Machine Learning as a Daily Work for a Programmer- Volodymyr Vorobiov
•
2 likes•798 views
Ruby Meditation #15
Recommended
Recommended
More Related Content
Similar to Machine Learning as a Daily Work for a Programmer- Volodymyr Vorobiov
Similar to Machine Learning as a Daily Work for a Programmer- Volodymyr Vorobiov (20)
More from Ruby Meditation
More from Ruby Meditation (20)
Recently uploaded
Recently uploaded (20)
Machine Learning as a Daily Work for a Programmer- Volodymyr Vorobiov
- 1. RUBYGARAGE2017 TECHNOLOGYMATTERS What It Is And How It Works Machine Learning Volodymyr Vorobiov Software Development Consultant at RubyGarage
- 2. Machine learning is a subset of artificial intelligence whose goalis to give computers the ability to teach themselves, whereas artificial intelligence is a general concept of smart machines. In other words, artificial intelligence is implemented through machine learning or - to be more precise - through machine learning algorithms. RUBYGARAGE2017 TECHNOLOGYMATTERS artificial intelligence TEACH YOUR COMPUTER
- 3. EXAMPLES OF HOW MACHINE LEARNING IS USED IN THE REAL WORLD RUBYGARAGE2017 TECHNOLOGYMATTERS - Facial recognition - Voice recognition - Text recognition - Diagnostics into medicine, - Self-driving cars - Robots behavior adjustment - Ads targeting, - Predictions in financial trading - Virtual and augmented reality - Astronomy and space ???
- 4. The 21st century is the age of data. It’s literally everywhere. In fact, there has been an exponential growth in the volume of data over the past decade; the total amount of data doubles every two years. Most of it, however, isn’t used. Huge volumes of data can be tagged, structured, and analyzed, revealing a lot of valuable information. Only machine learning algorithms can easily cope with this task. RUBYGARAGE2017 TECHNOLOGYMATTERS WHY THE FUTURE BELONGS TO MACHINE LEARNING
- 5. RUBYGARAGE2017 TECHNOLOGYMATTERS HOW MACHINE LEARNING WORKS Preprocessing Learinng Evaluation Prediction Labels Raw Data Labels Labels Final Model New DataTraining Dataset Test Dataset Learning Algorithm (putting data into the necessary shape) (creating a model with the help of training data) (model assessment using test data) application of the model)
- 6. TOOLS RUBYGARAGE2017 TECHNOLOGYMATTERS - Python - Pandas - Powerful data analysis library for Python Pandas is a powerful data analysis Python library that provides flexible and fast data structures for processing “relational” or “labeled” data. This is a fundamental data analysis toolkit in Python. - Scikit-learn - Machine Learning in Python These are simple and effective open-source tools for data mining and analysis. - Statsmodels This is a Python module providing functions and classes to estimate different statistical models as well as to conduct tests and explore statistical data. The Statsmodels module offers a comprehensive list of result statistics. - Matplotlib Matplotlib is a Python 2D plotting library that releases publication quality figures in multiple formats and interactive environments in different platforms.
- 7. The quality of the data and the amount of useful information that it contains are key factors that determine how well a machine learning algorithm can learn. Therefore, it is absolutely critical that we make sure to examine and preprocess a dataset before we feed it to a learning algorithm. - Removing and imputing missing values from the dataset - Getting categorical data into shape for machine learning algorithms - Selecting relevant features for the model construction RUBYGARAGE2017 TECHNOLOGYMATTERS DATA PREPROCESSING
- 8. DATA PREPROCESSING DATASET PRESENTATION RUBYGARAGE2017 TECHNOLOGYMATTERS Independent variables Dependent variables
- 11. RUBYGARAGE2017 TECHNOLOGYMATTERS DEALING WITH MISSING DATA Most computational tools are unable to handle such missing values or would produce unpredictable results if we simply ignored them. Therefore, it is crucial that we take care of those missing values before we proceed with further analyses.
