缩略语（Acronyms）¶
| 缩写 | 含义 |
|---|---|
| ACK | 确认（Acknowledge） |
| ACO | 蚁群优化（Ant Colony Optimization） |
| AI | 人工智能（Artificial Intelligence） |
| ANN | 人工神经网络（Artificial Neural Network） |
| API | 应用程序编程接口（Application Programming Interface） |
| ASIC | 专用集成电路（Application-specific Integrated Circuit） |
| BDA | 战损评估（Battle Damage Assessment） |
| CNN | 卷积神经网络（Convolutional Neural Network） |
| CPU | 中央处理器（Central Processing Unit） |
| CR | 认知无线电（Cognitive Radio） |
| CRN | 认知无线电网络（Cognitive Radio Network） |
| CTS | 发送清除（Clear-to-Send） |
| DM | 决策制定（Decision Making） |
| DoD | 美国国防部（Department of Defense, USA） |
| DSA | 动态频谱访问（Dynamic Spectrum Access） |
| EA | 电子攻击（Electronic Attack） |
| EBM | 电子战场管理（Electronic Battle Management） |
| ECM | 电子对抗措施（Electronic Countermeasures） |
| EMS | 电磁频谱（Electromagnetic Spectrum） |
| EMSO | 电磁频谱作战（Electromagnetic Spectrum Operations） |
| EP | 电子防护（Electronic Protect） |
| ES | 电子支援（Electronic Support） |
| EW | 电子战（Electronic Warfare） |
| EW BDA | 电子战战损评估（Electronic Warfare Battle Damage Assessment） |
| EWO | 电子战军官（Electronic Warfare Officer） |
| FFT | 快速傅里叶变换（Fast Fourier Transform） |
| FPGA | 现场可编程门阵列（Field Programmable Gate Array） |
| GAs | 遗传算法（Genetic Algorithms） |
| GMM | 高斯混合模型（Gaussian Mixture Model） |
| GPU | 图形处理器（Graphics Processing Unit） |
| I/Q | 同相与正交（In-phase and Quadrature） |
| IP stack | 网络协议七层栈（Internet Protocol 7-layer stack） |
| ISR | 情报、侦察与监视（Intelligence, Surveillance, Reconnaissance） |
| kNN | k近邻（k-Nearest Neighbor） |
| MAC | 中介访问控制层（Medium Access Layer in the IP stack） |
| MANET | 移动自组网（Mobile Ad-hoc Network） |
| MDP | 马尔可夫决策过程（Markov Decision Process） |
| ML | 机器学习（Machine Learning） |
| NLP | 自然语言处理（Natural Language Processing） |
| PHY | 物理层（Physical layer in the IP stack） |
| POI | 被截获概率（Probability of Intercept） |
| POMDP | 部分可观测马尔可夫决策过程（Partially-observable MDP） |
| Pd | 探测概率（Probability of detection） |
| Pfa | 虚警概率（Probability of false alarm） |
| QoS | 服务质量（Quality of Service） |
| RF | 射频（Radio Frequency） |
| RL | 强化学习（Reinforcement Learning） |
| RNN | 循环神经网络（Recurrent Neural Network） |
| RTS | 请求发送（Request-to-Send） |
| SA | 态势评估（Situation Assessment） |
| SAR | 合成孔径雷达（Synthetic Aperture Radar） |
| SD | 场景驱动器（Scenario Driver） |
| SO | 策略优化器（Strategy Optimizer） |
| SDR | 软件定义无线电（Software Defined Radio） |
| SEI | 特定辐射体识别（Specific Emitter Identification） |
| SVM | 支持向量机（Support Vector Machines） |
| SWaP | 尺寸、重量与功耗（Size, Weight, and Power） |
关于作者¶
Karen Zita Haigh博士是Mercury Systems公司的人工智能首席技术研究员（fellow chief technologist, AI），致力于推动人工智能（AI）和机器学习（ML）在具身嵌入式系统（embodied embedded systems）中的应用。Haigh博士曾任职于霍尼韦尔（Honeywell）、BBN和L3公司，在各类复杂系统领域拥有丰富经验，包括认知射频（Cognitive RF）系统、智能家居、网络安全、喷气发动机、炼油厂以及航天系统。Haigh博士是三个如今在全球广泛应用领域的先驱，其在这些领域的工作如下所述：
-
自动驾驶车辆（Autonomous vehicles，闭环规划与自主机器人上的机器学习）：Haigh博士在卡内基梅隆大学（Carnegie Mellon University）攻读博士学位期间，首次将人工智能风格的符号规划（symbolic planning）部署到真实硬件机器人上，并利用机器学习更新规划模型（即首次实现了自主机器人同时进行规划与学习的完整闭环）。这一能力对当今全球开展的所有自动驾驶车辆研究至关重要。
-
面向老年人照护的智能家居（Smart homes for elder care，通过被动行为监测辅助老年人居家生活）：Haigh博士曾担任霍尼韦尔公司“独立生活助手”（Independent LifeStyle Assistant™，ILSA）项目首席研究员。ILSA是一套智能、自适应的家庭自动化系统，具备先进的态势感知（Situation Awareness, SA）与决策（Decision Making, DM）能力，能够综合分析来自多种传感器、医疗设备和“智能”家电的数据，帮助体弱或年长用户安全地在家中独立生活。ILSA是一个多智能体系统（multiagent system），集成了统一的感知模型、基于概率推导的态势感知、意图识别、分层任务网络响应规划（hierarchical task network-response planning）、实时动作选择控制、机器学习以及人因工程（human factors）。ILSA是该领域首个此类系统，而该领域如今已成为研究热点，并促使霍尼韦尔在该方向进行了战略性收购。
