Brainchip Fundamentals

Figured I might share some research, most of you have read this over the years but I figured we are at a good position where it would be useful to have it in one location for new BRN investors or just for those genuinely interested in what has been going on with the company over the past few years.

We are at a very exciting point in the company.

Enjoy~~



BrainChip Research


Peter Van Der Mades book:

Higher Intelligence: How to Create a Functional Artificial Brain

https://www.amazon.com.au/Higher-Intelligence-Create-Functional-Artificial/dp/1922204153/

Higher Intelligence details a decade-long quest to research the human brain with the aim to develop an artificial brain chip. Computers are great tools for number crunching, statistical analysis or surfing the internet but their usefulness is limited when it comes to Artificial Intelligence. For nearly 70 years, computer scientists have tried to develop programs that approach the intellect of a human, but to no avail. Humanoid intelligent robots are still fiction. Even the most sophisticated robots are nothing but programmed dolls. Can machines act with intelligence, think and show initiative? What is the definition of intelligence? How can we build a learning, intelligent machine? Once we have built such an intelligent machine, what will its impact on society be? Why has learning, a prime function of the brain, until now never been part of the Artificial Intelligence process? Higher Intelligence explores all these questions and gives insights in the function of the human brain. It sheds light on a new digital technology, an artificial brain chip that has the capability to learn and evolve.

Patents:

Autonomous learning dynamic artificial neural computing device and brain inspired system

https://patents.google.com/patent/US8250011B2/en

Abstract

A hierarchical information processing system is disclosed having a plurality of artificial neurons, comprised of binary logic gates, and interconnected through a second plurality of dynamic artificial synapses, intended to simulate or extend the function of a biological nervous system. The system is capable of approximation, autonomous learning and strengthening of formerly learned input patterns. The system learns by simulated Synaptic Time Dependent Plasticity, commonly abbreviated to STDP. Each artificial neuron consisting of a soma circuit and a plurality of synapse circuits, whereby the soma membrane potential, the soma threshold value, the synapse strength and the Post Synaptic Potential at each synapse are expressed as values in binary registers, which are dynamically determined from certain aspects of input pulse timing, previous strength value and output pulse feedback.

Method and A System for Creating Dynamic Neural Function Libraries

https://patents.google.com/patent/US20190012597A1/en

Abstract

A method for creating a dynamic neural function library that relates to Artificial Intelligence systems and devices is provided. Within a dynamic neural network (artificial intelligent device), a plurality of control values are autonomously generated during a learning process and thus stored in synaptic registers of the artificial intelligent device that represent a training model of a task or a function learned by the artificial intelligent device. Control Values include, but are not limited to, values that indicate the neurotransmitter level that is present in the synapse, the neurotransmitter type, the connectome, the neuromodulator sensitivity, and other synaptic, dendric delay and axonal delay parameters. These values form collectively a training model. Training models are stored in the dynamic neural function library of the artificial intelligent device. The artificial intelligent device copies the function library to an electronic data processing device memory that is reusable to train another artificial intelligent device.

Spiking neural network

https://patents.google.com/patent/US20200143229A1/en

Abstract

Disclosed herein are system, method, and computer program product embodiments for an improved spiking neural network (SNN) configured to learn and perform unsupervised extraction of features from an input stream. An embodiment operates by receiving a set of spike bits corresponding to a set synapses associated with a spiking neuron circuit. The embodiment applies a first logical AND function to a first spike bit in the set of spike bits and a first synaptic weight of a first synapse in the set of synapses. The embodiment increments a membrane potential value associated with the spiking neuron circuit based on the applying. The embodiment determines that the membrane potential value associated with the spiking neuron circuit reached a learning threshold value. The embodiment then performs a Spike Time Dependent Plasticity (STDP) learning function based on the determination that the membrane potential value of the spiking neuron circuit reached the learning threshold value.

Low power neuromorphic voice activation system and method

https://patents.google.com/patent/US10157629B2/en

Abstract

The present invention provides a system and method for controlling a device by recognizing voice commands through a spiking neural network. The system comprises a spiking neural adaptive processor receiving an input stream that is being forwarded from a microphone, a decimation filter and then an artificial cochlea. The spiking neural adaptive processor further comprises a first spiking neural network and a second spiking neural network. The first spiking neural network checks for voice activities in output spikes received from artificial cochlea. If any voice activity is detected, it activates the second spiking neural network and passes the output spike of the artificial cochlea to the second spiking neural network that is further configured to recognize spike patterns indicative of specific voice commands. If the first spiking neural network does not detect any voice activity, it halts the second spiking neural network.