- 12. - Eliminating samples or features with missing values The easiest solution to this problem is simply to remove samples with missing values from a dataset. However, this seemingly handy approach has a number of drawbacks. For example, removing too many of such samples is likely to compromise the quality of the analysis. - Imputing missing values The solution is to use various interpolation techniques that help to “guess” the missing values from other samples in a dataset. RUBYGARAGE2017 TECHNOLOGYMATTERS DEALING WITH MISSING DATA
- 14. IMPUTING MISSING VALUES RESULTS
- 16. ENCODE LABELS
- 19. DUMMY VARIABLES
- 22. PARTITIONING A DATASET INTO TRAINING AND TEST SETS
- 23. TRAINING AND TEST SETS RESULTS
- 24. BRINGING FEATURES ONTO THE SAME SCALE
- 26. TRAINING AND SELECTING A PREDICTIVE MODEL RUBYGARAGE2017 TECHNOLOGYMATTERS - Supervised learning - Regression - Classification - Unsupervised learning - Clustering - Dimensionality Reduction - Reinforcement Learning - Association Rule Learning - Natural Language Processing - Deep Learning - Model Selection
- 27. SUPERVISED LEARNING RUBYGARAGE2017 TECHNOLOGYMATTERS For making predictions about the future Regression For predicting continuous outcomes Classification For predicting class labels
- 28. REGRESSION RUBYGARAGE2017 TECHNOLOGYMATTERS Regression models (both linear and non-linear) are used for predicting a real value, like salary for example. If your independent variable is time, then you are forecasting future values, otherwise your model is predicting present but unknown values.
- 29. SIMPLE LINEAR REGRESSION RUBYGARAGE2017 TECHNOLOGYMATTERS y x Constant Coefficent Dependent variable (DV) Independent variable (IV) y = b + b*x1 10
- 30. DATASET PRESENTATION. EXPERIENCE AND SALARY. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 31. SIMPLE LINEAR REGRESSION TRAINING
- 32. SIMPLE LINEAR REGRESSION TRAINING RUBYGARAGE2017 TECHNOLOGYMATTERS
- 33. MULTIPLE LINEAR REGRESSION RUBYGARAGE2017 TECHNOLOGYMATTERS Constant Coefficent Dependent variable (DV) Independent variable (IVs) y = b + b*x + b*x ... + b*x1 1 2 2 n n0
- 34. DATASET PRESENTATION. INVESTMENT FUND STATISTIC. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 35. MULTIPLE LINEAR REGRESSION TRAINING
- 36. EVALUATING REGRESSION MODELS PERFORMANCE RUBYGARAGE2017 TECHNOLOGYMATTERS 1. All-in 2. Backward Elimination 3. Forward Selection 4. Bidirectional Elimination 5. Score Comparison Stepwise Regression
- 37. BACKWARD ELIMINATION RUBYGARAGE2017 TECHNOLOGYMATTERS STEP 1: Select a significance level to stay in the model (e.g. SL = 0.05) STEP 2: Fit the full model with all possible predictors STEP 3: Consider the predictor with the highest P-value. If P > SL, go to STEP 4, otherwise go to FIN STEP 4: Remove the predictor STEP 5: Fit model without this variable*
- 39. BACKWARD ELIMINATION TRAINING STEP 1
- 40. BACKWARD ELIMINATION TRAINING STEP 4
- 41. EVALUATING PERFORMANCE R-SQUARED RUBYGARAGE2017 TECHNOLOGYMATTERS SUM (y - y^) -> min 2 ii Experience Simple Linear Regression: Salary ($) y^i yi
- 42. EVALUATING PERFORMANCE R-SQUARED RUBYGARAGE2017 TECHNOLOGYMATTERS SS = SUM (y - y^) 2 i ires SS = SUM (y - y ) 2 i avgtot yavg Experience Simple Linear Regression: Salary ($)
- 43. EVALUATING PERFORMANCE ADJUSTED R-SQUARED RUBYGARAGE2017 TECHNOLOGYMATTERS p - number of regressors n - sample size
- 44. ADJUSTED R-SQUARED STEP 3
- 45. ADJUSTED R-SQUARED STEP 4
- 46. ADJUSTED R-SQUARED STEP 5
- 47. POLYNOMIAL REGRESSION RUBYGARAGE2017 TECHNOLOGYMATTERS y x y = b + b x + b x1 1 2 1 2 0
- 48. POLYNOMIAL REGRESSION. DATASET PRESENTATION. BLUFFING DETECTOR RUBYGARAGE2017 TECHNOLOGYMATTERS
