Interactive Transcribed Dialog Data Normalization
|
|
|
- 은비 오
- 5 years ago
- Views:
Transcription
1 [ Introduction ] Deep Learning 정상근
2 Applications WHAT MACHINES CAN DO? (2015)
3 Video Understanding (Real-time Genre Detection) Google
4 Image Understanding Google
5 DNN for Image Understanding Image to Natural Language (By Google)
6 Semantic Guessing :: DNN 을통해 Symbol 을공간상에 Mapping 가능하게됨으로써 Symbol 들간의관계를 수학적 으로추측해볼수있는여지가있음 Ex) King Man + Woman Queen :: List of Number 가 Semantic Meaning 을포함하고있음을의미 Microsoft, Linguistic Regularities in Continuous Space Word Representations, 2013
7 Semantic Guessing - Demo Ex) korea kimchi china -?
8 Image Completion :: Shape Boltzman machine 을통해학습한모델에 Constraint 을부여하여원하는방식의이미지를복원 :: 데모 -
9 Hand Writing by Machine 기계에의해씌어진글씨 :: 사람의필체를흉내내어, 필기체를직접쓸수있다. :: 데모 -
10 Music Composition :: Recurrent Neural Network 를사용하여악보 Generation :: 데모
![translation. [ http://en.wikipedia.org/wiki/machine_tr anslation] :: http://devblogs.nvidia.](/docs-images/103/157670920/images/11-2.jpg "com/parallelforall/introduction-neural-machine-translation-with-gpus/ :: Neural Network 단위에서 Language to")
11 Neural Machine Translation Bernard Vauquois' pyramid showing comparative depths of intermediary representation, interlingual machine translation at the peak, followed by transfer-based, then direct translation. [ anslation] :: :: Neural Network 단위에서 Language to Language 번역을시도 :: 데모
12 Play for Fun Learn how to play game :: 게임기의메모리를직접읽어서딥러닝을이용해플레이방법을스스로학습
13 Overview ARTIFICIAL INTELLIGENCE
14 Overview Artificial Intelligence & Cognitive Science (1913) 보어 : 원자모델 (1915) 아인슈타인 : 상대성이론 (1936) 튜링 : 튜링머신 (1939~45) 폰노이만 : 컴퓨터구조 (1948) 쉐논 : 이진법, 정보이론 (1955) 촘스키 : 논리적언어학 (1957) 로젠블렛 : Perceptron (1st Neural Network) (1960) Back-propagation Algo. ( 신경망학습 ) (1980) 존설 : Chinese Room 논제 (1989) 버너스리 : Word Wide Web [ Computer Science ] [ Cognitive Science ] 컴퓨터구조정립 기호주의인공지능 (Computational ism) 연결주의인공지능 (Connectionism) 순수통계적인공지능 * 연결주의인공지능 [ 규칙기반 AI ] [ 신경망기반 AI ] [ 통계기반 AI ] [ DNN ] 마음 = Computer 인지과학태동 인지주의 체화된인지주의 심신이원론 심신일원론 * : 사람의두뇌구조를고려하지않고순수통계학적방법으로만인공지능을구현하려는시도
15 Historical View - Artificial Intelligence 컴퓨터구조정립 Rule Based AI Rule Based AI Decision Tree Statistic Only based AI (HMM, CRF, SVM..) NN based AI DNN based AI Data Driven AI 기계는사람만큼지능적일수있다. Strong AI Weak AI 기계는부분적으로만사람의지능을흉내낼수있다.
16 계산주의 Vs. 연결주의 계산주의 (Computationalism) 뇌구조를추상화 / 기호화 개개의기호및그들간의규칙에주목 기호조작을통해 Mental Activity 를설명가능하다고봄 특정영역에특화된규칙을이용한학습추구 연결주의 (Connectionism) 뇌구조자체를저수준에서모델링 외부환경과의자극에따른뉴론의학습에주목 기호조작들만으로는 Mental Activity 를충분히설명하지못한다고봄 여러분야에통용되는일반적학습방법추구 [ Representation ] One-Hot Representation Cat Distributed Representation [0, 0, 0, 1, 0, ] [ 34.2, 93.2, 45.3, ]
17 Artificial Intelligence ( 전통적의미의 ) 사람의지능을기계에구현하려고하는모든시도. 사람의생각, 기억, 이해, 학습, 인지, 조절기능등모든분야를다룸 분야연구비고 Knowledge Representation Knowledge Representation, Commonsense Knowledge Ontology 와연계됨 Planning Automated planning Game, Robotics, Multi-Agent Cooperation Learning Machine Learning 다방면에사용됨 Communication Perception Motion and Manipulation Natural Language Processing Computer Vision, Speech Recognition Robotics Interface :: 전통적의미의 AI 는사람의지능을구현하려는시도. 최근들어사람이부족한지능을강화시켜주는지능에대한연구도활발해짐
18 Machine Learning 지능중 학습 * 에관련된부분을기계에구현하려고하는시도 주목적 : Prediction / Inference Training Time Running Time Known Data Known Responses Model Model New Data Predicted Responses Reproduce known knowledge * 기계를학습시킨다고해서, 꼭사람과같게만든다는것은아님에유의. 사람이전혀못하는것도기계는잘하게만들게학습시키는것도 machine learning 의목표 - Data 가공, 추출과정, 최종결과물해석에 Data Mining 기법이사용될수있음 응용분야 :
19 Data Mining 데이터에서 Pattern 을발견하고자하는시도 주목적 : Pattern Discovery Unknown Data Miner Pattern Produce unknown knowledge - Unknown knowledge 를발견한다는측면에서 Machine Learning 분야중 Unsupervised Learning 과긴밀한연관이있음. - Data Mining 의일부는 사람의지능적발견 에관련된부분도있지만, 사람이발견하지못하는것 에대한것도많음 응용분야 :
20 Machine Learning and Other AI Tasks 최근의 AI 문제들은대부분 Empirical Data 를활용하여풀고자하는경향이있음. 이러한측면에서 Machine learning 의기술이다른분야로전파되기도함. 혹은, 다른분야의연구결과물이 Machine Learning 의새로운문제발견과해결에영향을주기도함 Ex) 음성인식 Task 와 ML 음성녹음데이터 전사 Script HMM Training (EM Algo.) HMM Model HMM model training 에사용되는 EM Algorithm 은대표적인 Machine Learning 의파라미터훈련알고리즘중하나 Ex) 형태소분석 Task 와 ML 형태소태깅데이터 CRF (LBFGS Algo.) CRF Model 형태소태깅의대표적인기술중하나인 CRF 는 Machine Learning 커뮤니티에발표되었고, 자연어처리전분야에성공적으로적용된대표적인알고리즘 ML Community 에서는대부분수학적, 통계적고도화위주의연구를진행하며, Toy Example ( 자연어, Vision, Speech..) 에해당알고리즘을적용하여성능이향상되는것을보이는방법으로연구를진행함. 혁신적이면서도유망한 ML 기술이발표되면, 다른분야에해당기술이전파되는경우가있음
21 Summary Statistics quantifies numbers Data Mining explains patterns Machine Learning predicts with models Artificial Intelligence behaves and reasons Cognitive Science is the scientific study of the mind and it s process Cognitive Science 인간 - 인간, 인간 - 동물, 인간 - 인공물간의정보처리활동을다룸 지능을다룸 ( 사람의지능 사람 +α) Artificial Intelligence 학습능력을다룸 Machine Learning 많은부분의기술을공유 Data Mining
22 Statistical View ARTIFICIAL INTELLIGENCE PROBLEMS
23 Problem Formulation ( 통계적입장에서본문제정의 ) 최종형태는세가지형태에서크게벗어나지않음 Selection Grouping Learning Act Recognize Understand Learn Plan Communicate Human Process Raw data
24 Selection & Grouping 하나를고르는문제 Classification Selection 여러개를골라서순서대로세우는문제 Ranking 여러개를 Grouping Clustering Grouping 여러개를구조화하여 Grouping Hierarchical Clustering
25 Statistical Approach to Classification 가장단순한 Classification 은 선긋기 문제이다. 두개의그룹을나누는선을긋는문제 좌 / 우, 상 / 하, 내 / 외, + / - Y = ax + b 문제 (Linear)
26 Regression 회귀분석 Regression : 사전적의미 - Go back to an earlier and worse condition :: Francis Galton (1822 ~ 1911) 은부모의키와자녀의키사이의상관관계 (928 명 ) 를조사하는과정에서, 키는무한정커기거나작아지는것이아니라 전체키평균으로돌아가려는경향 이있음을발견. 이를회귀분석이라명명함. :: Karl Pearson (1903) 은 1078 명의부자키를조사하여선형함수관계를도출 아버지키 = * 아들키
27 Support Vector Machine(SVM) 가장성공적인 Classifier 중하나 선을어떻게그을것인가? Maximum Margin (1963 Vapnik) 직선으로구분이안되는문제는어떻게풀것인가? Kernel Trick (1992 Vapnik) : support vector :: 두개의 Class(black/blank) 를구분짓는선을그을때그선과 support vector 들사이의거리가최대가되도록 :: 원래공간에있던각점들을 Kernel Function 을이용해새로운차원으로이동시키면, 직선으로구분가능한문제로바뀔수있다.