Neural processor based accelerator system and method

https://patents.google.com/patent/US20170024644A1/en

Abstract

A configurable spiking neural network based accelerator system is provided. The accelerator system may be executed on an expansion card which may be a printed circuit board. The system includes one or more application specific integrated circuits comprising at least one spiking neural processing unit and a programmable logic device mounted on the printed circuit board. The spiking neural processing unit includes digital neuron circuits and digital, dynamic synaptic circuits. The programmable logic device is compatible with a local system bus. The spiking neural processing units contain digital circuits comprises a Spiking Neural Network that handles all of the neural processing. The Spiking Neural Network requires no software programming, but can be configured to perform a specific task via the Signal Coupling device and software executing on the host computer. Configuration parameters include the connections between synapses and neurons, neuron types, neurotransmitter types, and neuromodulation sensitivities of specific neurons.

Intelligent biomorphic system for pattern recognition with autonomous visual feature extraction

https://patents.google.com/patent/US20170236027A1/en

Abstract

Embodiments of the present invention provide a hierarchical arrangement of one or more artificial neural networks for recognizing visual feature pattern extraction and output labeling. The system comprises a first spiking neural network and a second spiking neural network. The first spiking neural network is configured to autonomously learn complex, temporally overlapping visual features arising in an input pattern stream. Competitive learning is implemented as spike time dependent plasticity with lateral inhibition in the first spiking neural network. The second spiking neural network is connected by means of dynamic synapses with the first spiking neural network, and is trained for interpreting and labeling output data of the first spiking neural network. Additionally, the output of the second spiking neural network is transmitted to a computing device, such as a CPU for post processing.

Secure Voice Communications System

https://patents.google.com/patent/US20190188600A1/en

Abstract

Disclosed herein are system and method embodiments for establishing secure communication with a remote artificial intelligent device. An embodiment operates by capturing an auditory signal from an auditory source. The embodiment coverts the auditory signal into a plurality of pulses having a spatio-temporal distribution. The embodiment identifies an acoustic signature in the auditory signal based on the plurality of pulses using a spatio-temporal neural network. The embodiment modifies synaptic strengths in the spatio-temporal neural network in response to the identifying thereby causing the spatio-temporal neural network to learn to respond to the acoustic signature in the acoustic signal. The embodiment transmits the plurality of pulses to the remote artificial intelligent device over a communications channel thereby causing the remote artificial intelligent device to learn to respond to the acoustic signature, and thereby allowing secure communication to be established with the remote artificial intelligent device based on the auditory signature.

Intelligent Autonomous Feature Extraction System Using Two Hardware Spiking Neutral Networks with Spike Timing Dependent Plasticity

https://patents.google.com/patent/US20170236051A1/en

Abstract

Embodiments of the present invention provide an artificial neural network system for feature pattern extraction and output labeling. The system comprises a first spiking neural network and a second spiking neural network. The first spiking neural network is configured to autonomously learn complex, temporally overlapping features arising in an input pattern stream. Competitive learning is implemented as spike timing dependent plasticity with lateral inhibition in the first spiking neural network. The second spiking neural network is connected with the first spiking neural network through dynamic synapses, and is trained to interpret and label the output data of the first spiking neural network. Additionally, the labeled output of the second spiking neural network is transmitted to a computing device, such as a central processing unit for post processing.

System and Method for Spontaneous Machine Learning and Feature Extraction

https://patents.google.com/patent/US20180225562A1/en

Abstract

Embodiments of the present invention provide an artificial neural network system for improved machine learning, feature pattern extraction and output labeling. The system comprises a first spiking neural network and a second spiking neural network. The first spiking neural network is configured to spontaneously learn complex, temporally overlapping features arising in an input pattern stream. Competitive learning is implemented as Spike Timing Dependent Plasticity with lateral inhibition in the first spiking neural network. The second spiking neural network is connected with the first spiking neural network through dynamic synapses, and is trained to interpret and label the output data of the first spiking neural network. Additionally, the output of the second spiking neural network is transmitted to a computing device, such as a CPU for post processing.

