Software, visits, and talks

Hi all,

We just wanted to briefly share some new work from our lab: check out https://github.com/Nth-iteration-labs/contextual. Robin is working on this amazing [R] package for (contextual) bandit simulation, supporting both online and offline simulations. The package is very versatile; we hope to be releasing a proper vignette and demo soon. But right now, have a look and let us know if you have any comments!

Also, we are pretty excited that Matthew Pratola will be visiting JADS and our lab in the coming months. We will be working together with Reza Mohammadi on BART. Also, Matthew will give a talk (see here for details).

Cheers!

Lock in learning

We would like to share this small code snippet regarding lock-in feedback and the opportunity of using it for (online) learning of supervised learning models. For more info please see: https://www.nature.com/articles/palcomms201682 and its supplementary materials. However, linRegLif is an R script that nicely demonstrates some of the possibilities. Note that this depends on the S4 class for Lock-in Feedback: lifclass.

Lianne Ippels wins GOR thesis award!

We are happy to announce that lab-alumni Lianne Ippel won the GOR thesis award 2018 for her thesis “Multilevel Modeling for Data Streams with Dependent Observations”!

You can find the announcement here. Also, find a copy of Lianne’s thesis here, and see her publications here.

Congratulations Lianne!

Streaming estimation of (mixtures of) logistic regression models

Hi all,
We (Maurits Kaptein & Paul Ketelaar) are happy to share the following ArXiv pre-print with you: http://arxiv.org/abs/1802.10529.
The paper describes a novel method for fitting mixtures of logistic regression models in data streams. The paper is accompanied by an [R] package to carry out the analysis: https://github.com/Nth-iteration-labs/ofmlr.
The method is useful and promising for analyzing (mixture) models in data streams. Also, the paper is quite didactical in nature; we have tried to explain in quite some detail how one can deal with streaming data, and how our online version of EM works (and why).
However, since we do not have a clear case study for the project, and this was a bit of a “pet project” of both of us we are advocating the work here, but we will not be seeking further publication. We hope you enjoy the work! And, let us know if you end up using / extending it!
Maurits

Update

Hi all,

We have a bunch of updates to share:

  1. From May onwards Hongyi Chen will be joining our lab; for updates see people.
  2. Two papers that were accepted already can  now be found online; find our paper on behavioral vs. personality measures to predict persuasive outcomes here: http://www.emeraldinsight.com/toc/jcm/0/ja, and find our (pretty awesome) paper on automated adaptive selling here http://www.emeraldinsight.com/toc/ejm/0/0.
  3. On March 15th we have Robert Neal visiting at JADS. Robert is Principal Software Engineer on Experimentation and Analytics Platform at Udemy  and will be talking about (Bayesian) methods for large scale experimentation. (see meetup group).
  4. Last Tuesday we had a great presentation by Robin van Emden on the new package contextual (see our labs GitHub); You can find the slides here: http://www.robinvanemden.dds.nl/rdevslides/).
  5. Over the last week we had the honor of having Dean Eckles visit our lab. It was a great time (including some skiing); for more on Deans work see http://www.deaneckles.com.

Finally, we have updated our project description page.

Inaugural speech

This Friday I had the great honor to be able to provide my inaugural speech at JADS. Thanks to the whole team at JADS the event was a great success. And, the venue looked great:

For those interested, please see the full text of the speech here:

Kaptein, M.C. (2018) Computational Personalization: Data science methods for personalized health. Inaugural Address at the University of Tilburg / JADS Den Bosch.

Additionally, here are my slides, and [1][2][3][4] are the snippets of [R] code I used to run the simulations presented in the talk. Do contact me if you have any questions.

Thanks!

Maurits

Paper accepted, grants, and software

Hi all,

We are happy to announce that the paper “Automated Adaptive Selling” by prof. dr. M.C. Kaptein, prof. dr. Petri Parvinen, and prof. dr. Rick McFarland was accepted for publication at the European Journal of Marketing. We will be posting a pre-print of the paper here as soon as it is available, but do contact us if you are interested in receiving a copy directly. The paper describes tree (large-scale) field evaluations of the effect of automatically adapting, at the individual level, influence strategies in e-commerce. The paper present both the empirical evaluation, as well as extensive simulations showing the effectiveness of the algorithms used.

Next, we are proud to say that Xynthia Kavelaars her research proposal with NWO has progressed into the next round; congratulations to Xynthia!

Finally, we would like to share some of the new software projects we are working on (see https://github.com/Nth-iteration-labs). Primarily, see the “contextual” repository for a great [R] package by Robin van Emden created to perform evaluations of bandit policies.

That’s it for now, enjoy the holiday!

Maurits

Updates

Hi,

A few updates:

    1. We are very proud to be hosting a health-data science meetup at JADS: see for updates on our events and activities: https://medium.com/data-science-backchannel
    2. We are proud to announce that Bas Willemse is starting his PhD project at the computational personalization group at JADS; welcome Bas!
    3. We have updated some of our software downloads. Have a look around.