28 Statistical Learning 다양한형태가있지만대부분아래의형태를따른다. Feature Extraction Prediction Evaluation Function Distance between (Reference ~ Prediction) Prediction How closely predicted? Parameter Update θ θ θ Inference Learning
29 Feature Design / Evaluation Function / Parameter Update Feature Design Features describe a real world object 잘설계된 feature 를쓰는것이통계적기계학습의핵심 (Feature Engineering) 최근에는이조차도기계가알아서학습 (DNN) Evaluation Function Distance between Predication and Reference Parameter Update How to update parameter to fit data
30 WHY DEEP LEARNING?
31 Why Deep Learning - Learning Representation No more handcraft feature engineering! color = red shape = round leafs = yes dot = yes Numbers - 사과를 사과 로구별짓는표현방식을스스로학습
32
33 Why Deep Learning - Distributed Representation (1) :: DNN 가기존 AI 방법론들에비해큰의미가있는것은실세계에있는실제 Object 를표현할때 Symbol 에의존하지않는다. [ Representation ] One-Hot Representation Cat Distributed Representation [0, 0, 0, 1, 0, ] [ 34.2, 93.2, 45.3, ]
34 Why Deep Learning - Distributed Representation (2) Apple = 001 Pear = 010 Ball = 100 Distance(Apple ~ Pear) = Distance(Apple ~ Ball) - 유사한것은 유사하게 표현되어야함 - Curse of Dimensionality 를극복가능해야
35 Why Deep Learning - Reusable Learning Result 정규화 형태소분석 정규화 형태소분석 구문분석 구문분석 - 기존에는각각의문제를풀었고, 그결과물은유기적결합이어려웠음 - Deep Learning 은다른도메인에서풀었던문제를현재문제에그대로가져다사용할수있음
36 Why Deep Learning - Design Network Solve Problem Meaning : Apple on Plate ~ on ~ Plate 인식 Apple 인식 - 어떠한 Intelligence 를어떻게결합하는가에따라새로운문제를풀어낼수있다.
37 Why Deep Learning - Unlabeled Data >>>>>>>>>>>>>>>>>> Tagged Data [ Previous Machine Learning ] [ Deep Learning ] Small Tagged Data Large Raw Data P(x) Small Tagged Data P( y x) P( y x) - 수많은 Unlabeled Data 를활용할수있는 learning 방법
38 Review NEURAL NETWORK
39 One Learning Algorithm Nero-Rewiring Experiment Auditory Cortex Auditory cortex learns to see [Roe et al., 1992] :: 청각과연결되어있는신경망을끊고, 이부분에시신경과연결된신경망을연결하면, Auditory Cortex 가 볼수 있게된다. Slide from Andrew Ng
40 One Learning Algorithm Somatosensory Cortex Somatosensory cortex learns to see :: 촉감과연결되어있는부분을끊고, 이를시신경과연결된신경망에연결하면 Somatosensory Cortex 가 볼수 있게된다. [Metin & Frost, 1989] Slide from Andrew Ng
41 One Learning Algorithm Low resolution gray-scale Image Seeing with your tongue 전기신호로바꿈 해당전기신호를혀에계속해서전달 어느순간부터혀로 볼수 있게됨 Slide from Andrew Ng
42 Neurons Firing Off in Real-time
43 Neurons in Brain 뉴론은계속해서시그널을받아 - 그것을조합 - sum 하고, - 특정 threshold 가넘어서면 - fire 를한다.
44 Illustrative Example ( Apple Tree ) Size Day - 어떤사과나무에대해서몇년에걸쳐날짜별로사과들의크기를측정, 기록 - 농부는특정크기가넘을때만시장에사과를내다팔수있다고할때, - Q : 올해 Day -50 에사과를내다팔수있을까? 없을까?
45 Illustrative Example Default size = 5 size = 10 size = 15 size = 20 size = 25 If size > 30, sell an apple! Sell Day 0 Day 10 Day 20 Day 30 Day 40 상황 1 : 작년까지이사과나무는위의경향대로사과열매를맺었다. 조건 : 사과의크기가 30 이넘으면팔수있다. Question : 올해 Day-50 에사과를팔수있을까? Regression Problem
46 Illustrative Example Default size = 5 size = 10 size = 15 size = 20 size = 25 If size > 30, sell an apple! Sell Day 0 Day 10 Day 20 Day 30 Day y = ax + b Size = 0.5*day Activation point to sell an apple! Regression learn the parameter a and b from the data
47 Apple Selling Example Neural Network Framework y = ax + b 입력값을변형해새로운값계산 (day size ) Y = WX + b 정규화 If y > 30 sell an apple 새로운값을다시해석해최종결과산출 (size 팔까 (1)/ 말까 (0)) Activation function Step Function
48 Perceptron Simplest ANN (1) input 0 processor +1 or -1 output input 1 A perceptron consists of one or more inputs, a processor, and a single output. Step 1: Receive inputs. Step 2: Weight inputs. Step 3: Sum inputs. Step 4: Generate output. The Perceptron Algorithm: 1) For every input, multiply that input by its weight. 2) Sum all of the weighted inputs. 3) Compute the output of the perceptron based on that sum passed through an activation function (the sign of the sum). Sum = W 0 * input 0 + W 1 * input 1 if (sum > 0) return 1; else return -1;
49 Perceptron Learning Rule :: 동영상 ( ) :: 기울기 (w) 와 Bias(b) 를계속해서바꿔가면서 O 와 X 를구분하는선을탐색
50 Limitation of Perceptron Perceptron can do. Linearly Separable! Perceptron cannot do. Not Linearly Separable!
51 What if multiple perceptron? input OR (solver) output XOR input NOT AND (solver)
52 Multilayer Perceptron (MLP) The single-hidden layer Multi-Layer Perceptron (MLP). An MLP can be viewed as a logistic regressor, where the input is first transformed using a learnt non-linear transformation [ Softmax Function ] G : Scoring Function for top-layer [ tanh Function ] S : Activation Function for hidden layer x D is the size of input vector x L is the size of output vector f(x) Feed Forward Propagation
53 학습진행방향은? 오류 정답 ~ 비교 예측한답 오류가작아지는방향으로
54 정답 ~ 비교 예측한답 오류 오류가작아지는방향이란어느쪽인가? 얼마나나의지식을고쳐야오류를작아지게할수있을까?
55 방향을결정하는방법 (1) 이곡선이전체의오류를표현한다고하면 이지점에서의오류가작아지는방향을결정해야한다.
