Machine Learning Overview
Download as PPTX, PDF•12 likes•1,303 views
The document provides an overview of machine learning, its applications, problem definitions, and approaches, highlighting significant advancements such as self-driving cars and speech recognition. It discusses various types of learning, including supervised, unsupervised, and reinforcement learning, as well as the evolution of machine learning as a service, particularly through Microsoft Azure. Additionally, it includes resources and examples for practical applications, such as the Titanic survival prediction project.
Machine Learning Overview
- 1. www.eleks.comwww.eleks.com Machine Learning overview: applications, problems, approaches, services Mykhailo Koval Database Developer/Tech Lead, Eleks [email protected]
- 3. Abstract Machine learning is the science of getting computers to act without being explicitly programmed. In the past decade, machine learning has given us self-driving cars, practical speech recognition, effective web search, and a vastly improved understanding of the human genome. Machine learning is so pervasive today that you probably use it dozens of times a day without knowing it. We will review some modern machine learning applications, understand variety of machine learning problem definitions, go through particular approaches of solving machine learning tasks. This year 2015 Amazon and Microsoft introduced services to perform machine learning tasks in cloud. Microsoft Azure Machine Learning offers a streamlined experience for all data scientist skill levels, from setting up with only a web browser, to using drag and drop gestures and simple data flow graphs to set up experiments. We will briefly review Azure ML Studio features and run machine learning experiment.
- 4. Agenda ● Machine Learning definition ● Applications Overview ● Types of Problems and Tasks ● Machine Learning as a Service (Azure) ● Machine Learning experiment in Azure ML Studio
- 6. Machine Learning Machine Learning (ML) focuses on the development of computer programs that can teach themselves to grow and change when exposed to new data.
- 7. Machine Learning vs Data Mining Focuses on the discovery of (previously) unknown properties in the data Focuses on prediction, based on known properties learned from the training data
- 9. Machine Learning Uses training data to improve the program's future actions. Reproduces known knowledge.
- 11. Computer vision
- 15. Credit card fraud detection
- 16. Spam filtering
- 20. Junior, a robotic Volkswagen Passat, at Stanford University in October 2009 Autonomous Vehicles
- 22. Reinforcement learning Machine learning problem categories Three broad categories, depending on the nature of the learning "signal" or "feedback" available to a learning system: Supervised learning Unsupervised learning
- 25. Supervised learning goal The computer is presented with example inputs and their desired outputs, given by a "teacher", and the goal is to learn a general rule that maps inputs to outputs
- 26. Problem definition x11 x12 … x1n x21 x22 … x2n x31 x32 … x3n … … … … xm1 xm2 … xmn y1 y2 y3 ym Training Examples “input” variable / features “output” variable h(x) = h(x1, x2, …, xn) – hypothesis function and solution of a supervised learning problem, where h(x) ≈ y (as close as possible)
- 27. Regression problem Predicting results within a continuous output, meaning that we are trying to map input variables to some continuous function.
- 28. Linear regression hθ(x) – linear hypothesis function, m – number of training examples, Θ – vector of coefficients of the linear function hθ(x) Cost Function
- 29. Classification problem Predicting results in a discrete output. In other words, we are trying to map input variables into discrete categories.
- 30. Supervised learning diagram Pre- Processing Sampling Training Dataset Pre- Processing Learning Algorithm Training Parameter Optimization Post- Processing Final Model Feature Selection Feature Scaling Dimensionality Reduction Test Dataset New Data Missing Data Feature Extraction Performance Metrics Model Selection Split Cross-Validation Refinement Final Model Evaluation Prediction
- 31. Unsupervised learning No labels are given to the learning algorithm. The goal is to find hidden structure in its input. Since the examples given to the learner are unlabeled, there is no error or reward signal to evaluate a potential solution
- 32. Supervised vs Unsupervised learning
- 33. Problem definition x11 x12 … x1n x21 x22 … x2n x31 x32 … x3n … … … … xm1 xm2 … xmn y1 y2 y3 ym Training Examples “input” variable / features no “output” variable So no hypothesis function!
