GoConqr has a new look! Let us know what you think.

Sign up free

Flashcards

DM

Abolanle Sule English 105 cards

Cards in this deck (105)

  1. 1We are in the ERA of AI, name some sample AI apps?- User Interface
    - Customized AD
    - Insect Bug
    - Mind Reader
    - 3D Hologram
    - Retina Scan
  2. 2Name the 4 industrial revolutions?1st - Mechanical Production, 1750
    2nd - Mass Production, 1850
    3rd - Automated Production, 1950
    4th - Artificial Intelligence & Big Data, 2000.
  3. 3What is the name of the female writer of Technological Revolution & Financial Capital?Carolota Perez
  4. 4What are the generations of computers?- Vacuum Tubes
    - Transistors
    - Integrated Circuits IC
    - VLSI Very Large Scale Integrated Circuits
  5. 5What is AI ?Its the science & engineering making intelligent computer machines.
  6. 6What is machine learning ?Machine learning is a type of artificial intelligence (AI) that allows software applications to become more accurate in predicting outcomes without being explicitly programmed.
  7. 7Mention examples of machine learning ?Robotics Control
    Speech Recognition &
    Computer Vision
    Medical Analysis
    Computational Biology
  8. 8Association deals with ---Support and Confidence
  9. 9SL Classification 1 has following applicationFraud detection
    Medical Analysis
    Credit Approval
  10. 10Classification 2 uses ----Classifier/ Model
  11. 11Classification 3 ---Model for unseen data.
  12. 12Classification 4 we have ----Decision Tree
    Bayesian Classification
    Rule-Based Classification
    SVM
    Neural Network
  13. 13A Cluster is a Collection of data objects that are -------- Similar to one another within the same cluster.
    - Dissimilar to the objects in other clusters.
  14. 14Finding similarities between data according to the characteristics found in the data and grouping similar data objects into clusters -------Cluster Analysis.
  15. 15Unsupervised learning has ---------no predefined classes.
  16. 16Cluster Analysis can be applied as ----- or --------As a stand-alone tool to get insight into data distribution.
    As a pre-processing step for other algorithms.
  17. 17Examples of Clustering 2Marketing
    Land Use
    Insurance
    City-Planning
    Earth-Quake studies
  18. 18Examples of Clustering 3K- means
  19. 19Examples of Clustering 4, Use distance matrix as clustering criteria. This method does not require the number of clusters kas an input, but needs a termination conditionHierarchical Clustering
  20. 20Why Deep Learning ?Biological Plausibility
    Hastad Proof
  21. 21IBM’s Artificial Intelligence computer system • Capable of answering questions in natural language • Competed against champions on Jeopardy and wonIBM Watson
  22. 22AI Applications?ID Cards
    Bio-metric Identification
    Prediction Problem-stock market
    Language Problem-translation
  23. 23What is Data Mining?-Automated analysis of massive data-sets.
    - Knowledge discovery from data.
  24. 24We are in the age of _____ data?Zetta Byte
  25. 25Data volume is growing -------?Exponentially
  26. 26What are the step in KDD?Data cleaning
    Data integration from multiple sources
    Warehousing the data
    Data cube construction
    Data selection for data mining
    Data Mining
    Presentation of the Mining results.
    Patterns and knowledge to be used or stored into knowledge-base
  27. 27Data is a collection of ------Information.
  28. 28Data ----- makes it easier to understand data.Visualization
  29. 29Name 5 types of datasets ------Record
    Graph & Network
    Ordered
    Spatial
    Image
    Multimedia.
  30. 30An attribute is ----a data field rep. a Xteristics/Feature of a Data object.
  31. 31The Normal Distribution with mean of zero is ---- or --- distribution graph.Gaussian or Symmetric
  32. 32Name 2 types of skewed graphs?Positively Skewed
    Negatively Skewed
  33. 33Scatter Plot measures the -------relationship btw 2 attributes.
  34. 34p(A,B) = p(A|B) . p(B)Product Rule
    (event A&B can co-exist)
  35. 35p(A|B) = p(A) or p(A,B) = p(A) . p(B)Independence Rule
  36. 36A vector is a --------point in space/
    segmented magnitude.
  37. 37Epoch is the -----# of iterations during a training.
  38. 38Why do we need to Pre-process data?Better Accuracy
    Consistency
    To suit the algorithm
  39. 39Mention Data Pre-processing Tasks?Data Cleaning
    Data Reduction
    Integration
    Transformation/Discretization.
  40. 40--- helps to analysis 2 attributes that are strongly implies to one another?Correlation Analysis.
  41. 41Chi-Square is measure as ---------(Observed - Expected)^2 / Expected.
  42. 42Co-variance (A,B) is measured as ------E ( (A-A) . (B-B) ) / n - 1.
    where n is d sample size.
  43. 43PCA - Principal Component Analysis is interested in the --- eigen value?biggest
  44. 44----- Algorithm measures the frequency of item-sets?Apriori Algorithm.
  45. 45The ff. process model construction requires/not a labeled data? Classification - SL - UL -needs label
    needs label
    label is unknown.
  46. 46Mention d 3 stages of CNN?Convolution Stage
    Non- Linearity
    Pooling.
  47. 47We re-use the weight and sequences ?RNN
  48. 48We use different sequences ?CNN
  49. 492 types of Evaluation method ?Hold-Out
    Cross-Validation.
  50. 50Evaluates Classifier's Accuracy ? small dataset sampling wt error margin.Bootstrap