56 방향을결정하는방법 (2) 이지점에서의기울기방향을구해서, 기울기가작아지는방향으로간다면, 오류를작게할수있을것 미분 기울기 = Gradient
57 Gradient Descent A brief introduction to neural network, David Kriesel, dkriesel.com
58 Slide from Andrew Ng Gradient Descent Best-case J( 0, 1 ) 1 0
59 Slide from Andrew Ng Gradient Descent Local Minimum J( 0, 1 ) 0 1
60 Minima and Maxima
61 어떻게오류를고칠것인가? 오류에기여 오류에기여 얼마만큼오류에기여했는가? = 미분 오류 오류를수정하는방향으로얼마나움직일까? = 학습가중치 x 미분값 아래쪽으로반복 = Back - Propagate 해서오류수정
62 MLP Training (Weight Optimization) - How to learn the weights?? Backpropagation Algorithm 최종결과물을얻고 그결과물과우리가원하는결과물과의차이점을찾은후 Feed Forward and Prediction Cost Function 그차이가무엇으로인해생기는지 Differentiation ( 미분 ) 역으로내려가면서추정하여 새로운 Parameter 값을배움 Back Propagation Weight Update Cf) 속도 의미분값이 가속도 가속도 로인해 속도 변화
63 Summary : Neural Network Core Components Output Score S Decision : Scoring Function Hidden Layer z 1 z 2 z 3 Fire : Activation Function Summation : Matrix Production Neuron structure : Edge Connection Visible Layer x 1 x 2 x 3 x 4 x 5 x 6 Sensing : Vector Form Representation Input Raw Data 이해를돕기위해 Single Hidden Layer NN 을표현
64 Summary : Neural Network Process Application Specific 연산 Ex) 예상주식값 예상값과실제값의오류만큼을아래네트워크로전파 Ex) 오류 = 실제 - 예상 1 Matrix 연산 vector Parameter Update 2 vector vector vector vector vector Raw Data Raw Data Raw Data
65 DEEP NEURAL NETWORK
66 Why old ANN was not successful? Initialization Local Minima Computation Power Data Pre-Training Distributed Representation Initialization Techniques Activation Function Understanding ANN Big Data Deep Learning 대표적인 Bottleneck 만표시
67 Remind # of Parameters & Local Minima W = w 11 w 12 w 13 w 21 w 22 w 23 x = w 31 w 32 w 33 x 1 x 2 x 3 a 1 = f(w 11 x 1 + W 12 x 2 + W 13 x 3 + b 1 ) a 2 = f(w 21 x 1 + W 22 x 2 + W 23 x 3 + b 2 ) a 3 = f(w 31 x 1 + W 32 x 2 + W 33 x 3 + b 3 ) In Matrix notation z = Wx + b a = f(z) - 네트워크가깊어지고, 복잡해질수록 parameter 수가많아짐. - Parameter 가많아질수록 Local Minima 에빠질가능성이높아짐
68 Initialization Problem Output Score S W Hidden Layer z 1 z 2 z 3 최초의 Summation Weight 을어떻게결정? W Random Initialization Visible Layer x 1 x 2 x 3 x 4 x 5 x 6 hello 라는 Symbol 의 Vector Form? Input Raw Data
69 Deeper Network, Harder Learning - Network 가깊으면깊을수록최종성능이좋다는것은밝혀짐 - 단, 깊어지면깊어질수록 Error Propagation 이어려워짐 - Vanishing gradient problem
70 Pre-Training Unsupervised Learning Large Raw Data P(x) Pretraining Supervised Learning Small Tagged Data P( y x) - Pretraning 의개발로, NN 의성능이비약적으로향상 - AutoEncoder 계열과, Restricted Boltzmann Machine 계열이있음 - RBM is not covered today.
71
72
73 우리가 사다리 를알고있다면다시복원할수도있지않을까? 생성 = Generation
74 무엇이사다리를생성해내게끔하는가? 사다리 를구성하는핵심, 골격, 정보 (Essence) 핵심정보는원래사다리보다더작은양의정보일것 ( 군더더기없는 )
75 Illustrative Image 원데이터를설명할수있는핵심정보를추출 핵심정보로부터데이터를재생
76 Deep Learning Auto Encoder Encoding Decoding Original Data Abstracted Data H Original Data 원래의데이터 X 를 H 로프로젝션시킨후, H 로부터 X 를다시생성시킴 - 비교 - 압축알고리즘 (Zip, MPEG, PNG ) - Principle Component Analysis (PCA) - Kernel Function in SVM ( original space hyper space ) X X X H X 에서 Difference(X, X ) 가적으면적을수록추상화는완벽하게이루어진것이라생각할수있다. 그러한 Projection 이완벽하게훈련된다면, - Abstracted Data 는그자체로원래데이터를설명하는 Feature 라고볼수있을것이다. - Feature Learning 이자동으로이루어지는것이라할수있음
77 Deep Learning Auto Encoder Input Hidden reconstruction Prediction Z (predication of x ) x Encoding/Decoding Error real value case Cross-entropy bit vector or vectors of bit probability Note that it is purely unsupervised learning!
78 Deep Learning Auto Encoder for Weight Optimization (1) First, you would train a sparse autoencoder on the raw inputs x (k) to learn primary features h (1)(k) on the raw input. Next, you would feed the raw input into this trained sparse autoencoder, obtaining the primary feature activations h (1)(k) for each of the inputs x (k). You would then use these primary features as the "raw input" to another sparse autoencoder to learn secondary features h (2)(k) on these primary features.
79 Deep Learning Auto Encoder for Weight Optimization (2) You would then treat these secondary features as "raw input" to a softmax classifier, training it to map secondary features to digit labels. Next, you would feed the raw input into this trained sparse autoencoder, obtaining the primary feature activations h (1)(k) for each of the inputs x (k). Finally, you would combine all three layers together to form a stacked autoencoder with 2 hidden layers and a final softmax classifier layer capable of classifying the MNIST digits as desired.
80 Deep Learning Auto Encoder Denoising Auto Encoder Encoding Decoding Original Data Noisy Data Abstracted Data Original Data 데이터 X 에 Noise 를추가한 NX 를만들어낸후, NX 를 H 로프로젝션시킨후, H 로부터 X 를다시생성시키도록훈련 - Noise 가추가됨에도불구하고 Original Data 를복구시킬있다면, 그것이 중요한 정보다. - 오류에강건한 Feature 를학습함 Vincent, H. Larochelle Y. Bengio and P.A. Manzagol, Extracting and Composing Robust Features with Denoising Autoencoders
81 Illustration Denoising Auto-Encoding 원본데이터 오류추가 Hidden Layer 복원데이터 Encode Abstracted Info. Decode 원래의데이터를그대로복원할수있도록 Hidden Layer 를학습시킴
82 Deep Learning Auto Encoder Denoising Auto Encoder Vincent, H. Larochelle Y. Bengio and P.A. Manzagol,Extracting and Composing Robust Features with Denoising Autoencoders,
83 Deep Generative Models Generation Abstraction Learning Deep Generative Models, Ruslan Salakhutdinov
84 Generated Numbers by machine Here are the samples generated by the RBM after training. Each row represents a mini-batch of negative particles (samples from independent Gibbs chains) steps of Gibbs sampling were taken between each of those rows.
85 보다직관적인 Deep learning 이해 DEEP LEARNING INTRO 2
86 How? 어떻게 DNN 은사물의특징을스스로파악할수있을까?
87 Latent Variable Deep Neural Network 의핵심 Essence of Modern Machine Learning Hidden Variable
88 x 실세계에존재하는관측가능한것 관측가능 Count 가능 P(x)
89 h 이세상에존재하지않는가상의값 간접적으로추측만가능 무엇이든될수있는값
90 h 가가질수있는전체의미영역
91 x h 두개의변수를묶어주고
92 x h 두개가같이나오도록 P x, h P x, h) = P x h P(h : 같이나타날횟수 P x = h P x h p h dh : continuous P x = P x h p h : discrete h x 와같이잘나타나는 h 가되도록탐색
93 x h x 와연관된 h 가가질수있는전체의미영역 h 가가질수있는전체의미영역
94 x h 여전히 h 는어떤값도될수있음 x 의원인 x 의결과 x 와연관된 h 가가질수있는의미영역 h 가가질수있는전체의미영역
95 x h 같이많이나타나는 h 를찾을때사용되는 x 의개수가 100 개라면? 1,000 개라면? 10,000 개라면? 100,000 개라면? 1,000,000 개라면? 10,000,000 개라면?