- 49. POLYNOMIAL REGRESSION. FITTING THE DATASET
- 50. POLYNOMIAL REGRESSION. TRAINING THE MODEL
- 52. SUPPORT VECTOR REGRESSION BASED ON SUPPORT VECTOR MACHINE
- 53. SUPPORT VECTOR REGRESSION. RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 55. DECISION TREE REGRESSION RUBYGARAGE2017 TECHNOLOGYMATTERS Split 4 Split 2 Split 1 Split 3 200 20 40 170 X1 X2 1023 0.7-64.1300.5 65.7 Y
- 56. DECISION TREE REGRESSION RUBYGARAGE2017 TECHNOLOGYMATTERS X < 201 X < 2002 300.5 65.7 1023 -64.1 0.7 X < 1702 X < 401 yes no yes no yes no yes no
- 57. DECISION TREE REGRESSION TRAINING
- 58. DECISION TREE REGRESSION RESULT RUBYGARAGE2017 TECHNOLOGYMATTERS
- 59. ENSEMBLE LEARNING. RANDOM FOREST REGRESSION. RUBYGARAGE2017 TECHNOLOGYMATTERS STEP 1: Pick at random K data points from the Training set. STEP 2: Build the Decision Tree associated to these K data points. STEP 3: Choose the number Ntree of trees you want to build and repeat STEPS 1 & 2 STEP 4: For a new data point, make each one of your Ntree trees predict the value of Y to for the data point in question, and assign the new data point the average across all of the predicted Y values.
- 60. RANDOM FOREST REGRESSION TRAINING
- 61. RANDOM FOREST REGRESSION RESULT RUBYGARAGE2017 TECHNOLOGYMATTERS
- 62. REGRESSION MODELS. PROS AND CONS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 63. CLASSIFICATION RUBYGARAGE2017 TECHNOLOGYMATTERS Unlike regression where you predict a continuous number, you use classification to predict a category. There is a wide variety of classification applications from medicine to marketing.
- 64. LOGISTIC REGRESSION RUBYGARAGE2017 TECHNOLOGYMATTERS This is new: Action (Y/N) Age We know this: Salary ($) Experience y = b0 + b1*x
- 65. LOGISTIC REGRESSION RUBYGARAGE2017 TECHNOLOGYMATTERS Action (Y/N) Action (Y/N) Age Age
- 70. DATASET PRESENTATION. SOCIAL NETWORK ADS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 73. LOGISTIC REGRESSION. TRAINING SET RESULTS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 74. LOGISTIC REGRESSION. TEST SET RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 76. K-NEAREST NEIGHBORS RUBYGARAGE2017 TECHNOLOGYMATTERS STEP 1: Choose the number K of neighbors STEP 2: Take the K nearest neighbors of the new data point, according to the Euclidean distance STEP 3: Among these K neighbors, count the number of data points in each category STEP 4: Assign the new data point to the category where you counted the most neighbors Your Model is Ready
- 78. K-NEAREST NEIGHBORS RUBYGARAGE2017 TECHNOLOGYMATTERS Category 1: 3 neighbors Category 2: 2 neighbors
- 80. K-NEAREST NEIGHBORS. TRAINING SET RESULTS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 81. K-NEAREST NEIGHBORS. TEST SET RESULTS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 83. SUPPORT VECTOR MACHINES TRAINING
- 84. RUBYGARAGE2017 TECHNOLOGYMATTERS SUPPORT VECTOR MACHINES. TRAINING SET RESULTS.
- 85. RUBYGARAGE2017 TECHNOLOGYMATTERS SUPPORT VECTOR MACHINES. TEST SET RESULTS.
- 95. NAIVE BAYES. TRAINING SET RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 96. NAIVE BAYES. TEST SET RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 99. NAIVE BAYES. BAYES THEOREM. WALKS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 100. NAIVE BAYES. BAYES THEOREM. DRIVES. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 107. NAIVE BAYES. NEW WALKER. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 109. NAIVE BAYES. TRAINING SET RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 110. NAIVE BAYES. TEST SET RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 112. DECISION TREE CLASSIFICATION. TRAINING.