  51. 51One of the benefits of Decision Trees is ----makes classification easier.
  52. 52p(C|X) = P(X|C). P(C) / P(X)Naives Bayes Classifier
  53. 53Naives Bayes assumes that -----data are Independent.
  54. 54Mention ways by which we can optimize?Gradient Descend.
    EMIL
  55. 55Gradient Descend is an iterative algorithm, applied to ----Linear functions.
  56. 56Sigma is between ----- and ----- and its -----0 and 1 and its continuous.
  57. 57The Goal of Linear Regression is to ----------- thetha is a vector of weights to adjust such dt theta is minimized. Minimization is partial derivates of theta wrto theta.. During each step of the GD loop, d values of theta comes close to d values dt will minimize d cost fnc.minimize the Cost Function.
  58. 58The hypothesis is given asy= hx(w) = xw + xw + ... + xwn
  59. 59-------- ignores some of d nodes and focus on the important ones?Drop-Out.
  60. 60------ makes it meaningful?Activation Function.
  61. 61SVM meansSupport Vector Machines.
  62. 62SVM is a classification method based on finding a --------- fnc which maximizes d dist. btw 2 points.discriminative function
  63. 63What are the solutions for Over-fitting? - Avoiding Over-fitting?Regularization-kind of smoothing.
    More training data.
    Reduce the # feature.
    - Pre-prunning & Post-prunning.
  64. 64----- is the process of running two or more related but different analytical models and then synthesizing the results into a single score or spread in order to improve the accuracy of d predictive analytics and data mining applications.Ensemble.
    (+3 ~ 5% better accuracy).
  65. 653 Axioms of Probability:Non Negativity.
    Sum of all possible outcome equals 1.
    Addivity of Disjoint Events.
  66. 66Name 2 types of Probability Space ?1. Discrete Space - Finite (Analysis involves summation).
    2. Continuous Space - Infinite (Analysis involves integral).
  67. 67-------- is the covariance normalized by d standard deviations of 2 variables. Always lies btw 1 and -1.Pearson's Correlation Co.efficient.
  68. 68---- is a way to find conditional prob. for one variable when conditional prob. for another variable is known?Bayes Rule.
  69. 69------ measures tendency of 2 variables to deviate from their means in same/opposite directions at d same time.Covariance.
  70. 70A Data Object represents an -----entity.
  71. 71Data Objects are described by -------attributes.
  72. 72Attributes reps........Xteristics/Features.
  73. 734 Types of AttributesNominal
    Binary
    Numeric
    Ordinal
  74. 74Numeric has ------Interval Scaled -no true 0 pt
    Ratio Scaled - inherent 0 pt.
  75. 75Ordinal has ----meaningful order/ranking but unknown magnitude.
  76. 763 Types of Mode ----Unimodal
    Bimodal
    Trimodal
  77. 77Emperical Formular is ---mean - mode OR
    3 * (mean -median)
  78. 78Q1 is25th Percentile.
  79. 79Q3 is75th Percentile.
  80. 80IQR isQ3 -Q1
  81. 81Five number summary isMin.
    Q1
    Median
    Q3
    Max.
  82. 82Outliner is usually a value higher/lower than ----1.5 * IQR
  83. 83Properties of Normal Distribution ----68% - meu
    95% - 2meu
    99.7% - 3meu
  84. 84Numerical Measure of how 2 data objects are alike ?Similarity
  85. 85Numerical Measure of how 2 data objects are different ?Dissimilarity
  86. 86---- refers to Similarity or Dissimilarity ?Proximity.
  87. 87------ is used to reduce dimensional of data set to 2D or 3D.PCA
  88. 88Find pattern in high-dimensional dataPCA
  89. 89Visualize data of high-dimensionalityPCA
  90. 90PCA applications ------Face Recognition
    Image Compression
    Gene Expression Analysis.
  91. 91Vectors having same direction are called --Eigen Vectors.
  92. 92Its coefficient is calledEigen Value
  93. 93PC1largest absolute value/
    largest variance.
  94. 94PC2direction wt max. variance, orthogonal to PC1.
  95. 95How to Handle Noisy Data ?- Binning
    - Regression.
  96. 963 steps of Binning ---Sort d data.
    Partition into equal freq. bins.
    Smooth bins by mean/median/bin boundaries.
  97. 97Regression we haveLinear Regression
    Multiple Linear Regression.
  98. 98Data Integration combines -----data from multiple sources into coherent store.
  99. 99Obtaining a reduced rep. of data set much smaller in volume but yield same analytic results?Data Reduction.
  100. 100How do you avoid the Zero-Probability Problem?Using Laplacian Correction/Estimator.....
    Data set of 1000 Tuples, we have:
    P(low) =0 +1 = 1/1003
    P(medium)=990+1= 991/1003
    P(high)=10+1 = 11/1003
  101. 101Entropy - Information Theory...A measure of the degree of uncertainty associated a random variable.
    Higher Entropy - Higher Uncertainty!
    Lower Entropy - Lower Uncertainty!
  102. 102Name 4 Activation Functions?Step Fnc - produces binary results 0&1.
    Sigmoid Fnc -produces values btw 0&1.
    Tan h Fnc - produces values btw -1&1.
    ReLU fnc -depends on size of z&x.
  103. 103Features of ReLU?Converges 6x faster than Sigmoid & Tan h.
    Does not saturate.
    Computationally very efficient.
  104. 104how do you solve the annoying problem of Dead Neurons?Using Leaky ReLU.
  105. 105—

Information

Description
Flashcards on DM, created by Abolanle Sule on 28/09/2017.
Created by
Abolanle Sule
Views
1
Rating

Feedback

What is it about?

What do you think of the new look?

How do you like it?