정보관리기술사&컴퓨터시스템응용기술사 Vol 9: 인공지능

권영식

저자권영식
-성균관대학교정보통신대학원정보보호학과졸업(공학석사)
-삼성종합기술원연구원
-삼성전자선임/책임/수석연구원
-컴퓨터시스템응용기술사
-정보시스템수석감리원
-정보통신특급기술자
-과학기술정보통신부IT멘토
-데이터관리인증심사원(DQC-M)
-韓(한)·日(일)기술사교류회위원
-http://cafe.naver.com/96starpe운영자

PART1.인공지능(人工知能,ArtificialIntelligence)의개요
1.인공지능(ArtificialIntelligence)의역사
2.인공지능
3.약인공지능(WeakAI),강인공지능(StrongAI),초인공지능(SuperAI)
4.인공지능(AI)의특이점(Singularity)
5.아실로마(ASILOMA)AI(인공지능)원칙
6.규칙기반모델
7.추천엔진(RecommendationEngine)
8.전문가시스템(ExpertSystem)
9.정규표현식과유한오토마타
10.유한오타마타(FiniteAutomata)
11.튜링테스트(TuringTest)
12.에이전트(Agent)-1교시형답안
13.에이전트(Agent)-2교시형답안
14.킬스위치(KillSwitch)
15.트롤리딜레마(TrolleyDilemma)
16.인공지능(AI)윤리의개념,주요사례,고려사항및추진방향
17.이용자중심의지능정보사회를위한원칙

PART2.인공지능알고리즘(Algorithm)
18.유전자알고리즘(GeneticAlgorithm)
19.그리디알고리즘(GreedyAlgorithm)
20.상관분석(CorrelationAnalysis)
21.회귀분석(RegressionAnalysis)
22.로지스틱회귀분석(LogisticRegressionAnalysis)
23.군집분석(ClusterAnalysis)-1교시형답안
24.군집분석(ClusterAnalysis)-2교시형답안
25.계층적군집분석(HierarchicalClustering)
26.자카드(Jaccard)계수
27.해밍거리(HammingDistance)
28.유클리디안거리(EuclideanDistance)
29.유클리디안거리(EuclideanDistance)를계산하시오.
30.마할라노비스거리(MahalanobisDistance)를계산하시오.
31.Apriori(연관규칙)알고리즘
32.지지도(Support),신뢰도(Confidence),향상도(Lift)
33.사례1(TV구입시DVD구입),사례2(우유구입시주스구입)에대해연관규칙(지지도,신뢰도,향상도)을제시하시오.
34.앙상블학습(EnsembleLearning)
35.머신러닝(MachineLearning)에활용,앙상블(Ensemble)기법
36.Bagging과Boosting비교
37.랜덤포레스트(RandomForest)
38.의사결정트리(DecisionTree)
39.K-NN(K-NearestNeighbor)
40.시계열분석
41.시계열분석(ARIMA)
42.SVM(SupportVectorMachine)-1교시형답안
43.SVM(SupportVectorMachine)-2교시형답안
44.베이즈(Bayes)정리
45.크기와모양이같은공이상자A에는검은공2개와흰공2개,상자B에는검은공1개와흰공2개가들어있다.두상자A,B중임의로선택한하나의상자에서공을1개꺼냈더니검은공이나왔을때,그상자에남은공이모두흰공일확률은?(베이즈(Bayes)정리를활용하시오)
46.K-Means
47.DBSCAN(DensityBasedSpatialClusteringwithApplicationNotes)
48.차원축소(DimensionalityReduction)
49.특징추출(FeatureExtraction)
50.주성분분석,PCA(PrincipalComponentAnalysis)
51.독립성분분석,ICA(IndependentComponentAnalysis)
52.마르코프결정프로세스(MarkovDecisionProcess,MDP)
53.은닉마르코프모델(HMM-HiddenMarkovModel)
54.몬테카를로트리탐색(MCTS)
55.Q-Learning
56.Tokenization(토근화),N-gram
57.Word2Vec
58.Word2Vec학습모델,CBOW(ContinuousBagOfWords),Skip-gram

PART3.심층신경망상세
59.일반적인프로그램방식과MachineLearning(기계학습)프로그래밍방식
60.AI(ArtificialIntelligence),ML(MachineLearning),DL(DeepLearning)
61.기계학습(MachineLearning)
62.지도학습(SupervisedLearning)
63.비지도(비감독)(UnsupervisedLearning)학습
64.강화학습(ReinforcementLearning)
65.딥러닝(DeepLearning)
66.MCP(McCulloch-Pitts)뉴런(Neuron)
67.헵규칙(HebbRule)
68.퍼셉트론(Perceptron)
69.아달라인(Adaline-AdaptiveLinearNeutron)
70.활성화함수(ActivationFunction)-1
71.활성화함수(ActivationFunction)-2
72.신경망학습-FFNN(FeedForwardNeuralNetwork)
73.딥러닝(DeepLearning)의파라미터(Parameter)와하이터파라미터(Hyperparameter)를비교하고하이퍼파라미터의튜닝방법을설명하시오
74.역전파법(Back-Propagation)
75.기울기소실문제(VanishingGradientProblem)
76.경사하강법(GradientDescent)
77.과적합(Overfitting)과부적합(Underfitting),적합(Bestfitting)
78.과적합(Overfitting)과부적합(Underfitting)해결방안
79.Dropout
80.ANN(ArtificialNeuralNetwork)
81.DNN(DeepNeuralNetwork)
82.CNN(ConvolutionNeuralNetwork)
83.RNN(RecurrentNeuralNetwork)
84.LSTM(LongShort-TermMemory)
85.GRU(GatedRecurrentUnit)
86.RBM(RestrictedBoltzmannMachine)
87.DBN(DeepBeliefNetwork)
88.DQN(DeepQ-Network)
89.GAN(GenerativeAdversarialNetworks)〈GAN의이해〉
90.DL4J(DeepLearning4J)
91.혼동행렬(ConfusionMatrix)
92.MachineLearning(기계학습)의평가방법-Accuracy(정확도),Recall(재현율),Precision(정밀도)
93.F1Score