- 113. DECISION TREE CLASSIFICATION. TRAINING SET RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 114. DECISION TREE CLASSIFICATION. TEST SET RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 115. RUBYGARAGE2017 TECHNOLOGYMATTERS RANDOM FOREST CLASSIFICATION STEP 1: Pick at random K data points from the Training set. STEP 2: Build the Decision Tree associated to these K data points. STEP 3: Choose the number Ntree of trees you want to build and repeat STEPS 1 & 2 STEP 4: For a new data point, make each one of your Ntree trees predict the category to which the data point belongs, and assign the new data point to the category that wins the majority vote.
- 116. RANDOM FOREST CLASSIFICATION. TRAINING
- 117. RUBYGARAGE2017 TECHNOLOGYMATTERS RANDOM FOREST CLASSIFICATION. TRAINING SET RESULTS
- 118. RUBYGARAGE2017 TECHNOLOGYMATTERS RANDOM FOREST CLASSIFICATION. TEST SET RESULTS.
- 119. EVALUATING CLASSIFICATION MODELS PERFORMANCE. FALSE POSITIVES & FALSE NEGATIVES. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 120. RUBYGARAGE2017 TECHNOLOGYMATTERS EVALUATING CLASSIFICATION MODELS PERFORMANCE. CONFUSION MATRIX.
- 121. CLASSIFICATION MODELS. PROS AND CONS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 122. CLUSTERING RUBYGARAGE2017 TECHNOLOGYMATTERS Clustering is similar to classification, but the basis is different. In Clustering you don’t know what you are looking for, and you are trying to identify some segments or clusters in your data. When you use clustering algorithms on your dataset, unexpected things can suddenly pop up like structures, clusters and groupings you would have never thought of otherwise.
- 124. K-MEANS CLUSTERING RUBYGARAGE2017 TECHNOLOGYMATTERS STEP 1: Choose the number K of clusters STEP 2: Select at random K points, the centroids (not necessarily from your dataset) STEP 3: Assign each data point to the closest centroid -> That forms K clusters STEP 4: Compute and place the new centroid of each cluster STEP 5: Reassign each data point to the new closest centroid. If any reassignment took place, go to STEP 4, otherwise go to FIN. Your Model is Ready
- 133. K-MEANS CLUSTERING RANDOM INITIALIZATION PROBLEM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 134. K-MEANS CLUSTERING RANDOM INITIALIZATION PROBLEM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 135. RUBYGARAGE2017 TECHNOLOGYMATTERS K-MEANS CLUSTERING RANDOM INITIALIZATION PROBLEM
- 136. RUBYGARAGE2017 TECHNOLOGYMATTERS K-MEANS CLUSTERING RANDOM INITIALIZATION PROBLEM
- 137. K-MEANS SELECTING THE NUMBER OF CLUSTERS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 138. K-MEANS SELECTING THE NUMBER OF CLUSTERS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 139. DATASET PRESENTATION. MALL CUSTOMERS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 140. K-MEANS. TRAINING. OPTIMAL NUMBER OF CLUSTERS.
- 141. K-MEANS. OPTIMAL NUMBER OF CLUSTERS RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 142. K-MEANS TRAINING
- 145. RUBYGARAGE2017 TECHNOLOGYMATTERS HIERARCHICAL CLUSTERING AGGLOMERATIVE STEP 1: Make each data point a single-point cluster That forms N clusters STEP 2: Take the two closest data points and make them one cluster That forms N-1 clusters STEP 3: Take the two closest clusters and make them one cluster That forms N-2 clusters STEP 4: Repeat STEP 3 until there is only one cluster FIN
- 159. DENDROGRAMS OPTIMAL NUMBER OF CLUSTERS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 160. DENDROGRAMS OPTIMAL NUMBER OF CLUSTERS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 161. DENDROGRAM. FINDING THE OPTIMAL NUMBER OF CLUSTERS.
