Profile

Cover photo
Jean-François Puget
Works at IBM
Attended École Normale Supérieure
Lives in France
295 followers|33,418 views
AboutPostsPhotosVideos
People
Have him in circles
295 people
Emmanuel Hebrard's profile photo
Aditya Singh's profile photo
murielle dupont's profile photo
Mark Eastwood's profile photo
Marie Debora's profile photo
Albert Soufi's profile photo
Marie Courtois's profile photo
Aurelie Thiele's profile photo
Angelika Parrisse's profile photo
Education
  • École Normale Supérieure
    1983
Basic Information
Gender
Male
Work
Employment
  • IBM
    Distinguished Engineer, ILOG Optimization, 2009 - present
    Technical executive at IBM Software Group. Currently in charge of making analytics and mathematical optimization used extensively within IBM Industry Solutions. Helps drive IBM agenda on Analytics. In addition to the above role, technical lead of a highly skilled team (including more than 30 PhDs) in charge of IBM optimization products (CPLEX and ODME). Skills in mathematical optimization and constraint programming. Well connected in the academic community. Has established ILOG as the leader in the optimization software market. Has grown ILOG Optimization business inside IBM. Co-chair of IBM Business Analytics and Optimization Architecture Board Member of IBM Software Group Architecture Board Steering Committee Member of IBM Academy of Technology Leadership Team Has published over 60 scientific papers in refereed journals and international conferences with peer reviews.
Places
Map of the places this user has livedMap of the places this user has livedMap of the places this user has lived
Currently
France
Links
Other profiles
Contributor to

Stream

Jean-François Puget

Shared publicly  - 
 
Co-authored by Jean-Francois Puget (@JFPuget) Machine Learning represents the new frontier in analytics, and is the answer of how many companies can capitalize on the data opportunity. Machine Learning was first defined by Ar...
1
Add a comment...

Jean-François Puget

Shared publicly  - 
 
Have you ever heard of overfitting?  I bet you did if you are using machine learning one way or another.  Avoiding overfitting is a key concern for machine learning practitioners and researchers, and it led to the development of many techniques to avoid it, such as cross validation and regularization.  I claim that OR practitioners, especially those using mathematical optimization solvers, should be equally worried about overfitting. In a nutshel...
1
Add a comment...

Jean-François Puget

commented on a post on Blogger.
Shared publicly  - 
 
CLP isn't that bad, but ... CPLEX was used with one thread.

It confirms that clp is by far the best open source solver.
http://orinanobworld.blogspot.com/2016/06/using-clp-with-java.html CLP is not doing so poorly on this problem (note the log/log scale however).
3
Add a comment...

Jean-François Puget

Shared publicly  - 
 
Can you explain me what machine learning is?  I often get this question from colleagues and customers, and answering it is tricky.  What is tricky is to give the intuition behind what machine learning is really useful for.  I'll review common answers and give you my preferred one.  Cognitive Computing The first category of answer to the question is what IBM calls cognitive computing .  It is about building machines (computers, software, robots, w...
1
Add a comment...

Jean-François Puget

Shared publicly  - 
 

How easy it is to build a learning machine?  Shouldn't one just hire some Machine Learning PhDs and have them run their algorithms?  Well, this is most probably a good idea, but it won't be enough.  I'll try to explain why in this blog entry.
1
Add a comment...

Jean-François Puget

Shared publicly  - 
 

XGBoost is a recent implementation of Boosted Trees.  It is a machine learning algorithm that yield great results on recent Kaggle competitions.  I decided to install it on my computers to give it a try.  Installation on OSX was straightforward using these instructions.  Installation on Windows was not as straightforward.  I am sharing what worked for me in case it might help others.  I describe how to install for the Anaconda Python distribution, but it might work as-is for other Python distributions.
developerWorks blogs allow community members to share thoughts and expertise on topics that matter to them, and engage in conversations with each other.  You can browse for and follow blogs, read recent entries, see what others are viewing or recommending, and request your own blog.
1
Jean-François Puget's profile photoAnup Agarwal's profile photo
6 comments
 
Yes, I guess you are right. Thank you so much for helping me out.
Add a comment...
Have him in circles
295 people
Emmanuel Hebrard's profile photo
Aditya Singh's profile photo
murielle dupont's profile photo
Mark Eastwood's profile photo
Marie Debora's profile photo
Albert Soufi's profile photo
Marie Courtois's profile photo
Aurelie Thiele's profile photo
Angelika Parrisse's profile photo

Jean-François Puget

commented on a post on Blogger.
Shared publicly  - 
 
HI John,  I think that what you describe applies to data mining, i.e. the automated discovery of correlations.  Indeed, as you point out, the more data, the more likely there will be correlations.

I don't think this applies to (supervised) machine learning where you define a priori rthe correlation you are looking for, typically how to compute a target feature from other features.  Then, the more data, the better. 
http://tinyurl.com/jeyjtna Data driven decision making has proven to be key for organisational performance improvements. This stimulates organisations to gather data, analyse it and use decision support models to improve the...
1
Add a comment...

Jean-François Puget

Shared publicly  - 
 
Do you get what overfitting means in machine learning?  If you don't, then you better learn about it if you want to use or leverage machine learning.  Why?  Because overfitting can ruin the effectiveness of machine learning.  I wrote this blog because I found existing explanations of overfitting to be too technical.  I hope this one is more consumable by non specialists.  The Machine Learning Workflow Machine learning involves a fairly complex wo...
1
Add a comment...

Jean-François Puget

commented on a post on Blogger.
Shared publicly  - 
 
This is why indicator variables were introduce din CPLEX.  It moves modeler away from having to chose the value for big M, while staying focused on what they need to model.
A major issue in MIP modeling is choosing good values for big-M constants. The poorly chosen name 'big-M' indicates we should use a really big value which is exactly the opposite of what we should do, Here we see an extreme e...
2
Add a comment...

Jean-François Puget

Shared publicly  - 
 
Instructions to install one of the trendiest machine learning algorithm on OSX.
developerWorks blogs allow community members to share thoughts and expertise on topics that matter to them, and engage in conversations with each other.  You can browse for and follow blogs, read recent entries, see what others are viewing or recommending, and request your own blog.
1
Add a comment...

Jean-François Puget

Shared publicly  - 
 

How to mislead people with p-values hacking.
1
Add a comment...