-
认知射频系统（Cognitive RF systems，利用机器学习控制复杂的多目标通信系统）：Haigh博士在认知射频系统领域已深耕超过15年（参见示例7.1）。她为美国国防高级研究计划局（DARPA）资助的“自适应动态无线电开源智能团队”（Adaptive Dynamic Radio Open-source Intelligent Team, ADROIT）项目设计了认知控制器。ADROIT是首个已知在真实环境中（非仿真）利用机器学习动态控制移动自组织网络（MANET）的系统。此外，Haigh博士还为DARPA的“通信专家”（CommEx）项目设计了认知引擎（cognitive engine），该引擎可在面对先前未知的干扰和电子干扰（jamming）条件下优化通信网络性能。CommEx项目首次实现了嵌入式电子防护（electronic protect）系统在任务执行过程中的实时在线学习（real-time in-mission learning）。
这是Haigh博士的第三本书。此前，她于1997年在线出版了《晚餐合作社食谱集》（The Dinner Co-op Recipe Collection），并于2008年与Dana Moore和Michael Thome合著、由Wiley出版社出版了《脚本化你的世界：Second Life脚本官方指南》（Scripting Your World: The Official Guide to Scripting in Second Life）。
Julia Andrusenko 是约翰斯·霍普金斯大学应用物理实验室（Johns Hopkins University APL）的高级通信工程师，同时也是战术通信系统（Tactical Communications Systems）小组的总工程师。Andrusenko在通信理论、无线网络、卫星通信、射频（RF）传播预测、通信系统脆弱性、通信系统计算机仿真、进化计算、遗传算法/编程（genetic algorithms/programming）、多输入多输出（MIMO）技术以及毫米波（millimeter-wave）技术等领域拥有超过19年的经验。她还在为多种先进商用通信系统和军用数据链开发电子战（Electronic Warfare, EW）方法方面具有丰富经验。Andrusenko发表了大量技术论文，并合著了《无线互联网络：通过Wiley/IEEE出版社理解互联网络挑战》（Wireless Internetworking: Understanding Internetworking Challenges）一书。Andrusenko女士在费城德雷塞尔大学（Drexel University）获得电气工程学士和硕士学位。她是IEEE通信学会（IEEE Communications Society）会员，同时也是IEEE 1900.5动态频谱接入应用中认知无线电管理策略语言与架构工作组（IEEE 1900.5 Working Group on Policy Language and Architectures for Managing Cognitive Radio for Dynamic Spectrum Access Applications）的投票成员。
索引（Index）¶
Ablation testing, 198-200, 216 Ablation trials, 174, 199 Accuracy confusion matrix, 202 contextual factors, 212 as metric, 203 weighted, 204-5 Accuracy computation about, 200 classification and confusion matrices, 201-5 regression and nRMSE and, 201, 202 strategy performance evaluation and, 206-9 See also Test and evaluation Action monitoring, 140 Action selection, 150 Action uncertainty, 117-19 Active learning, 149 Activity recognition, 72 Adequacy, 206-9 Adjustable autonomy, 29-30 ADROIT, 10, 184-85, 186 Adversarial intent recognition, 72 Adversarial plan recognition, 74 Adversarial testing, 216 Algorithmic game theory, 126 Anomaly detection, 69-71 Ant colony optimization (ACO), 92 Anti-access area-denial (A2/AD), 147 Anytime algorithms, 99-100
Architectures ADROIT, 184-85, 186 broker approach, 185, 186 cognitive, 150 hardware, 189-91 modular approach, 185 software (interprocess), 183-87 threading and shared memory approaches, 187-89 Argumentation, 120, 130 Artificial intelligence (AI) domain characteristics and, 14 ensemble methods, 46 EW domain challenges viewed from, 6-13 generalization and meta-learning, 48-49 heuristics, 87 randomized search, 87 subfields, 42 symbolic, 47 traditional solutions versus, 13-16 Artificial neural networks (ANNs) about, 44 architectures, 45-46 DeepNets, 44-45 elements of, 44-45 in-mission learning, 150-51 target tracking and, 74 verification, 213 ASICs, 189-91, 223 Association, 71
Autoencoders, 46 Automatic target recognition and tracking, 74-75