Exclusively Licensed Patents:

Method, digital electronic circuit and system for unsupervised detection of repeating patterns in a series of events

https://patents.google.com/patent/US20190286944A1/en

Abstract

A method of performing unsupervised detection of repeating patterns in a series of events, includes a) Providing a plurality of neurons, each neuron being representative of W event types; b) Acquiring an input packet comprising N successive events of the series; c) Attributing to at least some neurons a potential value, representative of the number of common events between the input packet and the neuron; d) Modify the event types of neurons having a potential value exceeding a first threshold TL; and e) generating a first output signal for all neurons having a potential value exceeding a second threshold TF, and a second output signal, different from the first one, for all other neurons. A digital electronic circuit and system configured for carrying out such a method is also provided.

Published Research:

Unsupervised learning of repeating patterns using a novel STDP based algorithm

https://jov.arvojournals.org/article.aspx?articleid=2651951

Computational vision systems that are trained with deep learning have recently matched human performance (Hinton et al). However, while deep learning typically requires tens or hundreds of thousands of labelled examples, humans can learn a task or stimulus with only a few repetitions. For example, a 2015 study by Andrillon et al. showed that human listeners can learn complicated random auditory noises after only a few repetitions, with each repetition invoking a larger and larger EEG activity than the previous. In addition, a 2015 study by Martin et al. showed that only 10 minutes of visual experience of a novel object class was required to change early EEG potentials, improve saccadic reaction times, and increase saccade accuracies for the particular object trained. How might such ultra-rapid learning actually be accomplished by the cortex? Here, we propose a simple unsupervised neural model based on spike timing dependent plasticity, which learns spatiotemporal patterns in visual or auditory stimuli with only a few repetitions. The model is attractive for applications because it is simple enough to allow the simulation of very large numbers of cortical neurons in real time. Theoretically, the model provides a plausible example of how the brain may accomplish rapid learning of repeating visual or auditory patterns using only a few examples.

Unsupervised Feature Learning With Winner-Takes-All Based STDP

https://www.frontiersin.org/articles/10.3389/fncom.2018.00024/full

Abstract

We present a novel strategy for unsupervised feature learning in image applications inspired by the Spike-Timing-Dependent-Plasticity (STDP) biological learning rule. We show equivalence between rank order coding Leaky-Integrate-and-Fire neurons and ReLU artificial neurons when applied to non-temporal data. We apply this to images using rank-order coding, which allows us to perform a full network simulation with a single feed-forward pass using GPU hardware. Next we introduce a binary STDP learning rule compatible with training on batches of images. Two mechanisms to stabilize the training are also presented : a Winner-Takes-All (WTA) framework which selects the most relevant patches to learn from along the spatial dimensions, and a simple feature-wise normalization as homeostatic process. This learning process allows us to train multi-layer architectures of convolutional sparse features. We apply our method to extract features from the MNIST, ETH80, CIFAR-10, and STL-10 datasets and show that these features are relevant for classification. We finally compare these results with several other state of the art unsupervised learning methods.

A Hardware-Deployable Neuromorphic Solution for Encoding and Classification of Electronic Nose Data

https://www.mdpi.com/1424-8220/19/22/4831

Abstract

In several application domains, electronic nose systems employing conventional data processing approaches incur substantial power and computational costs and limitations, such as significant latency and poor accuracy for classification. Recent developments in spike-based bio-inspired approaches have delivered solutions for the highly accurate classification of multivariate sensor data with minimized computational and power requirements. Although these methods have addressed issues related to efficient data processing and classification accuracy, other areas, such as reducing the processing latency to support real-time application and deploying spike-based solutions on supported hardware, have yet to be studied in detail. Through this investigation, we proposed a spiking neural network (SNN)-based classifier, implemented in a chip-emulation-based development environment, that can be seamlessly deployed on a neuromorphic system-on-a-chip (NSoC). Under three different scenarios of increasing complexity, the SNN was determined to be able to classify real-valued sensor data with greater than 90% accuracy and with a maximum latency of 3 s on the software-based platform. Highlights of this work included the design and implementation of a novel encoder for artificial olfactory systems, implementation of unsupervised spike-timing-dependent plasticity (STDP) for learning, and a foundational study on early classification capability using the SNN-based classifier.

