Machine Learning and (some) Deep Learning

2019-05 Metropolis Hastings

• Manual implementation of the Metropolis Hastings Random Walk algorithm 201905_metropolis-hastings.ipynb

2019-04 Gaussian Processes

• Step by step explanation of getting the conditional distribution of unknown points given data using gaussian processes and an exponential quadratic kernel
  201904_gaussian-processes.ipynb

2018-09 Forward-Backward Algorithm for Hidden Markov Models

• Implements the algorithm to compute the probabilities of hidden states given observations
  201809_forward-backward.ipynb

2018-08 Appliances Energy Prediction

• Tries various models to predict the consumption of home appliances.
  201808_appliances_energy.ipynb

2018-02 Probabilistic Graphical Models

• Manual fun implementation of factors for bayesian network. Application with generic disease (simplified Cystic Fibrosis)
  Genetic_Bayesian_Network.ipynb

• Learning Structure and Parameters of the Flu dataset
  PGMs_flu.ipynb

• A Tour of Bayesian Networks with pgmpy with a Restaurant CPD example
  PGM_restaurant_tour.ipynb

2017-09 Predicting Customer Calls from Network Anomalies

• Network elements emit health measurements (CPU temperature, signal quality, number of connected devices...). When measurement's Z-Score is above some threshold compared to expected value ==> anomaly
• This notebook shows how a neural network can learn to predict from historical data increases of customer calls when network anomalies are detected
  Predict_calls_from_anomalies.ipynb

2017-03 Singular Value Decomposition

• Example of Singular Value Decomposition applied to Image Compression
  SVD_recommendation_and_img_compression.ipynb

2017-02 EKG Anomaly Detection with Clustering

• Shows how K-Means clustering can be used to build an expected signal, can be used later to detect anomalies
  AnomalyDetection_EKG_with_kmeans.ipynb

2016-07 Badges

• Find how the badges at a 1994 machine learning conference were labeled
  badges.ipynb

2016-07 Study of Adult Income Data Set

• This is a study of this data set: https://archive.ics.uci.edu/ml/datasets/Adult This tries several methods to get the best F1 score. Focuses on various feature selection methods
  adult-income.ipynb

2015-02 SVM with RBF Kernel

• Shows how SVM with RBF kernel can better fit non linear data (toy example)
  SVM_RBF_Kernel.ipynb

2014-09 Step by Step Gradient Descent

• Manual by-the-book implementation of gradient descent
  gradient-descent.ipynb

2014-09 Naive Bayes

• Manual implementation of Naive Bayes classification to understand what's under the hood!
• Application to email spam classification with Naive Bayes
  naive-bayes.ipynb

2014-09 K Nearest Neighbors

• Using KNN algo to detect non linear boundaries
  knn.ipynb