Back-propagation algorithm, 44 Bagging, 46 Barrier-to-adoption, 2 Battery lifetime constraints metric, 31 Battle damage assessment (BDA) about, 141 as core concept, 6 effectiveness, 141 feedback computation by, 32 metrics, 36 SA for, 7 scoring engine, 141 Bayesian-model averaging, 46 BBN SO, xiii, 142-43, 149, 151, 172, 198-99, 207 Bias, data, 170-71 Bit error rate (BER), 30, 32 Black-box attacks, 176 Bounded rationality, 91
Causal models, 51 Causal relationships, 71 CEESIM, 228 Centralized reasoning, 100, 101 Characterization, 58, 59-60 Class bias, 170 Classical planning, 113, 115 Classification confusion matrices and, 201-5 defined, 58 open-set, 48 waveform, 60-61 Closed-loop test environment, 224 Clustering environments, 24-28 Cognition loop, 2-4 Cognitive architectures, 150 Cognitive EW systems about, 2 CR and, 11-12 design questions, 12-13 distributed, designing, 11 evaluating, 209-17 introduction to, 1-17 learning assurance and, 209-17 Cognitive radio (CR), 10, 11-12 Collective anomalies, 70
Commanders, 129 Commander's intent (CI), 129 Communication-free learning, 102 Communications, 121, 122, 225 Complex anomalies, 70 Computational learning theory (CLT), 214 Concept drift, 70 Conditional planning, 113, 147 Confusion matrices about, 201 accuracy, 202 classification and, 201-5 illustrated, 204 modified, 206 multiple classes and, 204 precision and, 203 recall and, 203 weighted accuracy and model quality and, 204-5 Consensus propagation, 102 Controllables about, 21, 28 adjustable autonomy and, 29-30 costs, 29 hierarchical, 28 intentional exposure of, 28 values, 29 Convex transductive experimental design (CTED), 174 Convolutional neural networks (CNNs) about, 45 in-mission learning, 150-51 radar emitters using, 60 SVMs and, 75 Cooperative games, 126 Cost factors, 30 Cost metric, 31 Counterfactuals, 71 Course-of-Action Display and Evaluation Tool (CADET), 118 CPUs, 151, 189-91, 223 Cross-entropy method (CEM), 93-94
Data bias, 170 cleaning, 224 coreset of, 173 credibility of, 169 curation, 169 in embedded system, 171, 172 forgetting, 171 interoperability, 162 lifecycle, 160 metadata, 160, 162-67 offline, 223-24 provenance of records, 169 raw, 166 replaying with SD, 197 required, identifying, 223 "supply chain," 160 transmitter, 175 See also Tools and data Data augmentation about, 174-75 adversarial attacks and, 176 approaches, 175-76 oversampling and, 175 Data distribution service (DDS), 185 Data diversity, 171, 172-74 Data engineers, bringing in, 223 Data fusion about, 65 approaches, 66-67 challenges, 65 distributed, 68-69 JDL/DFIG model and, 66 localization example, 67-68 multi-intelligence, 65-69 scalability and, 67 See also Electronic support (ES) Data management about, 159-60 data augmentation and, 174-76 diversity and, 171, 172-74 forgetting and, 176-77 goals, 161 learning assurance and, 211 metadata, 162-67 practice, 171-78 process, 160-69 security and, 177-78 semantics, 160, 161, 167-68 traceability, 169 Data security, 177-78 Datasets, ML, 227 Data sharing, 163 Decentralized coordination, 100, 102 Decision confidence metric, 30 Decision-making (DM) about, 1 AI-based, 15 in cognition loop, 3 for EA, 8-10 for EBM, 8-10 for EP, 8-10 EW design question, 12 planning