- 34. Unsupervised learning examples Social network analysis Astronomical data analysis
- 35. Unsupervised learning examples Market segmentation
- 36. Clustering The result of a cluster analysis shown as the coloring of the squares into three clusters
- 37. Clustering methods (over 100) k-Means (Centroid-based) Expectation-Maximization (Distribution-based) DBSCAN (Density-based)
- 38. Reinforcement learning A Toyota Prius modified by Google to operate as a driverless car.
- 39. Reinforcement learning Performing a certain goal (such as driving a vehicle) in a dynamic environment, without a teacher explicitly telling it whether it has come close to its goal or not.
- 40. Reinforcement learning differences No correct input/output pairs Sub-optimal actions aren’t explicitly corrected Instead it maximizes some notion of cumulative reward There is a focus on on-line performance Finds a balance between exploration (of uncharted territory) and exploitation (of current knowledge)
- 41. Basic reinforcement learning model a set of environment states a set of actions rules of transitioning between states rules that determine the scalar immediate reward of a transition rules that describe what the agent observes
- 42. Machine Learning As a Service (Azure)
- 43. Machine Learning evolution SQL Server Data Mining Spam filtration Gestures under- standing in Microsoft Kinect Azure Machine Learning Using Data Mining in search engines Bing Maps started to use ML for traffic estimate Voice recognition 1999 201220082004 201420102005
- 44. Azure Machine Learning HDInsight Azure Storage Desktop Data PowerBI / DashboardsMobile AppsWeb Apps ML API service Developer ML Studio Data Analyst
- 48. On April 15, 1912, during her maiden voyage, the Titanic sank after colliding with an iceberg, killing 1502 out of 2224 passengers and crew. https://www.kaggle.com/c/titanic
- 49. PassengerId Survived Pclass Name Sex Age SibSp Parch Ticket Fare Cabin Embarked 1 0 3 Braund, Mr. Owen Harris male 22 1 0 A/5 21171 7.25 S 2 1 1 Cumings, Mrs. John Bradley (Florence Briggs Thayer) female 38 1 0 PC 17599 71.2833 C85 C 3 1 3 Heikkinen, Miss. Laina female 26 0 0 STON/O2. 3101282 7.925 S 4 1 1 Futrelle, Mrs. Jacques Heath (Lily May Peel) female 35 1 0 113803 53.1 C123 S 5 0 3 Allen, Mr. William Henry male 35 0 0 373450 8.05 S 6 0 3 Moran, Mr. James male 0 0 330877 8.4583 Q rows 891 columns 12Titanic Dataset https://www.kaggle.com/c/titanic/data sibsp - Number of Siblings/Spouses Aboard parch - Number of Parents/Children Aboard embarked - Port of Embarkation: C = Cherbourg Q = Queenstown S = Southampton
- 51. Recommended courses https://www.coursera.org/learn/machine-learning/home/info# Machine Learning by Stanford University Andrew Ng Associate Professor, Stanford University; Chief Scientist, Baidu; Chairman and Co-founder, Coursera
- 52. Useful Links Free eBook: Microsoft Azure Essentials: Azure Machine Learning http://blogs.msdn.com/b/microsoft_press/archive/2015/04/15/free-ebook-microsoft-azure- essentials-azure-machine-learning.aspx Azure Machine Learning для Data Scientist http://habrahabr.ru/company/microsoft/blog/254637/ Azure Machine Learning: Get started now http://azure.microsoft.com/uk-ua/services/machine-learning/ Tutorial: Building a classification model in Azure ML http://gallery.azureml.net/Experiment/01b2765fa75147ce99679e18482d280f
- 54. www.eleks.comwww.eleks.com Machine Learning overview: applications, problems, approaches, services Mykhailo Koval Database Developer/Tech Lead, Eleks [email protected]