96 x h 여전히 h 는어떤값도될수있음 x 의원인 x 의결과 많은수의 x 와연관된 h 가가질수있는의미영역 x 와연관된 h 가가질수있는의미영역 h 가가질수있는전체의미영역
97 또다른변수 y 를연관시켜본다면? x h y
98 x h y 세개가같이나오도록 P x, y, h
99 x h y 많은수의 x, y 와연관된 h 가가질수있는의미영역 많은수의 x 와연관된 h 가가질수있는의미영역 x 와연관된 h 가가질수있는의미영역 h 가가질수있는전체의미영역
100 또다른변수 z 를연관시켜본다면? 또다른변수 z 1 를연관시켜본다면? 또다른변수 z 2 를연관시켜본다면? x h y.... z
101 Latent Variable 의의미영역을축소시킬수도구 1) 많은수의데이터 2) 구조적연관성 x h y z
102 Latent Variable In DNN [ Task ] [ What We Want ] 사과 y Something Describe x and Cause y x x x x x x x x x 3x3 = 9
103 Latent Variable In DNN [ Design Structure ] Under-complete y h h h h h X 를 - Abstraction - Encoding - Semantic Extraction - Summary -. - 하기위해서 - dim(h) < dim(x) x x x x x x x x x
104 Latent Variable In DNN y h h h h h h 하나가모든 x 와연결되도록 x x x x x x x x x
105 Latent Variable In DNN y h h h h h x x x x x x x x x Single Layer
106 Latent Variable In DNN y h h h h h h h h x x x x x x x x x Multilayer - 2
107 Latent Variable In DNN y h h h.. h h h h h x x x x x x x x x Multilayer - N Number of h >>>> number of x, y
108 Intuitive Interpretation of Latent Variable in DNN 사과
109 Intuitive Interpretation of Latent Variable in DNN 사과 y h h Abstraction h h h Abstraction h h h h Abstraction x x x x x x
110 Intuitive Interpretation of Latent Variable in DNN 사과 y Class Something Representation x x x x x x Observation
111 Intuitive Interpretation of Latent Variable in DNN 사과 y Class Representation Representation x x x x x x Observation 잘설계된구조와수많은데이터를통해학습된 ( 찾아낸 ) Latent Variable 은사물의특징을설명할수있게된다.
112 ? Representation Learning 이우리에게주는의미는?
113 Classical Machine Learning Vs. Deep Learning based ML 사람이만든규칙에의한 사물특징추출특징 color = red shape = round AI Algorithm 학습된파라미터에의한 사물특징연산숫자 AI Algorithm Numbers
114 사물 Number Le and Mikolov, Distributed Representations of Sentences and Documents Mikolov et al., Distributed Representations of Words and Phrases and their compositionality Document Level Document Embedding Sentence Level Sentence Embedding Phrase Level Phase Embedding Word Level [ Vision ] [ NLP ] Word Embedding
115 Observation 사물 현상 Semantic 숫자 Representation Learning 은 실세계의사물이나현상을숫자로바꿔주는 Semantic Filter, Semantic Glasses, Semantic Converter 를가능하게한다.
116 Analog to Digital Vs. Object to Semantic Analog to Digital Analog / Digital Converter Object to Semantic Semantic Converter Numbers Analog Digital 과 Object Semantic 의변화구조가유사함에주목
117 미래의정보처리흐름? 과거 Analog / Digital Converter 정보처리 미래 Analog / Digital Converter Digital / Semantic Converter Numbers 정보처리 Digital 정보를 Semantic 정보로바꿔주는 Converter 가 ICT 의핵심자산 Semantic Converter 는단시간에얻어질수있는것이아니며 Copy 도불가능함
118 Deep Learning PROBLEM SOLVING
119 S2S Arithmetic Calculation = Output : Numbers Sequence 2 Sequence Learning Input : Math Expression padding
120 Sequence Modeling for Arithmetic Calculation Output Layer RNN N to M Hidden Layer Input Layer RNN Symbol to Vector Lookup Table One-Hot
121 Sequence Modeling for Arithmetic Calculation Performance (1) 1 Accuracy add sub Difference add sub 50 Iteration 근처에서오차 1 미만으로수렴
122 Sequence Modeling for Arithmetic Calculation Performance (2) 1 Accuracy add sub sub->add add->sub Difference add sub sub->add add->sub 덧셈 예제를통해훈련한모델로시작해, 뺄셈 을훈련시키면훨씬빠르게훈련이수행된다. 마찬가지로 뺄셈 을통해훈련한모델로시작해, 덧셈 ' 을훈련시켜도빠르게훈련수행됨.
123 Pointer Networks Combinatorial Optimization Problem Convex Hull Delaunay triangulation Traveling Salesman Problem Pointer Networks. Vinyals et al. Attention model 활용
124 Pointer Networks - Idea Graph Sequence(Input) Algorithm Deep Learning Solution Sequence(Output)
125 Pointer Network - Performance
126 Pointer Network Performance (TSP Problem)
127 Deep Learning SUMMARY
128 Algorithm Finding
129 Summary Deep Learning = Representation Paradigm Shift Deep Learning = Design Architecture Deep Learning = Data, Data, Data Deep Learning = Beyond Pattern Recognition
130 Q/A 감사합니다. 정상근, Ph.D Intelligence Architect Senior Researcher, AI Tech. Lab. SKT Future R&D Contact : hugmanskj@gmail.com, hugman@sk.com
김기남_ATDC2016_160620_[키노트].key
![김기남_ATDC2016_160620_[키노트].key](/thumbs/85/91488855.jpg "김기남_ATDC2016_160620_[키노트].key")
metatron Enterprise Big Data SKT Metatron/Big Data Big Data Big Data... metatron Ready to Enterprise Big Data Big Data Big Data Big Data?? Data Raw. CRM SCM MES TCO Data & Store & Processing Computational
<4D6963726F736F667420576F7264202D20B1E2C8B9BDC3B8AEC1EE2DC0E5C7F5>
주간기술동향 2016. 5.18. 컴퓨터 비전과 인공지능 장혁 한국전자통신연구원 선임연구원 최근 많은 관심을 받고 있는 인공지능(Artificial Intelligence: AI)의 성과는 뇌의 작동 방식과 유사한 딥 러닝의 등장에 기인한 바가 크다. 이미 미국과 유럽 등 AI 선도국에서는 인공지능 연구에서 인간 뇌 이해의 중요성을 인식하고 관련 대형 프로젝트들을
<313120C0AFC0FCC0DA5FBECBB0EDB8AEC1F2C0BB5FC0CCBFEBC7D15FB1E8C0BAC5C25FBCF6C1A42E687770>
한국지능시스템학회 논문지 2010, Vol. 20, No. 3, pp. 375-379 유전자 알고리즘을 이용한 강인한 Support vector machine 설계 Design of Robust Support Vector Machine Using Genetic Algorithm 이희성 홍성준 이병윤 김은태 * Heesung Lee, Sungjun Hong,
Multi-pass Sieve를 이용한 한국어 상호참조해결 반-자동 태깅 도구
Siamese Neural Network 박천음 강원대학교 Intelligent Software Lab. Intelligent Software Lab. Intro. S2Net Siamese Neural Network(S2Net) 입력 text 들을 concept vector 로표현하기위함에기반 즉, similarity 를위해가중치가부여된 vector 로표현
example code are examined in this stage The low pressure pressurizer reactor trip module of the Plant Protection System was programmed as subject for
2003 Development of the Software Generation Method using Model Driven Software Engineering Tool,,,,, Hoon-Seon Chang, Jae-Cheon Jung, Jae-Hack Kim Hee-Hwan Han, Do-Yeon Kim, Young-Woo Chang Wang Sik, Moon
Ch 1 머신러닝 개요.pptx
Chapter 1. < > :,, 2017. Slides Prepared by,, Biointelligence Laboratory School of Computer Science and Engineering Seoul National University 1.1 3 1.2... 7 1.3 10 1.4 16 1.5 35 2 1 1.1 n,, n n Artificial
<4D6963726F736F667420576F7264202D20C3D6BDC52049435420C0CCBDB4202D20BAB9BBE7BABB>
주간기술동향 2016. 2. 24. 최신 ICT 이슈 인공지능 바둑 프로그램 경쟁, 구글이 페이스북에 리드 * 바둑은 경우의 수가 많아 컴퓨터가 인간을 넘어서기 어려움을 보여주는 사례로 꼽혀 왔 으며, 바로 그런 이유로 인공지능 개발에 매진하는 구글과 페이스북은 바둑 프로그램 개 발 경쟁을 벌여 왔으며, 프로 9 단에 도전장을 낸 구글이 일단 한발 앞서 가는
