Sep. 28, 2018 - Microsoft Ignite - Machine Learning with ML.NET

These are some notes from the talk at Microsoft Ignite.

Source:

Machine Learning with ML.NET

https://www.youtube.com/watch?v=yDa8zsyKJRw

Program the unprogrammable

Machine Learning creates a Model using this data

Many ML tasks

github demo label issue

Pre-built machine learning models (Azure Cognitive Services)

Limitations with pre-built machine learning models

Build your own custom machine learning models

pre-built vs custom machine learning models

custom machine learning models

ML.NET is a framework for building custom ML Models

ML.NET - https://github.com/dotnet/machinelearning

Machine Learning Samples - https://github.com/dotnet/machinelearning-samples

Sentiment Analysis Sample - https://github.com/dotnet/machinelearning-samples/tree/master/samples/csharp/getting-started/BinaryClassification_SentimentAnalysis

sentiment analysis - is it A or B problem

sentiment analysis - feature input and label output.png

Machine learning workflow

Text Featurizer - text to numeric - ngram extraction

important concepts - data, transformer, estimator, prediction

Transformer - n-gram

Estimator - create Model

Prediction - single input -> single output

demo sentiment analysis binary classification

- code

- metrics

- prediction function

demo github multiclassification

More data you have the better results you may get

iterative machine learning

different pipeline

- different user, repository (data)

- different learner

demo movie recommendations

Collaborative filtering approach

prediction label and scores