and scheduling techniques and, 15 problem components, 21 SA and, 184 unified approach to, 125 Decision theory, 86-87 Decision trees, 51 DeepNets about, 44-45 design trade-offs, 51 early stopping and activation dropout, 49 latent feature identification, 48, 49 learning before missions, 151 Deep neural networks (DNNs), 214-15 Deep Q-Networks (DQNs), 152-53 Dempster-Shafer theory, 120 Differential privacy (DP), 177 Direct assessment, 217 Discounted rewards, 118-19 Discretization, 178 Distributed data fusion, 68-69 Distributed optimization, 100-102 Distribute Operations (DistrO) program, 147 Diversified minibatch stochastic gradient descent (DM-SGD), 174 Diversity, data, 171, 172-74 Domain of expertise, 49 Domain ontology, 167 Dynamic spectrum access (DSA), 71 Dynamism, 15-16
Effectiveness concepts, 30 Electronic attack (EA) about, 1 from AI standpoint, 2 anytime algorithms and, 99-100 decision-maker and, 85-86 distributed optimization, 100-102 DM for, 8-10 EP and, 85 in EW, 5 optimization and, 87-96 planning and, 87 scheduling and, 87, 96-99
Electronic battle management (EBM) about, 1, 111 automated, 111 DM for, 8-10 in EW, 5 game theory and, 126 human-machine interface (HMI) and, 127-31 planning and scheduling techniques and, 113-25 Electronic protect (EP) from AI standpoint, 2 anytime algorithms and, 99-100 decision-maker and, 85-86 distributed optimization, 100-102 DM for, 1, 8-10 EA and, 85 in EW, 5 optimization and, 87-96 planning and, 87 scheduling and, 87, 96-99 Electronic support (ES) about, 57-58 anomaly detection, 69-71 causal relationships, 71 emitter classification and characterization, 58-63 in EW, 5 intent recognition, 72-75 multi-intelligence data fusion, 65-69 performance estimation, 63-65 SA for, 1, 7-8 Electronic warfare (EW) brief introduction to, 4-6 challenges of, 1, 6-13 core concepts, 5-6 defined, 5 objectives, 90 user requirements, \(10-11\) Electronic warfare officers (EWOs), 11 Embedded systems, data in, 171, 172 Emitter classification and characterization behavior characterization, 59-60 challenges impacting, 58-59 specific emitter identification and, 61-63 waveform classification, 60-61 Emitter descriptor words (EDWs), 61 Engine for Lexicalized Intent Reasoning (ELEXIR), 73 Enhanced mobile broadband (eMBB), 225
Ensemble methods, 46, 49, 144, 145-46 Entropy measurement and management, 174 Environments closed-loop test, 224 clustering, 24-28 description of, 26 in-mission learning and, 199 RF, 24 EW Planning and Management Tool (EWPMT), 130, 131 Exclusion bias, 170 Execution monitoring actions, 139-40 in cognition loop, 3 EW battle damage assessment, 141-44 Experiment reproducibility, 162
Feature engineering, 59 Federated learning system, 178 Find-fix-track-attack-assess (FFTTAA), 6 5G use cases, 225-27 Forgetting, 176-77 FPGAs, 28, 151, 164, 189-91, 223 Future-proofing, 162 Fuzzy logic, 120
Game theory, 126 Gaussian mixture models (GMMs), 74 General algorithms (GAs), 93 Generalization, 48, 178 Generative adversarial networks (GANs), 46, 176 Goal monitoring, 140 Goal recognition, 72 GPUs, 189-91 Gray-box attacks, 176 Ground-truth data file (GTDF), 194-96, 206