지능정보연구제 16 권제 1 호 2010 년 3 월 (pp.71~92),.,.,., Support Vector Machines,,., KOSPI200.,. * 지능정보연구제 16 권제 1 호 2010 년 3 월
지능정보연구제 16 권제 1 호 2010 년 3 월 (pp.71~92),.,.,., Support Vector Machines,,., 2004 5 2009 12 KOSPI200.,. * 2009. 지능정보연구제 16 권제 1 호 2010 년 3 월 김선웅 안현철 社 1), 28 1, 2009, 4. 1. 지능정보연구제 16 권제 1 호 2010 년 3 월 Support
ecorp-프로젝트제안서작성실무(양식3)
(BSC: Balanced ScoreCard) ( ) (Value Chain) (Firm Infrastructure) (Support Activities) (Human Resource Management) (Technology Development) (Primary Activities) (Procurement) (Inbound (Outbound (Marketing
(JBE Vol. 23, No. 2, March 2018) (Special Paper) 23 2, (JBE Vol. 23, No. 2, March 2018) ISSN
(Special Paper) 23 2, 2018 3 (JBE Vol. 23, No. 2, March 2018) https://doi.org/10.5909/jbe.2018.23.2.186 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) a), a) Robust Online Object Tracking via Convolutional
High Resolution Disparity Map Generation Using TOF Depth Camera In this paper, we propose a high-resolution disparity map generation method using a lo
High Resolution Disparity Map Generation Using TOF Depth Camera In this paper, we propose a high-resolution disparity map generation method using a low-resolution Time-Of- Flight (TOF) depth camera and
融合先验信息到三维重建 组会报 告[2]
![融合先验信息到三维重建 组会报 告[2]](/thumbs/99/140143252.jpg "融合先验信息到三维重建 组会报 告[2]")
[1] Crandall D, Owens A, Snavely N, et al. "Discrete-continuous optimization for large-scale structure from motion." (CVPR), 2011 [2] Crandall D, Owens A, Snavely N, et al. SfM with MRFs: Discrete-Continuous
untitled
전방향카메라와자율이동로봇 2006. 12. 7. 특허청전기전자심사본부유비쿼터스심사팀 장기정 전방향카메라와자율이동로봇 1 Omnidirectional Cameras 전방향카메라와자율이동로봇 2 With Fisheye Lens 전방향카메라와자율이동로봇 3 With Multiple Cameras 전방향카메라와자율이동로봇 4 With Mirrors 전방향카메라와자율이동로봇
김경재 안현철 지능정보연구제 17 권제 4 호 2011 년 12 월
지능정보연구제 17 권제 4 호 2011 년 12 월 (pp.241~254) Support vector machines(svm),, CRM. SVM,,., SVM,,.,,. SVM, SVM. SVM.. * 2009() (NRF-2009-327- B00212). 지능정보연구제 17 권제 4 호 2011 년 12 월 김경재 안현철 지능정보연구제 17 권제 4 호
(JBE Vol. 21, No. 1, January 2016) (Regular Paper) 21 1, (JBE Vol. 21, No. 1, January 2016) ISSN 228
(JBE Vol. 1, No. 1, January 016) (Regular Paper) 1 1, 016 1 (JBE Vol. 1, No. 1, January 016) http://dx.doi.org/10.5909/jbe.016.1.1.60 ISSN 87-9137 (Online) ISSN 16-7953 (Print) a), a) An Efficient Method
À±½Â¿í Ãâ·Â
Representation, Encoding and Intermediate View Interpolation Methods for Multi-view Video Using Layered Depth Images The multi-view video is a collection of multiple videos, capturing the same scene at
DIY 챗봇 - LangCon
without Chatbot Builder & Deep Learning bage79@gmail.com Chatbot Builder (=Dialogue Manager),. We need different chatbot builders for various chatbot services. Chatbot builders can t call some external
THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE Jul.; 29(7),
THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE. 2018 Jul.; 29(7), 550 559. http://dx.doi.org/10.5515/kjkiees.2018.29.7.550 ISSN 1226-3133 (Print) ISSN 2288-226X (Online) Human
Introduction to Deep learning
Introduction to Deep learning Youngpyo Ryu 동국대학교수학과대학원응용수학석사재학 youngpyoryu@dongguk.edu 2018 년 6 월 30 일 Youngpyo Ryu (Dongguk Univ) 2018 Daegu University Bigdata Camp 2018 년 6 월 30 일 1 / 66 Overview 1 Neuron
RNN & NLP Application
RNN & NLP Application 강원대학교 IT 대학 이창기 차례 RNN NLP application Recurrent Neural Network Recurrent property dynamical system over time Bidirectional RNN Exploit future context as well as past Long Short-Term
Problem New Case RETRIEVE Learned Case Retrieved Cases New Case RETAIN Tested/ Repaired Case Case-Base REVISE Solved Case REUSE Aamodt, A. and Plaza, E. (1994). Case-based reasoning; Foundational
untitled
Math. Statistics: Statistics? 1 What is Statistics? 1. (collection), (summarization), (analyzing), (presentation) (information) (statistics).., Survey, :, : : QC, 6-sigma, Data Mining(CRM) (Econometrics)
R을 이용한 텍스트 감정분석
R Data Analyst / ( ) / kim@mindscale.kr (kim@mindscale.kr) / ( ) ( ) Analytic Director R ( ) / / 3/45 4/45 R? 1. : / 2. : ggplot2 / Web 3. : slidify 4. : 5. Matlab / Python -> R Interactive Plots. 5/45
산선생의 집입니다. 환영해요
Biped Walking Robot Biped Walking Robot Simulation Program Down(Visual Studio 6.0 ) ). Version.,. Biped Walking Robot - Project Degree of Freedom : 12(,,, 12) :,, : Link. Kinematics. 1. Z (~ Diablo Set
Slide 1
Clock Jitter Effect for Testing Data Converters Jin-Soo Ko Teradyne 2007. 6. 29. 1 Contents Noise Sources of Testing Converter Calculation of SNR with Clock Jitter Minimum Clock Jitter for Testing N bit
PowerPoint 프레젠테이션
I. 문서표준 1. 문서일반 (HY중고딕 11pt) 1-1. 파일명명체계 1-2. 문서등록정보 2. 표지표준 3. 개정이력표준 4. 목차표준 4-1. 목차슬라이드구성 4-2. 간지슬라이드구성 5. 일반표준 5-1. 번호매기기구성 5-2. 텍스트박스구성 5-3. 테이블구성 5-4. 칼라테이블구성 6. 적용예제 Machine Learning Credit Scoring
Software Requirrment Analysis를 위한 정보 검색 기술의 응용
EPG 정보 검색을 위한 예제 기반 자연어 대화 시스템 김석환 * 이청재 정상근 이근배 포항공과대학교 컴퓨터공학과 지능소프트웨어연구실 {megaup, lcj80, hugman, gblee}@postech.ac.kr An Example-Based Natural Language System for EPG Information Access Seokhwan Kim
Manufacturing6
σ6 Six Sigma, it makes Better & Competitive - - 200138 : KOREA SiGMA MANAGEMENT C G Page 2 Function Method Measurement ( / Input Input : Man / Machine Man Machine Machine Man / Measurement Man Measurement
사회통계포럼
wcjang@snu.ac.kr Acknowledgements Dr. Roger Peng Coursera course. https://github.com/rdpeng/courses Creative Commons by Attribution /. 10 : SNS (twitter, facebook), (functional data) : (, ),, /Data Science
untitled
Logic and Computer Design Fundamentals Chapter 4 Combinational Functions and Circuits Functions of a single variable Can be used on inputs to functional blocks to implement other than block s intended
OR MS와 응용-03장
o R M s graphical solution algebraic method ellipsoid algorithm Karmarkar 97 George B Dantzig 979 Khachian Karmarkar 98 Karmarkar interior-point algorithm o R 08 gallon 000 000 00 60 g 0g X : : X : : Ms
4 CD Construct Special Model VI 2 nd Order Model VI 2 Note: Hands-on 1, 2 RC 1 RLC mass-spring-damper 2 2 ζ ω n (rad/sec) 2 ( ζ < 1), 1 (ζ = 1), ( ) 1
