Machine Learning

Machine Learning

1. 1.Supervised Learning

   • Classification problem
     • At training time , the clasifier uses labelled examples to learn rules for recognizing each element
     • After training, at prediction time, the trained model is used to predict the fruit type for new intances using the learnig rules
   • Regression problem
   • supervised learning algorithms because they use labeled examples in a training set to learn how to predict the labels for new, previously unseen examples.

2. 2. Unsupervised Learning

   • Finding useful structure or knowledge in data when no labels are available. Examples: Finding clusters of similar users; Detecting abnormal server access paterns. Example: clustering
   • ... where we don't have labels for the training data examples

3. Basic Machine Learning WorkFlow

   • 1. Representation
     • Choose : a feature representation and type of classifier to use
   • 2. Evaluation
     • Choose: what criterion distinguishes good vs bad classifiers ?
   • 3. Optimization
     • Choose: how to search for the settings / parameters that give the best classifier for this evaluation criterion

4. ML in Python

   • Pandas
   • Scikit-learn
     • Models
       • Supervised Learning
         • 1. Linear Regression
         • 2. Supported Vector Machines
         • 3. Naive Bayes
         • 4. KNN
       • Unsupervised Learning
         • 5. PCA ( Principal Component Analysis )
         • 6. K Means
   • SciPy
   • NumPY
   • MatplotLib
   • Anaconda Python Distribution
     • Jupyter - Notebook !

5. Methods

   • Momentum
   • Early Methods
   • Recent Advances
     • Convolutional layers for visual invariance
     • Use of DropUp
     • Introduction of faster Gadient descent methods: RMSProp
     • Use od Rectified Linear Unit (ReLU) activation functions
     • Introduction of Autoencoder architecture
       • to provide un-supervised model learning or pre-training

6. Computational advances

   • Skflow
   • TensorFlow
     • Main Classes
       • tf.Graph ( ) tf.operation ( ) tf.Tensor( ) tf.Session()
     • useful Functions
       • tf.get_default_session ( )
   • Keras