- 165. CLUSTERING MODELS. PROS AND CONS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 166. REINFORCEMENT LEARNING Reinforcement Learning is a branch of Machine Learning, also called Online Learning. It is used to solve interacting problems where the data observed up to time t is considered to decide which action to take at time t + 1. It is also used for Artificial Intelligence when training machines to perform tasks such as walking. Desired outcomes provide the AI with reward, undesired with punishment. Machines learn through trial and error. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 167. THE MULTI-ARMED BANDIT PROBLEM Hot to bet to maximize your return RUBYGARAGE2017 TECHNOLOGYMATTERS
- 168. THE MULTI-ARMED BANDIT PROBLEM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 169. THE MULTI-ARMED BANDIT PROBLEM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 170. UPPER CONFIDENCE BOUND ALGORITHM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 171. UPPER CONFIDENCE BOUND ALGORITHM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 172. UPPER CONFIDENCE BOUND ALGORITHM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 173. UPPER CONFIDENCE BOUND ALGORITHM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 174. UPPER CONFIDENCE BOUND ALGORITHM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 175. UPPER CONFIDENCE BOUND ALGORITHM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 176. UPPER CONFIDENCE BOUND ALGORITHM RUBYGARAGE2017 TECHNOLOGYMATTERS
- 177. RANDOM SELECTION
- 179. UPPER CONFIDENCE BOUND. TRAINING.
- 180. UPPER CONFIDENCE BOUND. RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 184. CREATING DISTRIBUTION BASED ON AN INITIAL DATA RUBYGARAGE2017 TECHNOLOGYMATTERS
- 185. PULLING RANDOM VALUES FROM DISTRIBUTIONS RUBYGARAGE2017 TECHNOLOGYMATTERS
- 186. ADJUSTING THE PERCEPTION OF THE WORLD RUBYGARAGE2017 TECHNOLOGYMATTERS
- 188. UCB VS THOMPSON SAMPLING RUBYGARAGE2017 TECHNOLOGYMATTERS
- 189. THOMPSON SAMPLING ALGORITHM. TRAINING.
- 190. THOMPSON SAMPLING ALGORITHM. RESULTS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 191. NATURAL LANGUAGE PROCESSING RUBYGARAGE2017 TECHNOLOGYMATTERS Natural Language Processing (or NLP) is applying Machine Learning models to text and language. Teaching machines to understand what is said in spoken and written word is the focus of Natural.
- 192. NATURAL LANGUAGE PROCESSING RUBYGARAGE2017 TECHNOLOGYMATTERS Natural Language Processing (or NLP) is applying Machine Learning models to text and language. Teaching machines to understand what is said in spoken and written word is the focus of Natural.
- 193. NATURAL LANGUAGE PROCESSING RUBYGARAGE2017 TECHNOLOGYMATTERS Language Processing. Whenever you dictate something into your iPhone / Android device that is then converted to text, that’s an NLP algorithm in action.
- 194. NATURAL LANGUAGE PROCESSING RUBYGARAGE2017 TECHNOLOGYMATTERS You can use NLP on an article to predict some categories of the articles you are trying to segment. You can use NLP on a book to predict the genre of the book.
- 195. NATURAL LANGUAGE PROCESSING RUBYGARAGE2017 TECHNOLOGYMATTERS A very well-known model in NLP is the Bag of Words model. It is a model used to preprocess the texts to classify before fitting the classification algorithms on the observations containing the texts.
- 196. DATASET PRESENTATION. RESTAURANT REVIEWS. RUBYGARAGE2017 TECHNOLOGYMATTERS
- 197. NLP. TRAINING. IMPORTING THE DATASET AND CLEANING THE TEXTS.
- 198. NLP. TRAINING. CLEANING THE TEXTS. RESULTS.
- 199. NLP. TRAINING. CREATING THE BAG OF WORDS MODEL.
- 200. NLP. CREATING THE BAG OF WORDS MODEL.
- 201. NLP. TRAINING. SPLITTING THE DATASET INTO THE TRAINING SET AND TEST SET.
- 202. NLP. TRAINING. FITTING NAIVE BAYES TO THE TRAINING SET.
- 203. NLP. TRAINING. PREDICTING AND MAKING THE CONFUSION MATRIX.
- 204. NLP. CONFUSION MATRIX. RESULTS.
- 206. HOW DO NEURAL NETWORKS LEARN? RUBYGARAGE2017 TECHNOLOGYMATTERS