: LabVIEW Control Design, Simulation, & System Identification LabVIEW Control Design Toolkit, Simulation Module, System Identification Toolkit 2 (RLC Spring-Mass-Damper) Control Design toolkit LabVIEW
2 : (Seungsoo Lee et al.: Generating a Reflectance Image from a Low-Light Image Using Convolutional Neural Network) (Regular Paper) 24 4, (JBE
2: (Seungsoo Lee et al.: Generating a Reflectance Image from a Low-Light Image Using Convolutional Neural Network) (Regular Paper) 24 4, 2019 7 (JBE Vol. 24, No. 4, July 2019) https://doi.org/10.5909/jbe.2019.24.4.623
°í¼®ÁÖ Ãâ·Â
Performance Optimization of SCTP in Wireless Internet Environments The existing works on Stream Control Transmission Protocol (SCTP) was focused on the fixed network environment. However, the number of
Page 2 of 5 아니다 means to not be, and is therefore the opposite of 이다. While English simply turns words like to be or to exist negative by adding not,
Page 1 of 5 Learn Korean Ep. 4: To be and To exist Of course to be and to exist are different verbs, but they re often confused by beginning students when learning Korean. In English we sometimes use the
Intra_DW_Ch4.PDF
The Intranet Data Warehouse Richard Tanler Ch4 : Online Analytic Processing: From Data To Information 2000. 4. 14 All rights reserved OLAP OLAP OLAP OLAP OLAP OLAP is a label, rather than a technology
PowerPoint 프레젠테이션
Verilog: Finite State Machines CSED311 Lab03 Joonsung Kim, joonsung90@postech.ac.kr Finite State Machines Digital system design 시간에배운것과같습니다. Moore / Mealy machines Verilog 를이용해서어떻게구현할까? 2 Finite State
Gray level 변환 및 Arithmetic 연산을 사용한 영상 개선
Point Operation Histogram Modification 김성영교수 금오공과대학교 컴퓨터공학과 학습내용 HISTOGRAM HISTOGRAM MODIFICATION DETERMINING THRESHOLD IN THRESHOLDING 2 HISTOGRAM A simple datum that gives the number of pixels that a
3 Gas Champion : MBB : IBM BCS PO : 2 BBc : : /45
3 Gas Champion : MBB : IBM BCS PO : 2 BBc : : 20049 0/45 Define ~ Analyze Define VOB KBI R 250 O 2 2.2% CBR Gas Dome 1290 CTQ KCI VOC Measure Process Data USL Target LSL Mean Sample N StDev (Within) StDev
Page 2 of 6 Here are the rules for conjugating Whether (or not) and If when using a Descriptive Verb. The only difference here from Action Verbs is wh
Page 1 of 6 Learn Korean Ep. 13: Whether (or not) and If Let s go over how to say Whether and If. An example in English would be I don t know whether he ll be there, or I don t know if he ll be there.
Structural SVMs 및 Pegasos 알고리즘을 이용한 한국어 개체명 인식
Deep Learning 차례 현재딥러닝기술수준소개 딥러닝 딥러닝기반의자연어처리 Object Recognition https://www.youtube.com/watch?v=n5up_lp9smm Semantic Segmentation https://youtu.be/zjmtdrbqh40 Semantic Segmentation VGGNet + Deconvolution
1-1-basic-43p
A Basic Introduction to Artificial Neural Network (ANN) 도대체인공신경망이란무엇인가? INDEX. Introduction to Artificial neural networks 2. Perceptron 3. Backpropagation Neural Network 4. Hopfield memory 5. Self Organizing
Microsoft PowerPoint - 27.pptx
이산수학 () n-항관계 (n-ary Relations) 2011년봄학기 강원대학교컴퓨터과학전공문양세 n-ary Relations (n-항관계 ) An n-ary relation R on sets A 1,,A n, written R:A 1,,A n, is a subset R A 1 A n. (A 1,,A n 에대한 n- 항관계 R 은 A 1 A n 의부분집합이다.)
정보기술응용학회 발표
, hsh@bhknuackr, trademark21@koreacom 1370, +82-53-950-5440 - 476 - :,, VOC,, CBML - Abstract -,, VOC VOC VOC - 477 - - 478 - Cost- Center [2] VOC VOC, ( ) VOC - 479 - IT [7] Knowledge / Information Management
19_9_767.hwp
(Regular Paper) 19 6, 2014 11 (JBE Vol. 19, No. 6, November 2014) http://dx.doi.org/10.5909/jbe.2014.19.6.866 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) RGB-Depth - a), a), b), a) Real-Virtual Fusion
Oracle Apps Day_SEM
Senior Consultant Application Sales Consulting Oracle Korea - 1. S = (P + R) x E S= P= R= E= Source : Strategy Execution, By Daniel M. Beall 2001 1. Strategy Formulation Sound Flawed Missed Opportunity
슬라이드 1
Pairwise Tool & Pairwise Test NuSRS 200511305 김성규 200511306 김성훈 200614164 김효석 200611124 유성배 200518036 곡진화 2 PICT Pairwise Tool - PICT Microsoft 의 Command-line 기반의 Free Software www.pairwise.org 에서다운로드후설치
첨 부 1. 설문분석 결과 2. 교육과정 프로파일 169
첨부 168 첨 부 1. 설문분석 결과 2. 교육과정 프로파일 169 Ⅰ-1. 설문조사 개요 Ⅰ. 설문분석 결과 병무청 직원들이 생각하는 조직문화, 교육에 대한 인식, 역량 중요도/수행도 조사를 인터넷을 통해 실 시 총 1297명의 응답을 받았음 (95% 신뢰수준에 표본오차는 ±5%). 조사 방법 인터넷 조사 조사 기간 2005년 5월 4일 (목) ~ 5월
<B3EDB9AEC1FD5F3235C1FD2E687770>
경상북도 자연태음악의 소박집합, 장단유형, 전단후장 경상북도 자연태음악의 소박집합, 장단유형, 전단후장 - 전통 동요 및 부녀요를 중심으로 - 이 보 형 1) * 한국의 자연태 음악 특성 가운데 보편적인 특성은 대충 밝혀졌지만 소박집합에 의한 장단주기 박자유형, 장단유형, 같은 층위 전후 구성성분의 시가( 時 價 )형태 등 은 밝혀지지 않았으므로
Delving Deeper into Convolutional Networks for Learning Video Representations - Nicolas Ballas, Li Yao, Chris Pal, Aaron Courville arXiv:
Delving Deeper into Convolutional Networks for Learning Video Representations Nicolas Ballas, Li Yao, Chris Pal, Aaron Courville arxiv: 1511.06432 Il Gu Yi DeepLAB in Modu Labs. June 13, 2016 Il Gu Yi
methods.hwp
1. 교과목 개요 심리학 연구에 기저하는 기본 원리들을 이해하고, 다양한 심리학 연구설계(실험 및 비실험 설계)를 학습하여, 독립된 연구자로서의 기본적인 연구 설계 및 통계 분석능력을 함양한다. 2. 강의 목표 심리학 연구자로서 갖추어야 할 기본적인 지식들을 익힘을 목적으로 한다. 3. 강의 방법 강의, 토론, 조별 발표 4. 평가방법 중간고사 35%, 기말고사
생들의 역할을 중심으로 요약 될 수 있으며 구체적인 내용은 다음과 같다. 첫째. 교육의 대상 면에서 학습대상이 확대되고 있다. 정보의 양이 폭발적으로 증가하고 사회체제의 변화가 가속화 되면서 학습의 대상은 학생뿐만 아니라 성인 모두에게 확대되고 있으며 평생학습의 시대가
Ⅰ. 사회패러다임과 교육패러다임의 변화 1. 사회패러다임변화 교육환경의 변화를 이해하기 위해서는 우선 21세기 사회패러다임의 변화에 대한 이해가 필요하다. 요즈음 우리사회에 자주 사용되는 말 가운데 하나가 패러다임 을 전환해야 한다., 21세기를 지향하는 새로운 패러다임을 갖추어야 한다. 는 등 등 패러다임이라는 말을 많이 사용하고 있다. 패러다임이란 말은
04_오픈지엘API.key
4. API. API. API..,.. 1 ,, ISO/IEC JTC1/SC24, Working Group ISO " (Architecture) " (API, Application Program Interface) " (Metafile and Interface) " (Language Binding) " (Validation Testing and Registration)"
Microsoft PowerPoint - AC3.pptx
Chapter 3 Block Diagrams and Signal Flow Graphs Automatic Control Systems, 9th Edition Farid Golnaraghi, Simon Fraser University Benjamin C. Kuo, University of Illinois 1 Introduction In this chapter,
Disclaimer IPO Presentation,. Presentation...,,,,, E.,,., Presentation,., Representative...