Hardware architectures, 189-91 Hardware design, 13 Heuristics, 22, 47, 87, 91, 94, 115 Hidden Markov models (HMMs), 71 Hierarchical task networks (HTN) planning, 116-17 Human-machine interface (HMI) about, 127 design, 127 HMT-enabled, 128 interaction points, 128-30 technological challenges of, 127-28 See also Electronic battle management (EBM) Human-machine teaming (HMT), 127 Hybrid ML, 47 Hyperparameters, 48-49
Information uncertainty, 120-22 Informativeness analysis, 174 In-mission learning about, 148-50 cognitive architectures, 150 Deep Q-Networks (DQNs), 152-53 Markov decision processes (MDPs), 152 Multiarmed Bandit (MAB), 151-52 neural networks, 150 in novel environments, 199 SVMs, 151 In-mission planning about, 139 execution monitoring, 139-44 learning, 148-53 replanning, 144-48 See also Planning In-mission replanning about, 144-46 conditional planning and, 147 failures and, 147 illustrated, 146 planners, 146-47 plan stability, 147 Intelligence, reconnaissance, and surveillance (ISR), 2 Intent recognition, 72-74 Interoperability, 162 Interval arithmetic, 215 Intervention, 71
Joint Directors of Laboratories/Data Fusion Information Group (JDL/DFIG) model, 66
Key performance indicators (KPIs), 67 \(K\)-fold cross-validation, 198 \(K\)-nearest neighbor ( \(k \mathrm{NN}\) ), 48, 51 Kohonen networks, 46
Label anonymization, 177 Latent embedding, 48 Learning active, 149 in cognition loop, 4 communication-free, 102 EW design question, 12-13 in-mission, 148-53, 199 Occam, 216 PAC, 214 preference, 130 reinforcement (RL), 42 semi-supervised, 42 supervised, 41-42, 148 unsupervised, 41-42 verification, 213 See also Machine learning (ML) Learning algorithms, 41-42 Learning assurance about, 209 data management and, 211 data verification and, 212 empirical and semiformal verification methods, 215-17 formal verification methods, 214-15 inference model verification and, 212 learning process management and, 211 learning process verification and, 210 model implementation and, 211 model training and, 210 multiple approaches, 214 as quality control process, 209 "W," 209, 211 See also Test and evaluation Leave-one-out testing, 198 LEXrec, 73-74 Limitation studies, 216 Logistic regression, 51 Low probability of intercept/low probability of detection (LPI/LPD), 31
Machine learning (ML) about, 1, 41-43 conclusions, 52 in EW system, 41 hybrid, 47 libraries, 25 open-set classification, 48 primer, 41-52 rule extraction, 4 See also ML algorithms Majority voting, 46 Management information base (MIB), 28
Markov decision processes (MDPs) about, 64 action uncertainty and, 117-19 in-mission learning, 152 RL and, 150 utility function for, 118-19 Massive machine-type communications (mMTC), 225 MATLAB RF Toolbox, 228 Maximum-margin hyperplane, 44 Measurement bias, 170 Meta-analysis, 162 Metadata about, 160 air traffic example, 164 annotation, 163-64, 165 global features, 163-64 hardware, firmware, software details, 166 RF fingerprinting example, 165 tasks supported by, 162-63 Metaheuristic methods, 91 Metalearning about, 94 benefits, 95 in design phase, 211 MAB heuristic and, 152 in model sensitivity estimation, 216 optimization, 94-96 in search process, 95-96, 132 Metrics about, 22, 30 