United States

This presentation is an overview of the recently developed SpikePropamine meta-learning framework submitted for publication in the Frontiers In Neurorobotics journal. Material from the original paper as well as more recent developments will be presented. 

Our goal in developing this framework is to leverage synaptic plasticity in Spiking Neural Networks (SNNs), the learning mechanism found in biological neural networks. Today, most engineered neural network only use fixed-weights despite the superior learning capabilities found in biological plastic neural networks. Therefore, SpikePropamine is a framework that harnesses the power of offline learning of fixed weights along with the online continual learning through synaptic plasticity. Specifically, along with the fixed weights, SpikePropamine additionally learns the rules that govern the online plastic learning and neuromodulation in SNNs through offline gradient descent. The capabilities of the framework are demonstrated through a series of challenging benchmarks. The parameters for variants of BCM, Oja’s and other plasticity and neuromodulated rules are learned. Results show that the SpikePropamine based networks are able to solve temporal learning tasks that fixed-weight SNNs fail to solve. Additional results highlight the robustness of SpikePropamine based SNNs to noise and novel conditions not seen during training. This is seen through minimal degradation of performance in a high-dimensional legged locomotion task governed by a SpikePropamine based SNN.


NAML 2021

SpikePropamine: Meta-Learning Synaptic Plasticity for Spiking Neural Networks

spiking neural network

neuromorphic

meta learning

reinforcement learning

ICPR 2020 is the premier international conference on Pattern Recognition, gathering leading researchers from around the world to present their latest achievements in Pattern Analysis and Recognition, applied Artificial Intelligence and Machine Learning. The topic of the year is " Putting Artificial intelligence to work on Patterns". 
For the first time, ICPR has employed a two-round review process similar to journal submissions, with new papers allowed to be submitted in either the first or second round and papers submitted in the first round and not accepted allowed to be revised and re-submitted for second round review. 

In this edition of the **ICPR 2020** Conference, there are 1407 Oral/Poster presentations organized in five Tracks: 
- Artificial Intelligence, Machine Learning for Pattern Analysis; 
- Biometrics, Human Analysis and Behavior Understanding; Computer Vision, Robotics and Intelligent Systems; 
- Document and Media Analysis; and 
- Image and Signal Processing. 

The **ICPR 2020** Main Conference program also includes 7 Keynotes from prominent world-renowned scientists, 8 Tutorials, 28 Demos, 8 Challenges, 2 Panels, as well as an Industry/Sponsor Exhibition. In addition, 41 Workshops on a range of important current and emerging research topics complete the ICPR 2020 program. 


ICPR 2020

technical paper

Naval Information Warfare Center Pacific is partnering with AFCEA San Diego chapter to host the workshop.

The NAML workshop is an opportunity for researchers from Navy research and development centers and our partner organizations to exchange technical ideas and to come together as a team.

The goals of NAML are:

* To spread awareness of current machine learning research relevant to Naval applications

* To connect machine learning researchers with experts in military requirements

* To build and strengthen collaborations within the defense research community

**Watch this video to hear about the AFCEA purpose.**

<iframe width=1280 height=706 frameborder="0" scrolling="no" src="https://screencast-o-matic.com/player/creleVVcvRh?v=6&ff=1&title=0&controls=1" allowfullscreen="true"></iframe>

It is well known that Deep Neural Network (DNN) implementations take massive volumes of data to sufficiently train their complex network architectures to perform high accuracy inferencing. The Machine Learning (ML) communities, academic and industry, have created sophisticated models of DNNs trained on millions of data samples to achieve robust classification performance. In the area of image analysis, to address the need for data, numerous efforts have led to standard curated datasets, such as MINST, CIFAR-10, Open Images, and ImageNet. In fact, the ImageNet dataset has become a de-facto image dataset for new algorithm evaluation. In a yearly competition, participants are invited to submit new and novel DNN architectures for assessment. The architectures are trained using the ImageNet database and then evaluated for classification performance. In order of development, some of the well-known winning networks originating from this competition are AlexNet, ZFNet, GoogleNet, VGG Net, ResNet, and Inception V4. As a benefit to the deep learning community, many of the competition submissions are released as open-source for others to use and refine. It is important to recognize that the key to the performance of these networks is that the data input into the network is similar in nature to the data on which the network was trained. Herein is the problem facing many Department of Defense (DoD) ML implementations; the lack of sufficient, operationally relevant data to train and innovate DNN architectures for the DoD community. 
As part of a 2020 Navy Small Business Technology Transfer (STTR) program, Areté and its academic partner, the University of Arizona, are developing an interactive scenario building tool capable of generating realistic synthetic 360° videos in real-time for use in DoD training simulators. This novel capability is accomplished by combining the latest advances in generative adversarial networks (GANs), Areté's expertise in physics-based modeling, and the advanced simulation capabilities of the Unreal Engine. In this talk, Areté will present examples of maritime relevant synthetic data compared to real-world data. Areté will then demonstrate how the data can be used to train DNN architectures directly or through transfer learning and how key technologies are marshaled together.