DEXTER STUDIOS INVESTOR RELATIONS 2015 Disclaimer IPO Presentation,. Presentation...,,,,, E.,,., Presentation,., Representative... Contents Prologue 01 VFX 02 China 03 Investment Highlights 04 Growth Engine
Journal of Educational Innovation Research 2018, Vol. 28, No. 1, pp DOI: A study on Characte
Journal of Educational Innovation Research 2018, Vol. 28, No. 1, pp.381-404 DOI: http://dx.doi.org/10.21024/pnuedi.28.1.201803.381 A study on Characteristics of Action Learning by Analyzing Learners Experiences
아트앤플레이군 (2년제) Art & Play Faculty 95 교육목표 95 군 공통(네트워크) 교과과정표 96 드로잉과 페인팅 Drawing & Painting Major Track 97 매체예술 Media Art Major Track 98 비디오 & 사운드 Video & Sound Major Track 99 사진예술 PHOTOGRAPHIC ART Major
#Ȳ¿ë¼®
http://www.kbc.go.kr/ A B yk u δ = 2u k 1 = yk u = 0. 659 2nu k = 1 k k 1 n yk k Abstract Web Repertoire and Concentration Rate : Analysing Web Traffic Data Yong - Suk Hwang (Research
UML
Introduction to UML Team. 5 2014/03/14 원스타 200611494 김성원 200810047 허태경 200811466 - Index - 1. UML이란? - 3 2. UML Diagram - 4 3. UML 표기법 - 17 4. GRAPPLE에 따른 UML 작성 과정 - 21 5. UML Tool Star UML - 32 6. 참조문헌
인켈(국문)pdf.pdf
M F - 2 5 0 Portable Digital Music Player FM PRESET STEREOMONO FM FM FM FM EQ PC Install Disc MP3/FM Program U S B P C Firmware Upgrade General Repeat Mode FM Band Sleep Time Power Off Time Resume Load
AT_GraduateProgram.key
Art & Technology Graduate Program M.A.S (Master of Arts & Science) in Art & Technology Why Art Tech Graduate Program? / + + X Why Sogang? - Art/Design + Technology 4 Art & Technology Who is this for? (
FMX M JPG 15MB 320x240 30fps, 160Kbps 11MB View operation,, seek seek Random Access Average Read Sequential Read 12 FMX () 2
FMX FMX 20062 () wwwexellencom sales@exellencom () 1 FMX 1 11 5M JPG 15MB 320x240 30fps, 160Kbps 11MB View operation,, seek seek Random Access Average Read Sequential Read 12 FMX () 2 FMX FMX D E (one
Journal of Educational Innovation Research 2018, Vol. 28, No. 3, pp DOI: NCS : * A Study on
Journal of Educational Innovation Research 2018, Vol. 28, No. 3, pp.157-176 DOI: http://dx.doi.org/10.21024/pnuedi.28.3.201809.157 NCS : * A Study on the NCS Learning Module Problem Analysis and Effective
DBPIA-NURIMEDIA
The e-business Studies Volume 17, Number 4, August, 30, 2016:319~332 Received: 2016/07/28, Accepted: 2016/08/28 Revised: 2016/08/27, Published: 2016/08/30 [ABSTRACT] This paper examined what determina
Microsoft PowerPoint - 실습소개와 AI_ML_DL_배포용.pptx
실습강의개요와인공지능, 기계학습, 신경망 < 인공지능입문 > 강의 허민오 Biointelligence Laboratory School of Computer Science and Engineering Seoul National University 실습강의개요 노트북을꼭지참해야하는강좌 신경망소개 (2 주, 허민오 ) Python ( 프로그래밍언어 ) (2주, 김준호
본문01
Ⅱ 논술 지도의 방법과 실제 2. 읽기에서 논술까지 의 개발 배경 읽기에서 논술까지 자료집 개발의 본래 목적은 초 중 고교 학교 평가에서 서술형 평가 비중이 2005 학년도 30%, 2006학년도 40%, 2007학년도 50%로 확대 되고, 2008학년도부터 대학 입시에서 논술 비중이 커지면서 논술 교육은 학교가 책임진다. 는 풍토 조성으로 공교육의 신뢰성과
SW¹é¼Ł-³¯°³Æ÷ÇÔÇ¥Áö2013
SOFTWARE ENGINEERING WHITE BOOK : KOREA 2013 SOFTWARE ENGINEERING WHITE BOOK : KOREA 2013 SOFTWARE ENGINEERING WHITE BOOK : KOREA 2013 SOFTWARE ENGINEERING WHITE BOOK : KOREA 2013 SOFTWARE ENGINEERING
03.Agile.key
CSE4006 Software Engineering Agile Development Scott Uk-Jin Lee Division of Computer Science, College of Computing Hanyang University ERICA Campus 1 st Semester 2018 Background of Agile SW Development
_KrlGF발표자료_AI
AI 의과거와현재그리고내일 AI is the New Electricity 2017.09.15 AI! 2 Near Future of Super Intelligence? *source l http://www.motherjones.com/media/2013/05/robots-artificial-intelligence-jobs-automation 3 4 I think
09권오설_ok.hwp
(JBE Vol. 19, No. 5, September 2014) (Regular Paper) 19 5, 2014 9 (JBE Vol. 19, No. 5, September 2014) http://dx.doi.org/10.5909/jbe.2014.19.5.656 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) a) Reduction
DBPIA-NURIMEDIA
FPS게임 구성요소의 중요도 분석방법에 관한 연구 2 계층화 의사결정법에 의한 요소별 상관관계측정과 대안의 선정 The Study on the Priority of First Person Shooter game Elements using Analytic Hierarchy Process 주 저 자 : 배혜진 에이디 테크놀로지 대표 Bae, Hyejin AD Technology
/ TV 80 () DAB 2001 2002 2003 2004 2005 2010 Analog/Digital CATV Services EPG TV ( 60 ) TV ( Basic, Tier, Premiums 60 ) VOD Services Movies In Demand ( 20 ) Education N- VOD (24 ) Digital Music
LCD Display
LCD Display SyncMaster 460DRn, 460DR VCR DVD DTV HDMI DVI to HDMI LAN USB (MDC: Multiple Display Control) PC. PC RS-232C. PC (Serial port) (Serial port) RS-232C.. > > Multiple Display
예제 1.1 ( 관계연산자 ) >> A=1:9, B=9-A A = B = >> tf = A>4 % 4 보다큰 A 의원소들을찾을경우 tf = >> tf = (A==B) % A
예제 1.1 ( 관계연산자 ) >> A=1:9, B=9-A A = 1 2 3 4 5 6 7 8 9 B = 8 7 6 5 4 3 2 1 0 >> tf = A>4 % 4 보다큰 A 의원소들을찾을경우 tf = 0 0 0 0 1 1 1 1 1 >> tf = (A==B) % A 의원소와 B 의원소가똑같은경우를찾을때 tf = 0 0 0 0 0 0 0 0 0 >> tf
강의록
Analytic CRM 2006. 5. 11 tsshin@yonsei.ac.kr Analytic CRM Analytic CRM Data Mining Analytical CRM in CRM Ecosystem Operational CRM Business Operations Mgmt. Analytical CRM Business Performance Mgmt. Back
강의10
Computer Programming gdb and awk 12 th Lecture 김현철컴퓨터공학부서울대학교 순서 C Compiler and Linker 보충 Static vs Shared Libraries ( 계속 ) gdb awk Q&A Shared vs Static Libraries ( 계속 ) Advantage of Using Libraries Reduced
2005CG01.PDF
Computer Graphics # 1 Contents CG Design CG Programming 2005-03-10 Computer Graphics 2 CG science, engineering, medicine, business, industry, government, art, entertainment, advertising, education and
APOGEE Insight_KR_Base_3P11
Technical Specification Sheet Document No. 149-332P25 September, 2010 Insight 3.11 Base Workstation 그림 1. Insight Base 메인메뉴 Insight Base Insight Insight Base, Insight Base Insight Base Insight Windows
PowerPoint 프레젠테이션
2003 CRM (Table of Contents). CRM. 2003. 2003 CRM. CRM . CRM CRM,,, Modeling Revenue Legacy System C. V. C. C V.. = V Calling Behavior. Behavior al Value Profitability Customer Value Function Churn scoring
(JBE Vol. 24, No. 1, January 2019) (Special Paper) 24 1, (JBE Vol. 24, No. 1, January 2019) ISSN 2287-
(Special Paper) 24 1 2019 1 (JBE Vol. 24 No. 1 January 2019) https//doi.org/10.5909/jbe.2019.24.1.58 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) a) a) a) b) c) d) A Study on Named Entity Recognition
PowerPoint 프레젠테이션
CRM Fair 2004 Spring Copyright 2004 DaumSoft All rights reserved. INDEX Copyright 2004 DaumSoft All rights reserved. Copyright 2004 DaumSoft All rights reserved. Copyright 2004 DaumSoft All rights reserved.
