Practical Machine Learning in F#
Machine learning is more popular than ever because there are several dataset available and we can use several tools at our disposal to learn an insight from this data.
In this session I shall show how F# can be used for several machine learning tasks and I will be using industry standard APIs
During this session participants will be solving several machine learning challenges from Kaggle like handwritten digit recognizer (https://www.kaggle.com/c/digit-recognizer)
During this session participants will write code in F# to solve real challenges like this one
https://gist.github.com/sudipto80/72e6e56d07110baf4d4d
and they will get to understand the process of machine learning system design pipeline.
Outline/Structure of the Demonstration
In this you shall learn how to use
In this Workshop you shall learn how to implement several algorithms to solve real world challenges from 
People will learn the following
1. F# Fundamentals (F# List comprehension)
2. Basics of Machine Learning
- Understanding Training/Test Corpus
3. Classification Algorithms implemented to solve real problems
- K Nearest Neighbor
- Logistic Regression
- Multi-class Logistic Regression
- Decision Trees
- Similarity Measurements
4. Linear Regression with F# to predict real values
- Simple Regression
- Multiple Regression
- Weighted Linear Regression
- Ridge Regression
- Multivariate Multiple Regression
5. Building Recommendation Systems
- Collaborative Filtering
- Item based filtering and recommendation
6. Clustering
- K Means Clustering to identify market segments
7. Ensemble Methods
- Random Forests
- AdaBoost
- Bagging
In general I shall show audience how to break a problem statement to a machine learning problem and how to solve it with existing APIs like WekaSharp, Accord.NET and F#. All the visualization will be implemented using FsPlot FsPlot helps draw the charts using HighCharts and is a very natural choice for using in F#. You can see the plot generated using FsPlot in action at http://recordit.co/SMNa4i7S8r
Learning Outcome
After attending this session
- People will be able to identify a problem as a regression/classification/recommendation problem
- Be able to find a solution using F# and other APIs (WekaSharp, FsPlot .NET, Accord.NET)
Target Audience
Software Developers, Architects, Data Scientists , Data Engineers
Video
Links
Sample code solving Digit Recognizer Challenge from Kaggle (https://www.kaggle.com/c/digit-recognizer)
schedule Submitted 8 years ago
People who liked this proposal, also liked:
-
keyboard_arrow_downSudipta Mukherjee - Agilie Data Analytics and Data Journalism
90 Mins
Demonstration
Intermediate
This is a demo session about Squirrel framework. Squirrel in an open source data analytics framework for .NET
https://github.com/sudipto80/Squirrel
Here are some of the case studies that will be demoed. These questions will be posed and then the solution will be provided using Squirrel
Hope that audience will be able to appreciate the elegance and simplicity of the solution that Squirrel provides.
1. Do women pay more tip than men ?
2. How many accidents happen becasue of bird strikes ?
3. Which country is famous for which sport in Olympics ?
Please visit Squirrel github page https://github.com/sudipto80/Squirrel to see some of these examples
At the end of this session, data from Squirrel will be taken and then some infographics will be generated that can make life lot simpler for data journalists.
In other words, it will be the attempt to make each developer realize that they can be a data journalist.
We shall have few T-Shirts to give away with the nice Squirrel logo and our motto. But I hope you shall have other motivation to come to this session :)
-
Naresh Jain
madhavan
Debasish Ghosh
Tejas Dinkar
Bhasker Kode