Using the Unreal Engine as a High Fidelity Simulation for Data Creation

This presentation places AI in a global context and provides perspectives that support the imperative for aggressively leading AI transformation at scale. It provides a strategic overview of AI as it relates to other exponential technologies, the expanding growth of data, bandwidth and computing power that is accelerating knowledge doubling around the globe. It discusses the significant economic implications of AI over the next 10 years, the state of global research and development investment and productivity as well as the state of national AI strategies and policies, research ecosystems and the AI-enabled future of the battlespace. A review of the National Security Commission Report on AI is provided to underscore the breadth and depth of impact and the requisite sense of urgency associated with AI and its applications.

AI in Global Context: The Imperative for Leding AI Transformation

Design, Demonstration, and Proof of Concept Using the Explainable Reinforcement Learning (XAI) of Soar for Combat Identification (CID) 


The performance of CPU-based and GPU- based systems is often low for PDE codes, where large, sparse, and often structured systems of linear equations must be solved. Iterative solvers are limited by data movement, both between caches and memory and between nodes. Here we describe the solution of such systems of equations on the Cerebras Systems CS-1, a wafer-scale processor that has the memory bandwidth and communication latency to perform well. We achieve 0.86 PFLOPS on a single wafer-scale system for the solution by BiCGStab of a linear system arising from a 7-point finite difference stencil on a 600 × 595 × 1536 mesh, achieving about one third of the machine’s peak performance. We explain the system, its architecture and programming, and its performance on this problem and related problems. We discuss issues of memory capacity and floating point precision. We outline plans to extend this work towards full applications. 


Fast Stencil-Code Computation on a Wafer-Scale Processor

Radio Modulation classification (RMC) is an important front-end step in a number of applications such as Signals Intelligence, spectrum monitoring, link adaptation, dynamic spectrum access, cognitive radio and 4G/5G cellular standards. We report on an RMC method which uses a novel multi-channel Time-Series Imaging input to a hybrid Residual CNN-BiLSTM Deep Neural Network. Training and testing on the RadioML 2018.01A 24-class dataset shows a peak accuracy of 96% at the highest SNRs and improvement at lower SNRs using the hybrid architecture. A new amateur radio satellite downlink collection effort for a modulation classification dataset with naturally varying SNR, Doppler shift and various bit rates will also be discussed.

Deep Neural Network Modulation Classification

As tech giants are advancing for gaining quantum supremacy, quantum technologies are becoming more and more prominent in different parts of the world, with applications that include the military sector. Promising applications, such as quantum machine learning, are starting to establish as fields of their own. Much like classical machine learning, the goal is to be able to generalize a process into multiple applications, but a deeper question remains as to whether there is computational advantage against classical processes via quantum-inspired algorithms; common ways to find such advantages involves proving that there is no best known classical algorithm that performs more efficiently than a quantum algorithm, and processing wise that there exists a runtime exponential advantage over classical computing [1]. While proving quantum advantage and supremacy is controversial and non-trivial, an additional challenge lies in the processing limitation imposed by currently available quantum computers, often referred to as Noisy Intermediate Scale Quantum (NISQ) processors [2]. As a result, it is widely accepted in academia and industry that there is not much advantage over classical processors given the current technological readiness of available quantum processors due to noise and quantum bits qubit de-coherence limiting the number of qubits and gate operations. These are some avenues to continue advancing the paradigm of quantum machine learning applications acknowledging hardware limitations: 1) theoretical explorations are done to demonstrate, via computational complexity theory, implementations that could get attained in the future once a fault-tolerant quantum computer becomes available, 2) via hybrid instances where some machine learning tasks are managed by a quantum computer while a classical processor takes the rest, such that, quantum machine learning algorithms are used to classify quantum states that are produced by a quantum simulator, and 3) programming a NISQ computer and assessing its quantum algorithm performance against a classical processor counterpart [3]. 
In this talk, we will discuss the challenges for implementing machine learning on NISQ computers and provide insight into what can be achieved accounting for present day hardware limitations. 
REERENCES
[1] Nielsen, Micheal A., and Issac L Chuang. Quantum Computation and Quantum Information. Cambridge: Cambridge University Press, 2000.
[2] Preskill, J. Quantum computing in the NISQ era and beyond. Quantum 2, 79, 2018.
[3] Schuld, Maria, and F. Petruccione. Supervised Learning with Quantum Computers. 1st ed. 2018. Cham: Springer International Publishing, 2018.