Real-Time Classification of Multivariate Olfaction Data Using Spiking Neural Networks

https://www.mdpi.com/1424-8220/19/8/1841

Abstract

Recent studies in bioinspired artificial olfaction, especially those detailing the application of spike-based neuromorphic methods, have led to promising developments towards overcoming the limitations of traditional approaches, such as complexity in handling multivariate data, computational and power requirements, poor accuracy, and substantial delay for processing and classification of odors. Rank-order-based olfactory systems provide an interesting approach for detection of target gases by encoding multi-variate data generated by artificial olfactory systems into temporal signatures. However, the utilization of traditional pattern-matching methods and unpredictable shuffling of spikes in the rank-order impedes the performance of the system. In this paper, we present an SNN-based solution for the classification of rank-order spiking patterns to provide continuous recognition results in real-time. The SNN classifier is deployed on a neuromorphic hardware system that enables massively parallel and low-power processing on incoming rank-order patterns. Offline learning is used to store the reference rank-order patterns, and an inbuilt nearest neighbor classification logic is applied by the neurons to provide recognition results. The proposed system was evaluated using two different datasets including rank-order spiking data from previously established olfactory systems. The continuous classification that was achieved required a maximum of 12.82% of the total pattern frame to provide 96.5% accuracy in identifying corresponding target gases. Recognition results were obtained at a nominal processing latency of 16ms for each incoming spike. In addition to the clear advantages in terms of real-time operation and robustness to inconsistent rank-orders, the SNN classifier can also detect anomalies in rank-order patterns arising due to drift in sensing arrays.

A Review of Current Neuromorphic Approaches for Vision, Auditory, and Olfactory Sensors

https://www.frontiersin.org/articles/10.3389/fnins.2016.00115/full

Abstract

Conventional vision, auditory, and olfactory sensors generate large volumes of redundant data and as a result tend to consume excessive power. To address these shortcomings, neuromorphic sensors have been developed. These sensors mimic the neuro-biological architecture of sensory organs using aVLSI (analog Very Large Scale Integration) and generate asynchronous spiking output that represents sensing information in ways that are similar to neural signals. This allows for much lower power consumption due to an ability to extract useful sensory information from sparse captured data. The foundation for research in neuromorphic sensors was laid more than two decades ago, but recent developments in understanding of biological sensing and advanced electronics, have stimulated research on sophisticated neuromorphic sensors that provide numerous advantages over conventional sensors. In this paper, we review the current state-of-the-art in neuromorphic implementation of vision, auditory, and olfactory sensors and identify key contributions across these fields. Bringing together these key contributions we suggest a future research direction for further development of the neuromorphic sensing field.


Application of a Brain-Inspired Spiking Neural Network Architecture to Odor Data Classification

https://www.mdpi.com/1424-8220/20/10/2756

Abstract
Existing methods in neuromorphic olfaction mainly focus on implementing the data transformation based on the neurobiological architecture of the olfactory pathway. While the transformation is pivotal for the sparse spike-based representation of odor data, classification techniques based on the bio-computations of the higher brain areas, which process the spiking data for identification of odor, remain largely unexplored. This paper argues that brain-inspired spiking neural networks constitute a promising approach for the next generation of machine intelligence for odor data processing. Inspired by principles of brain information processing, here we propose the first spiking neural network method and associated deep machine learning system for classification of odor data. The paper demonstrates that the proposed approach has several advantages when compared to the current state-of-the-art methods. Based on results obtained using a benchmark dataset, the model achieved a high classification accuracy for a large number of odors and has the capacity for incremental learning on new data. The paper explores different spike encoding algorithms and finds that the most suitable for the task is the step-wise encoding function. Further directions in the brain-inspired study of odor machine classification include investigation of more biologically plausible algorithms for mapping, learning, and interpretation of odor data along with the realization of these algorithms on some highly parallel and low power consuming neuromorphic hardware devices for real-world applications.

Neuromorphic engineering — A paradigm shift for future IM technologies

https://ieeexplore.ieee.org/abstract/document/8674627

Abstract

Recent developments in measurement science have mainly focused on enhancing the quality of measurement procedures, improving their efficiency, and introducing processes that can increase accuracy [1]. Simultaneously, the implementation of intelligent and adaptive distributed systems using the Internet of Things (IoT) has recently seen a rapid increase. The synergy of these developments requires an extensive use of sensing systems that can measure, analyze and provide accurate results in real-time [2]. However, an increase in the number of sensors leads to challenges such as a necessity for complex processing strategies to handle multivariate data that further leads to overall increases in power consumption and output latency.