computational independence, 37 computation of, 33 effectiveness, 32 to evaluate performance, 30-32 non-independent, 33 performance, 6, 63-64, 90, 141, 148 potential, examination of, 31 types of, 30-31 values, 31-32 weight, 31 See also specific metrics Mission data file (MDF), 28 Mission planners, 130 Mistake-bound model, 216 Mixed-initiative planning, 130 ML algorithms about, 43 artificial neural networks (ANNs), 44-46 design trade-offs, 51 instance-based methods, 43 model-based methods, 43 questions to address when choosing, 50 support vector machines (SVMs), 43-44 trade-offs, 49-52 See also Machine learning (ML) ML datasets, 227 ML toolkits, 222,224-27 Model monitoring, 140 Model-predictive control (MPC), 148 Modified confusion matrices, 206 Monte Carlo tree search, 73 Multiarmed Bandit (MAB), 151-52 Multi-objective optimization, 88-90 Multiple-version dissimilar software, 213
Native Bayes classifier, 51 Neural networks, in-mission learning, 150-51 Nodes, 37 Noise, 171 Noncooperative games, 126 Normalization, 178 Normalized RMSE (nRMSE), 201
Objective function. See Utility functions Observables about, 21, 24 clustering environments, 24-28 elements of, 23 visualization tools, 23 Observe, orient, decide, and act (OODA), 4 Observe-Orient-Plan-Decide-Act-Learn (OOPDAL), 6 Observer (confirmation) bias, 171 Occam learning, 216 Ontology, 167 Ontology-based knowledge graphs, 215 Open-set classification, 48 Optimization about, 87-88 ant colony (ACO), 92 distributed, 100-102 metalearning, 94-96 multi-objective, 88-90 particle swarm (PSO), 92 problem, 87 randomized algorithms, 92-94 schedulers and, 99 Overfitting, 48 Oversampling, 175
Pareto-optimal frontier, 88-89 Partially observable Markov decision processes (POMDPs), 96, 117, 119 Particle swarm optimization (PSO), 92 Perceptual attention, 150 Performance estimation, 63-65 Performance metrics, 6, 63-64, 90, 141, 148 Perturbation, 178 Plan monitoring, 140 Planning about, 87, 113 action uncertainty, 117-19 algorithms, 114 basics, 115-16 classical, 113, 115 conditional, 113, 147 contingent, 114 hierarchical task networks (HTN), 116-17 impact of communications on, 121 information uncertainty, 120-22 in-mission, 139-53 mixed-initiative, 130 multiple timescales and, 125 probabilistic, 113 resource management and, 122-25 support of, 112 temporal, 113, 122-25 See also Electronic battle management (EBM) Planning Domain Definition Language (PDDL), 115 Plan recognition, 72-73 Plan stability, 147 Playback interfaces, 130 Policy bundles, 33 Position, navigation, and timing (PNT), 2 Preference learning, 130 Preference relations, 130 Probabilistic hostile agent task tracker (PHATT), 73 Probabilistic planning, 113 Probabilistic structural causal models (SCMs), 71 Probability of detection constraints metric, 31 Probability-of-intercept (POI), 33, 36 Probably approximately correct (PAC) learning, 214 "Pseudo-classes," 174
Quality-of-service (QoS) metric, 31, 36