Promises And Challenges For Implementing Machine Learning On Noisy Intermediate Scale (NISQ) Computers

Knowledge graphs, or node-edge graphs representing entities and relationships are fundamental to a wide variety of cutting-edge Natural Language Processing (NLP) tasks including search, question-answering, salience determination, slot-filling, and event schema induction. Such tasks enable advanced analytics that allow users to derive new, meaningful insights about their data and are particularly useful in the realm of Big Data. Google’s Knowledge Graph, for example, forms the backbone of its search engine, enabling Google to make more intelligent inferences about results relevant to user queries. Knowledge graphs and their associated NLP tasks are widely applicable to large Navy text corpora including Navy missile standards documents, instrument logs, and cybersecurity and safety policy documents. 

The goal of this project is to build a Natural Language Processing (NLP) pipeline capable of ingesting a corpus of documents and automatically creating a Knowledge Graph containing the entities and relationships in the corpus. We are constructing an end-to-end pipeline composed of six modules: Coreference Resolution, Open Information Extraction, Canonicalization, Entity Linking, a Neo4j Graph Database, and a Graphical User Interface (GUI). 

We first perform Coreference Resolution (AllenNLP) on the corpus to create a single term for all expressions that refer to the same entity. We then extract all subject-predicate-object triplets using Open Information Extraction models by Stanford, AllenNLP, and the University of Washington and IIT Delhi; currently the user can choose a single model or a combination of models. Next we map each name to a canonical form, by clustering entities via cosine similarity, Jaccard similarity, or inverse document frequency. We then link our entities to existing entities with the same names in a pre-constructed knowledge base (KB), in effect ingesting all knowledge contained in the KB. Finally, we load the processed subject-predicate-object triplets into a Neo4j graph database, accessible to users via our GUI. 
The GUI provides an interface through which users can input text data, either by uploading files or typing raw text, choose pipeline modules to include and models and thresholds to use, query the database, run advanced analytics, produce interactive graph visualizations, and save data and visualizations. 

Through the aforementioned six modules, our project is an end-to-end pipeline of state-of-the-art methods for converting unstructured text data into a structured Knowledge Graph. 


Automated Knowledge Graphs

There is a critical need within the military to be able to identify outbreaks of communicable diseases early in order to be able to contain, mitigate, or track their spread. This is especially true when soldiers and personnel have to be kept in close quarters on a ship or in barracks. Wearable health monitoring devices provide the capability to record a variety of real time physiological time series data including, but not limited to: respiratory rate, respiratory rate variability, heart rate, steps, calorie intake and state of mind (or mental activity level). We hypothesize that these physiological measures may be used to predict the health status of individuals over the course of an acute respiratory infection (ARI) from healthy, infected, pre-symptomatic, symptomatic, and recovered. To test this hypothesis, we are using data collected using Spire HealthTag devices that were deployed during a community cohort contact tracing study conducted by the University of Maryland during the 2018-2019 influenza season. The objective of this study is to use the wearable data to predict the onset of acute respiratory infections due to viruses such as parainfluenza, coronavirus, adenovirus, influenza, and rhinovirus. We are using methods in time series analysis, topological data analysis, machine learning, and anomaly detection to develop predictors that link physiological variables collected from the Spire HealthTag device to disease onset based on perceived symptom scores and confirmed infections. A model being developed and its performance in predicting the onset of ARIs will be presented.

Learning Objectives

1. Understand the potential of wearable health data to predict the onset of illness.
2. Describe the model and our analytic approach.
3. Discuss this potential health monitoring system’s use in the field.

Predicting of Health Status from Wearable Physiological Monitoring Data (PreMonData)

PMW/ISEA teams are dealing with thousands of trouble tickets on a daily basis – currently the process is reliant on sailors or help desk self-assessing issue severity​. CASREP Prediction Model was develop to predict which Trouble Tickets are most likely to escalate into a CASREP or will take longer to resolve. This allows immediate identification of high-risk issues as soon as the trouble ticket is submitted. Enables team to prioritize response actions and resources, and potentially reduce system downtime. CASREP Prediction Model uses NAVWAR Help Desk ITSM data and five different AI/ML algorithms to work together to boost accuracy. 

The Trouble Ticket Duration model builds upon the framework of the CASREP Prediction Model but with the focus on predicting how long until a trouble ticket will be resolved. The Fleet will be potentially able to forecast downtime and adjust operational plans to compensate effectively. The PMW/ISEA teams are potentially able to use the forecast to prioritize tasking in resolving these trouble tickets.



Using Machine Learning to Predict CASREP Occurrence and Trouble Ticket Duration

Create synchronized data to train a machine learning algorithm to esimate turbulence in image data.

Premium content

SpikePropamine: Meta-Learning Synaptic Plasticity for Spiking Neural Networks

Downloads

Next from NAML 2021

Using the Unreal Engine as a High Fidelity Simulation for Data Creation

Similar lecture

Meta-Learning via Learned Loss