addressalign-toparrow-leftarrow-rightbackbellblockcalendarcameraccwcheckchevron-downchevron-leftchevron-rightchevron-small-downchevron-small-leftchevron-small-rightchevron-small-upchevron-upcircle-with-checkcircle-with-crosscircle-with-pluscontroller-playcrossdots-three-verticaleditemptyheartexporteye-with-lineeyefacebookfolderfullheartglobegmailgooglegroupshelp-with-circleimageimagesinstagramFill 1light-bulblinklocation-pinm-swarmSearchmailmessagesminusmoremuplabelShape 3 + Rectangle 1ShapeoutlookpersonJoin Group on CardStartprice-ribbonprintShapeShapeShapeShapeImported LayersImported LayersImported Layersshieldstartickettrashtriangle-downtriangle-uptwitteruserwarningyahoo

Lockless Tables, NegaMax, and Numerical Optimization in Machine Learning

Our second meetup! After a brief introduction from myself we have two excellent talks lined up (see below). 

A massive thank you to Skillsmatter for hosting us. You will be required to register attendance with them directly as well here.

We will kickoff at 6:30pm sharp so please be on time. Do join us in the pub afterwards.

Hashtag: #bigoldn

Any questions, get in touch. 

Dirk (@elazungu)

___

Talks:

• Numerical Optimization by Jurgen van Gaal (@jvangael), Datascience Director at Rangespan

Numerical Optimization is at the core of many sciences. My personal background is in machine learning and pretty much any algorithm in this space does numerical optimization of a multivariate function: e.g. neural network training . In this talk I want to give an overview of multivariate numerical optimization. I'll cover line search, gradient descent, Newton methods and quasi-Newton methods. The goal of the talk will be for people to understand the different properties of different algorithms.

• Implementing Lock-less Look Up Tables in a NegaMax Search by Okash Khawaja, R&D at Draycir Ltd

NegaMax search is used to find the best possible decision in a zero-sum game under limited time. In many of its applications, the search encounters same situations that it had previously evaluated. Therefore one way to optimize it is to store its results in a look up table. However, when the search is parallelized, a lock-less look up table is required. In this talk we will closely follow the example of a chess engine as an application of negamax search and explore the implementation of such a lock-less look up table.

Join or login to comment.

  • Hoogendoorn W.

    same as everybody, good topics, would like moretime for questions

    October 3, 2013

  • Dirk G.

    An excellent meetup with a great turnout. Thanks to skillsmatter for hosting and thanks to Jurgen and Okash for speaking. The videos are online here: http://skillsmatter.com/event-details/home/lockless-tables-negamax-and-numerical-optimization-in-machine-learning

    Next meetup is being planned. Please leave feedback, ratings, comments, etc.

    October 3, 2013

    • Dirk G.

      Thanks for coming. I was mistaken and it is Yoshua Bengio, not Yann LeCun, but still good :) Link is http://www.csml.ucl.a...­ Unfortunately the monday event is sold out but the others are still open.

      October 3, 2013

    • Dirk G.

      PS: all the meetup material will appear here: https://github.com/big...­

      October 3, 2013

  • A former member
    A former member

    Interesting topics. I would have liked a little more time for questions.

    October 3, 2013

    • Dirk G.

      Valid point John. We were running a bit behind schedule but I will probably adjust the schedule for the next one. Of course there is also always the pub session afterwards. Thanks for coming!

      October 3, 2013

  • JUNIOR S.

    Sorry can't make it, person who I was waiting for has just arrived.

    October 2, 2013

  • Dirk G.

    Sorry, linked fixed. Thanks for reporting.

    September 30, 2013

  • Hoogendoorn W.

    just saw i "A massive thank you to Skillsmatter for hosting us. You will be required to register attendance with them directly as well here." but the link doesn't work, what to do now?

    1 · September 30, 2013

  • A former member
    A former member

    I have been taking Algorithms: Design and Analysis, Parts 1 and 2 on Coursera.org. The teacher, Professor Tim Roughgarden of Stanford, is excellent. I highly recommend.

    September 18, 2013

Our Sponsors

People in this
Meetup are also in:

Sign up

Meetup members, Log in

By clicking "Sign up" or "Sign up using Facebook", you confirm that you accept our Terms of Service & Privacy Policy