reliability-oriented techniques periodically de-stress all neuron and synapse circuits in the hardware at fixed intervals, assuming worst-case operating conditions, without actually tracking their aging at run-time. To de-stress these circuits, normal operation must be interrupted, which introduces latency in spike generation and propagation, impacting the inter-spike interval and hence, performance (e.g., accuracy). We observe that in contrast to long-term aging, which permanently damages the hardware, short-term aging in scaled CMOS transistors is mostly due to bias temperature instability. The latter is heavily workload-dependent and, more importantly, partially reversible. We propose a new architectural techniquemore »
This content will become publicly available on October 22, 2023
Si-based self-programming neuromorphic integrated circuits for intelligent morphing wings
Unlike artificial intelligent systems based on computers, which must be programmed for specific tasks, the human brain can learn in real-time to create new tactics and adapt to complex, unpredictable environments. Computers embedded in artificial intelligent systems can execute arbitrary inference algorithms capable of outperforming humans at specific tasks. However, without real-time self-programming functionality, they must be preprogrammed by humans and will likely to fail in unpredictable environments beyond their preprogrammed domains. In this work, a Si-based synaptic resistor (synstor) was developed by integrating Al2Ox/TaOymaterials to emulate biological synapses. The synstors were characterized, and their operation mechanism based on the charge stored in the oxygen vacancies in the Al2Oxmaterial was simulated and analyzed, to understand the inference, learning, and memory functions of the synstors. A self-programming neuromorphic integrated circuit (SNIC) based on synstors was fabricated to execute inference and learning algorithms concurrently in real-time with an energy efficiency more than six-orders of magnitudes higher than those of standard digital computers. The SNIC dynamically modified its algorithm in a real-time learning process to control a morphing wing, thus successfully improving its lift-to-drag force ratio and recovering the wing from stall in complex aerodynamic environments. The synaptic resistor circuits can potentially circumvent more »
- Authors:
- ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; ; more »
- Publication Date:
- NSF-PAR ID:
- 10376577
- Journal Name:
- Journal of Composite Materials
- Volume:
- 56
- Issue:
- 30
- Page Range or eLocation-ID:
- p. 4561-4575
- ISSN:
- 0021-9983
- Publisher:
- SAGE Publications
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Cyber-Physical-Human Systems (CPHS) interconnect humans, physical plants and cyber infrastructure across space and time. Industrial processes, electromechanical systems operations and medical diagnosis are some examples where one can see the intersection of humans, physical and cyber components. Emergence of Artificial Intelligence (AI) based computational models, controllers and decision support engines have improved the efficiency and cost effectiveness of such systems and processes. These CPHS typically involve a collaborative decision environment, comprising of AI-based models and human experts. Active Learning (AL) is a category of AI algorithms which aims to learn an efficient decision model by combining domain expertise of the human expert and computational capabilities of the AI model. Given the indispensable role of humans and lack of understanding about human behavior in collaborative decision environments, modeling and prediction of behavioral biases is a critical need. This paper, for the first time, introduces different behavioral biases within an AL context and investigates their impacts on the performance of AL strategies. The modelling of behavioral biases is demonstrated using experiments conducted on a real-world pancreatic cancer dataset. It is observed that classification accuracy of the decision model reduces by at least 20% in case of all the behavioral biases.
-
psychology, and cognition has progressed sufficiently that the technology exists to develop a mutually beneficial exchange of information between a human and an AI. Dubbed “AI Symbiosis,” this process enables positive feedback between humans and adaptive computer algorithms in which both human and AI would “learn” how to perform tasks more efficiently than either could alone. Several new technologies and inventions al-low a vast array of augmented input and/or output between humans and AI, in-cluding mental activity wirelessly operating computers, manipulation of targeted neurons with or without implants, non-invasive, surface-level implants the size of a coin transmitting real-time neural activity of senses, real-time video feed of human mental images, and estimation of thoughts and emotions. A research pro-ject is planned to study students’ divided attention when they are learning content in on-line environments. The research will target eye-tracking, click timing, and task performance data to determine the levels of impact divided attention has on student learning. We believe that this line of research will also inform best prac-tices in on-line instructional settings.
-
Abstract Microelectronic computers have encountered challenges in meeting all of today’s demands for information processing. Meeting these demands will require the development of unconventional computers employing alternative processing models and new device physics. Neural network models have come to dominate modern machine learning algorithms, and specialized electronic hardware has been developed to implement them more efficiently. A silicon photonic integration industry promises to bring manufacturing ecosystems normally reserved for microelectronics to photonics. Photonic devices have already found simple analog signal processing niches where electronics cannot provide sufficient bandwidth and reconfigurability. In order to solve more complex information processing problems, they will have to adopt a processing model that generalizes and scales. Neuromorphic photonics aims to map physical models of optoelectronic systems to abstract models of neural networks. It represents a new opportunity for machine information processing on sub-nanosecond timescales, with application to mathematical programming, intelligent radio frequency signal processing, and real-time control. The strategy of neuromorphic engineering is to externalize the risk of developing computational theory alongside hardware. The strategy of remaining compatible with silicon photonics externalizes the risk of platform development. In this perspective article, we provide a rationale for a neuromorphic photonics processor, envisioning its architecture and amore »
-
Recent research suggests that in vitro neural networks created from dissociated neurons may be used for computing and performing machine learning tasks. To develop a better artificial intelligent system, a hybrid bio-silicon computer is worth exploring, but its performance is still inferior to that of a silicon-based computer. One reason may be that a living neural network has many intrinsic properties, such as random network connectivity, high network sparsity, and large neural and synaptic variability. These properties may lead to new design considerations, and existing algorithms need to be adjusted for living neural network implementation. This work investigates the impact of neural variations and random connections on inference with learning algorithms. A two-layer hybrid bio-silicon platform is constructed and a five-step design method is proposed for the fast development of living neural network algorithms. Neural variations and dynamics are verified by fitting model parameters with biological experimental results. Random connections are generated under different connection probabilities to vary network sparsity. A multi-layer perceptron algorithm is tested with biological constraints on the MNIST dataset. The results show that a reasonable inference accuracy can be achieved despite the presence of neural variations and random network connections. A new adaptive pre-processing technique ismore »