PART4.인공지능활용
94.음성인식기술,ASR(AutomaticSpeechRecognition),NLU(NaturalLanguageUnderstanding)
TTS(TexttoSpeech)
95.음성인식(VoiceRecognition)
96.챗봇(ChatBot)
97.가상개인비서(VirtualPersonalAssistant)
98.패턴인식(PatternRecognition)
99.머신러닝파이프라인(MachineLearningPipeline)
100.자연어처리
101.엑소브레인(Exobrain)
102.엑소브레인(Exobrain)과Deepview기술요소
103.딥뷰(Deepview)
104.SNA(SocialNetworkAnalysis)
105.텐서플로(Tensorflow)
106.파이션(Python)의특징및자료형(DataType)
107.패션의류용이미지를분류하는다층신경망을들려고한다.의류용이미지는바지,치마,셔츠등10가지유형의흑백이미지(32*32pixels)로구성되어있고,학습에투입할이미지데이터는검증및테스트용데이터를제외하고총48,000장이다.입력층,은닉층,출력층의완전연결(fullyconnected)3계층으로구성되어있고은닉층의뉴런개수는100개일때,다음에대하여설명하시오
가.신경망구성도
나.입력층의입력개수,출력층의뉴런개수,학습할가중치와절편의총개수
다.원핫인코딩(One-HotEncoding)과소프트맥스(Softmax)함수

PART5.기출및예상토픽
108.GPU(GraphicProcessingUnit)와CPU(CentralProcessingUnit)의차이점
109.머신러닝모델은학습과함께검증및평가과정이필요하다
가.교차검증(k-foldCrossValidation)기법에대해설명하시오
나.머신러닝모델의평가방법에대하여설명하시오
110.머신러닝보안취약점에대해설명하시오.
가.머신러닝학습과정에서의적대적공격4가지
나.각각적대적공격의방어기법
111.데이터어노테이션(DataAnnotation)
112.AIaaS(AIasaService)와도입시고려사항
113.전이학습(TransferLearning)
114.Pre-Crime
115.인공신경망의오류역전파(Backpropagation)알고리즘
116.머신러닝(Machinelearning)의학습방법은크게3가지[지도학습(SupervisedLearning),비지도학습(UnsupervisedLearning),강화학습(ReinforcementLearning)]로분류한다.인공지능소프트웨어개발프로세스를V모델기준으로도식화하고관련기술의최신동향및안전취약점을설명하시오
117.인공지능개발과정에서중점적으로점검할항목
118.인공지능데이터품질요구사항
119.몬테카를로(MonteCarlo)트리(Tree)탐색(MCTS)
120.디지털카르텔(DigitalCartel)
121.XAI(eXplainableAI)
122.인공지능(AI)데이터평가를위한고려사항

■책의특징
Part1.인공지능(人工知能,ArtificialIntelligence)의개요
인공지능의역사,인공지능의분류,특이점,인공지능원칙,전문가시스템,튜링테스트(TuringTest),Agent,인공지능윤리등에대한내용으로작성했습니다.[관련토픽-17개]

Part2.인공지능알고리즘
유전자알고리즘,그리디알고리즘,상관분석,회귀분석,군집분석,자카드계수,해밍거리,연관규칙,지지도/신뢰도/향상도,앙상블학습,Bagging과Boosting,RandomForest,DecisionTree,K-NN,시계열분석,SVM(SupportVectorMachine),K-Means,DBSCAN,차원축소,특징추출,PCA,ICA,Q-Learning,Word2Vec등에대해학습할수있도록하였습니다.[관련토픽-41개]

Part3.심층신경망상세
기계학습,지도학습,비지도(비감독)학습,강화학습,DeepLearning,Perceptron론,활성화함수,하이퍼파라미터,역전파법,기울기소실문제,경사하강법,과적합과부적합,Dropout,ANN,DNN,CNN,RNN,LSTM,GRU,RBM,DBN,DQN,GAN,DL4J,혼동행렬,기계학습의평가방법,정확도,재현율,정밀도,F1Score등에대해학습할수있도록하였습니다.[관련토픽-35개]

Part4.인공지능활용
음성인식기술,챗봇(ChatBot),가상개인비서,패턴인식,머신러닝파이프라인(MachineLearningPipeline),자연어처리,엑소브레인(Exobrain)과Deepview,SNA,텐서플로우,파이션의특징및자료형,패션의류용이미지를분류하는다층신경망예시등을수록했습니다.[관련토픽-14개]

Part5.기출및예상토픽
GPU와CPU,교차검증기법,머신러닝모델의평가방법,보안취약점,DataAnnotation,AIaaS(AIasaService),인공지능V모델,인공지능점검할항목,인공지능데이터품질요구사항,XAI,인공지능데이터평가를위한고려사항등에대해학습할수있습니다.[관련토픽-15개]