Radar classifier, 63 Rapid-response engine (RRE), 142-43, 187 Raytheon's EW Planning and Management Tool (EWPMT), 130, 131 Reader's guide, this book, 16-17 Recall bias, 171 Recurrent neural networks (RNNs), 45 Reinforcement learning (RL), 42 Reprogramming, EW, 6 Resource management approaches to, 123-25 augmenting planning with, 122-25 benefits of, 123 game-theoretic approaches to, 123-24 joint optimization and, 88 unified approach to, 125 RF data-generation tools, 227-28 RF environments, 24 RL systems, 148 Root-cause/ground-truth analysis, 162 Root-mean-square error (RMSE), 201, 202
Sample (selection) bias, 170 Satisfiability modulo theories (SMT), 214-15 Scenario definition, 222-23 Scenario driver (SD) adequacy evaluation, 207 clustering and, 200 data generation, 197-98 driving testing, 197 GTDF and, 194-96 performance evaluation, 200 performance table, 206 replaying data, 197 sequences, 196 structure of, 194, 197 test scenario and, 196 See also Test and evaluation Schedulers, 98-99 Scheduling about, 87, 96 automatic, 97-98 critical path methods, 96 designer's goal in, 98 Scikit-learn code, 25, 27 Security, 13, 126, 160, 171, 177-78 Self-organizing maps (SOMs), 46 Self-organizing networks (SONs), 226
Semantics, 160, 161, 167-68 Semi-supervised learning, 42 Sensitivity analysis/tuning studies, 216-17 Sensor monitoring, 140 Siamese neural nets, 46 Signal Metadata Format (SigMF), 167, 168 Simple anomalies, 69-70 Simplification, 216 Simulated annealing, 93 Situation-assessment (SA) about, 1 for battle damage assessment (BDA), 7 in cognition loop, 2-3 decision-making (DM) and, 184 for ES and EW BDA, 7-8 Software architecture EW design question, 13 interprocess, 183-87 intraprocess, 187-89 scalable, 224 Spectrum situation awareness (SSA), 57, 58 Spoofing, 58 Starting about, 221-22 development considerations, 222-24 steps in, 221-22 tools and data, 224 Statistical learning theory (SLT), 214 Strategy performance evaluation, 206-9 Supervised learning, 41-42 Supply-chain evaluation, 163 Support vector machines (SVMs) about, 43-44 CNNs and, 75 design trade-offs, 51 in-mission learning, 151 modeling performance space, 64 support vectors, 44 Symbolic AI, 47 System designers, 128-29 Systemic bias, 171
Tactical Expert Mission Planner (TEMPLAR), 112 Task reuse, 162-63 Temporal CNNs, 45 Temporal planning, 113, 122-25 Test and evaluation ablation testing, 198-200 about, 193-94 accuracy computation, 200-209 goal of, 193 learning assurance, 209-17 scenario driver (SD), 194-98 Timescales, multiple, 125 Time-to-live thresholds, 28 Tools and data ML data sets, 227 ML toolkits, 224-27 RF data-generation tools, 227-28 Traceability, 169
Ultra-reliable and low-latency communications (URLLC), 225 Uncertainty estimates, 215 Uncertainty metric, 30 Underfitting, 48 Uninformed algorithms, 115 Universal software radio peripheral (USRP) structure, 166 Unsupervised learning algorithms, 41-42 User requirements, EW, 10-11 User stories, 131 Utility functions about, 21, 22 creating, 32-36 design considerations, 36-38 EW system and, 22 for MDPs, 118-19 metrics computation, 33 nodes and, 37 preference learning and, 130 simplification of, 34-35 structure of, 32-33 symbols supporting construction of, 22
Value of information metric, 30 Vapnik-Chervonenkis (VC) dimension, 214 VITA-49, 168
Web ontology language (OWL), 167 White-box attacks, 176, 216 Wireless sensor networks (WSNs), 67
Zero-sum games, 126