Machine Learning

• Definition of Learning


  A change in human disposition or capability that persists over a period of time and is not simply ascribable to processes of growth
• Definition of Machine Learning

  A type of artificial intelligence (AI) that allows software applications to become more accurate at predicting outcomes without being explicitly programmed to do so
• Understanding Machine Learning


  • Deals with huge amounts of data from disparate sources
  • Utilizes AI and machine learning to process big data
  • Machine learning models identify data patterns and make predictions

  
• Terminologies of Machine Learning

  • Model
  • Feature
  • Feature Vector
  • Training
  • Prediction
  • Target (Label)
  • Overfitting
  • Underfitting

  
• Importance of Machine Learning

  • Allows companies to transform processes previously only possible for humans
  • Can scale to handle larger problems and technical questions
• Importance of Data for Machine Learning

  Quality data is necessary for machine learning models to operate efficiently
• Dataset in Machine Learning

  A collection of instances (rows of data)
• Machine learning


  Algorithms that can perform tasks that were previously only possible for humans to perform - e.g. responding to customer service calls, bookkeeping, reviewing resumes
• Machine learning


  Can scale to handle larger problems and technical questions - e.g. image detection for self-driving cars, predicting natural disaster locations and timelines, understanding potential drug interactions
• That's why machine learning is important
• Data


  Necessary for machine learning models to operate efficiently

  
• Machine Learning (ML)


  A branch of Artificial Intelligence (AI) that allows software applications to become more accurate at predicting outcomes without being explicitly programmed to do so
• Dataset in machine learning

  A collection of instances (rows) that all share a common attribute
• Training machine learning models

  1. Feed training datasets into the machine learning algorithm

  2. Use validation/testing datasets to ensure the model is interpreting the data accurately

  3. Subsequent datasets can then be used to sculpt the machine learning model going forward

  
• Machine learning algorithms


  Use historical data as input to predict new output values
• Alan Turing published a paper answering the question "Can Machine Think?"


  1950

  
• The 6 elements of machine learning

  • Data
  • Task
  • Model
  • Loss Function
  • Learning Algorithm
  • Evaluation

  
• Data


  Information in all types and formats, including text, audio-video, structured and unstructured
• Data curation


  The process of getting the required data for a machine learning project

  
• First neural network designed by Frank Rosenbatt, called the Perceptron Model


  1957
• Bernard Widrow and Marcian Hoff created two neural network models, including the Adaline (Adaptive Linear Neuron) model


  1959

  
• Resources for publicly available datasets

  • Google Dataset Search
  • Kaggle Datasets
  • Indian Government Data
• Crowdsourcing marketplaces for data collection

  • Amazon Mechanical Turk
  • Dataturks
• All data needs to be encoded as numbers before feeding it to computers
• Supervised learning


  Machine learning with input data and corresponding output data, to learn the relationship and predict outputs for new inputs

  
• Evelyn Fix and Joseph Hodges developed a non-parametric method for pattern classification, later expanded by Thomas Cover as the K-Nearest Neighbor algorithm


  1951
• Unsupervised learning


  Machine learning with only input data, used for tasks like generation and clustering
• Gerald DeJong introduced the concept of Explanation-Based Learning (EBL)


  1986
• Supervised learning has created 99% of economic value in AI
• In the 1990s, the approach shifted from knowledge-driven to more data-driven, with programs created for computers to analyze large amounts of data and draw conclusions or "learn" from the results
• Model


  A mathematical function that defines the relationship between input data and output data
• Big data


  The enormous amount of data becoming easily available and accessible due to advanced computing capabilities and cloud storage
• Loss function


  Measures the difference between the true value and the value predicted by a model, used to determine the best model

  
• Model


  The mathematical representation of a real-world process, built by a machine learning algorithm and training data

  
• Loss functions


  • Square Error Loss
  • Cross Entropy Loss
  • KL Divergence
• Feature


  A measurable property or parameter of the data-set
• Feature Vector


  A set of multiple numeric features used as input to the machine learning model for training and prediction
• Training


  The process where an algorithm takes training data as input, finds patterns, and trains the model for expected results
• Prediction


  The process of feeding input data to the trained machine learning model to provide a predicted output
• Learning algorithm


  Optimizes the parameters of a model to minimize the loss function