기획 1 서울공대생에게 물었다 글 재료공학부 1, 이윤구 재료공학부 1, 김유리 전기정보공학부 1, 전세환 편집 재료공학부 3, 오수봉 이번 서울공대생에게 물었다! 코너는 특별히 설문조사 형식으로 진행해 보려고 해 요. 설문조사에는 서울대학교 공대 재학생 121명, 비
2015 autumn 공대상상 예비 서울공대생을 위한 서울대 공대 이야기 Vol. 13 Contents 02 기획 서울공대생에게 물었다 극한직업 공캠 촬영 편 Fashion in SNU - 단체복 편 서울대 식당, 어디까지 먹어 봤니? 12 기획 연재 기계항공공학부 기계항공공학부를 소개합니다 STEP 01 기계항공공학부에 대한 궁금증 STEP 02 동문 인터뷰
04김호걸(39~50)ok
Journal of Environmental Impact Assessment, Vol. 22, No. 1(2013) pp.39~50 Prediction of Landslides Occurrence Probability under Climate Change using MaxEnt Model Kim, Hogul* Lee, Dong-Kun** Mo, Yongwon*
., (, 2000;, 1993;,,, 1994), () 65, 4 51, (,, ). 33, 4 30, 23 3 (, ) () () 25, (),,,, (,,, 2015b). 1 5,
* 4.,, 3,,, 3,, -., 3, 12, 27, 20. 9,,,,,,,,. 6,,,,,. 5,,,,.. * (2016),. (Corresponding Author): / / 303 Tel: 063-225-4496 / E-mail: jnj1015@jj.ac.kr ., (, 2000;, 1993;,,, 1994), 2000. 2015 () 65, 4 51,
public key private key Encryption Algorithm Decryption Algorithm 1
public key private key Encryption Algorithm Decryption Algorithm 1 One-Way Function ( ) A function which is easy to compute in one direction, but difficult to invert - given x, y = f(x) is easy - given
소프트웨어개발방법론
사용사례 (Use Case) Objectives 2 소개? (story) vs. 3 UC 와 UP 산출물과의관계 Sample UP Artifact Relationships Domain Model Business Modeling date... Sale 1 1..* Sales... LineItem... quantity Use-Case Model objects,
Buy one get one with discount promotional strategy
Buy one get one with discount Promotional Strategy Kyong-Kuk Kim, Chi-Ghun Lee and Sunggyun Park ISysE Department, FEG 002079 Contents Introduction Literature Review Model Solution Further research 2 ISysE
` Companies need to play various roles as the network of supply chain gradually expands. Companies are required to form a supply chain with outsourcing or partnerships since a company can not
PowerPoint Presentation
컴퓨터비전 및 패턴인식 연구회 2009.2.12 Support Vector Machines http://cespc1.kumoh.ac.kr/~nonezero/svm ws cvpr.pdf 금오공과대학교 컴퓨터공학부 고재필 1 Contents Introduction Optimal Hyperplane Soft-Margin SVM Nonlinear SVM with Kernel
<B9CCB5F0BEEEB0E6C1A6BFCDB9AEC8AD5F31322D32C8A35FBABBB9AE5FC3CAC6C731BCE25F6F6B5F32303134303531362E687770>
미디어 경제와 문화 2014년 제12권 2호, 7 43 www.jomec.com TV광고 시청률 예측방법 비교연구 프로그램의 장르 구분에 따른 차이를 중심으로 1)2) 이인성* 단국대학교 커뮤니케이션학과 박사과정 박현수** 단국대학교 커뮤니케이션학부 교수 본 연구는 TV프로그램의 장르에 따라 광고시청률 예측모형들의 정확도를 비교하고 자 하였다. 본 연구에서
,. 3D 2D 3D. 3D. 3D.. 3D 90. Ross. Ross [1]. T. Okino MTD(modified time difference) [2], Y. Matsumoto (motion parallax) [3]. [4], [5,6,7,8] D/3
![,. 3D 2D 3D. 3D. 3D.. 3D 90. Ross. Ross [1]. T. Okino MTD(modified time difference) [2], Y. Matsumoto (motion parallax) [3]. [4], [5,6,7,8] D/3](/thumbs/67/57222518.jpg ",. 3D 2D 3D. 3D. 3D.. 3D 90. Ross. Ross [1]. T. Okino MTD(modified time difference) [2], Y. Matsumoto (motion parallax) [3]. [4], [5,6,7,8] D/3")
Depth layer partition 2D 3D a), a) 3D conversion of 2D video using depth layer partition Sudong Kim a) and Jisang Yoo a) depth layer partition 2D 3D. 2D (depth map). (edge directional histogram). depth
15_3oracle
Principal Consultant Corporate Management Team ( Oracle HRMS ) Agenda 1. Oracle Overview 2. HR Transformation 3. Oracle HRMS Initiatives 4. Oracle HRMS Model 5. Oracle HRMS System 6. Business Benefit 7.
서론 34 2
34 2 Journal of the Korean Society of Health Information and Health Statistics Volume 34, Number 2, 2009, pp. 165 176 165 진은희 A Study on Health related Action Rates of Dietary Guidelines and Pattern of
2002년 2학기 자료구조
자료구조 (Data Structures) Chapter 1 Basic Concepts Overview : Data (1) Data vs Information (2) Data Linear list( 선형리스트 ) - Sequential list : - Linked list : Nonlinear list( 비선형리스트 ) - Tree : - Graph : (3)
Effects of baseball expertise and stimulus speeds on coincidence-anticipation timing accuracy of batting Jong-Hwa Lee, Seok-Jin Kim, & Seon-Jin Kim* Seoul National University [Purpose] [Methods] [Results]
SchoolNet튜토리얼.PDF
Interoperability :,, Reusability: : Manageability : Accessibility :, LMS Durability : (Specifications), AICC (Aviation Industry CBT Committee) : 1988, /, LMS IMS : 1997EduCom NLII,,,,, ARIADNE (Alliance
ETL_project_best_practice1.ppt
ETL ETL Data,., Data Warehouse DataData Warehouse ETL tool/system: ETL, ETL Process Data Warehouse Platform Database, Access Method Data Source Data Operational Data Near Real-Time Data Modeling Refresh/Replication
PowerPoint 프레젠테이션
Visual Search At SK-Planet sk-planet Machine Intelligence Lab. 나상일 1. 개발배경 2. 첫접근방법 3. 개선된방법 A. Visual recognition technology B. Guided search C. Retrieval system 개발배경 개발배경 상품검색을좀더쉽게 Key-word 트렌치코트버튺벨트
4 : (Hyo-Jin Cho et al.: Audio High-Band Coding based on Autoencoder with Side Information) (Special Paper) 24 3, (JBE Vol. 24, No. 3, May 2019
4 : (Hyo-Jin Cho et al.: Audio High-Band Coding based on Autoencoder with Side Information) (Special Paper) 24 3, 2019 5 (JBE Vol. 24, No. 3, May 2019) https://doi.org/10.5909/jbe.2019.24.3.387 ISSN 2287-9137
PowerPoint 프레젠테이션
Reasons for Poor Performance Programs 60% Design 20% System 2.5% Database 17.5% Source: ORACLE Performance Tuning 1 SMS TOOL DBA Monitoring TOOL Administration TOOL Performance Insight Backup SQL TUNING