Profile cover photo
Profile photo
Laurens van der Maaten
601 followers
601 followers
About
Laurens's posts

Post has attachment
Call for Participation: Benelearn 2015
==============================
June 19, 2015; Delft, The Netherlands
http://www.benelearn2015.nl

The 24th Belgium-Netherlands Conference on Machine Learning (Benelearn) will be held on June 19, 2015 in Delft, The Netherlands. We have a very exciting program with many high-quality contributions and three excellent keynote speakers: 
- Charles Elkan (University of California, San Diego / Amazon)
- Kilian Weinberger (Washington University in St. Louis / Cornell University)
- Tim Salimans (Kaggle Top 10 Data Scientist). 

Everyone is cordially invited to attend the conference! Registrations costs are just 25 euros + administration costs thanks to our great sponsors: GoDataDriven, PrimeVision, Delft Data Science, NWO, and BNVKI.

For additional information, please visit www.benelearn2015.nl or contact me.

Post has shared content
Nice PhD opportunity in Guelph.
PhD position in Deep Learning/Representation Learning, University of Guelph

I have a fully-funded PhD position open in Deep Learning and Representation Learning to start no later than January 2015. The position is partially funded by industry, so the project will involve some time on campus at the University of Guelph and some time at a startup in downtown Toronto (scheduling is flexible). The project involves developing new  representation learning methods for unstructured, event-based time series. 

Ideal candidates will have a MSc or equivalent degree in Computer Science, Physics, Mathematics, or in a related area, and have a strong interest and experience in machine learning, statistics and/or scientific computing.

Guelph is a vibrant university community 28 kilometres east of Waterloo and 100 kilometres west of downtown Toronto, with commuter train access to both. It is consistently rated as one of Canada’s best places to live. 

The position is open to Canadian and foreign students. To apply, please send the following documents in a single pdf-file to my attention.

(1) Your Curriculum Vitae
(2) A brief statement of relevant research interests and/or experience (preferably one page, at most two)
(3) Transcripts for previous degrees if available
(4) Names of two or three referees who are able to comment on the applicant's qualifications. I will contact references of short-listed candidates only.

Only short-listed candidates will be contacted.

Post has attachment
I am excited to announce that I will soon be joining Facebook AI Research to work on the next generation of artificial intelligence systems. I will start working in Facebook's New York office in February 2015. Danique will join Columbia University as a postdoc.

Post has shared content
Nice PhD opportunity in Amsterdam.
The Informatics Institute at the University of Amsterdam invites applications for a fully funded position for a PhD student in the area of machine learning and involves applying deep learning to astronomical data.

Supervisor: Max Welling
Salary indication: € 2.083,-- to € 2.664,-- gross per month, based on 38 hours per week

Job / Project Description:
Deep Learning for Big Data Analysis in Astronomy.
With an estimated Exabyte of data per day in 2024, the new “Square Kilometer Array” radio telescope will be one of the largest generators of big data ever. In this project we will train deep neural networks to reconstruct the image of radio sources from the sparse (Fourier transformed) data received at the telescopes. We will use “Stochastic Variational Bayesian” learning to jointly train the encoding network (observations to sources) as well as the generative network (sources to observations). Models will be trained from simulator data as well as from real observations. A major challenge will be to develop deep learning algorithms that can handle the enormous amounts of data generated by these telescopes and reconstruct the sources efficiently and accurately.
The PhD student will be conducting research in the area of machine learning on the topic described above. In addition s/he may have some teaching duties as a teaching assistant. The PhD student will be directly supervised by Prof. Dr. M. Welling.

Institution: 
The research group Machine Learning & Autonomous Intelligent Systems of IvI conducts research in the area of large scale modeling of complex data sources. This includes the development of new methods for probabilistic graphical models and nonparametric Bayesian models, the development of faster (approximate) inference and learning methods, deep learning, causal inference, reinforcement learning and multi-agent systems and the application of all of the above to large scale data domains in science and industry (“Big Data problems”). The Machine Learning and Autonomous Intelligent Systems group is embedded in the Intelligent Systems Lab Amsterdam (ISLA). ISLA conducts research in sensory information processing and autonomous systems. It produces theoretically as well as applied research. ISLA has an outstanding group of researchers and collaborates with national and international research institutes and companies.

Requirements
Candidates are required to have a master's degree in mathematics, statistics or physics (preferably with a specialization in computational methods) or computer science (preferably with a specialization in artificial intelligence and/or machine learning).
Necessary qualifications for candidates include excellent grades, proven research talent, affinity with computational statistics or machine learning and excellent programming skills. 
Candidates are expected to have an excellent command of English, and good academic writing and presentation skills. Applicants are kindly requested to motivate why they have chosen to apply for this specific position.

Further Information
Further information about this vacancy can be obtained from Prof. Dr. M. Welling, email: m.welling@uva.nl.

Appointment
The appointment will be full time (38 hours a week) for a period of four years (initial employment is 18 months and after a positive evaluation, the appointment will be extended further with 30 months) and should lead to a dissertation (PhD thesis). An educational plan that includes attendance of courses and national and international meetings will be drafted.  The PhD student is also expected to assist in teaching of undergraduates. The salary is in accordance with the university regulations for academic personnel. The PhD student salary will range from € 2.083,-- (first year) up to a maximum of  € 2.664,-- (last year) before tax per month (scale P) based on a full-time appointment.  There are also secondary benefits, such as 8% holiday allowance per year and the end of year allowance of 8.3%. The Collective Employment Agreement of Dutch Universities is applicable.
English is the working language within the Informatics Institute. Moreover, since Amsterdam is a very international city where almost everybody speaks and understands English, candidates need not be afraid of the language barrier.

Job Application
Applications may only be submitted by sending your applications to application-science@uva.nl. To process your application immediately, please mention the position you are applying for in the subject-line. Applications must include  a motivation letter explaining why you are the right candidate, curriculum vitae, a copy of your master’s thesis, a list of projects you have worked on (with brief descriptions of your contributions and the names and contact addresses of two academic references).
All these should be grouped in one PDF attachment. 
Applications will be accepted until 1 October 2014.
Photo

Post has shared content
What happens when you extract CNN codes from 50,000 ImageNet validation images and embed them with t-SNE in 2D? Find out:

http://cs.stanford.edu/people/karpathy/cnnembed/

:)
looks very pretty!

Post has attachment
I just posted a long paper of my work on regularization with +Minmin Chen , Stephen Tyree, and +Kilian Weinberger on Arxiv. The paper is an extension of our earlier ICML paper on regularizing linear models by adding exponential-family noise (such as dropout) to the training data, and marginalizing over this noise. We show that this leads to interesting new regularizers for exponential and logistic loss functions, which may improve the performance of your linear classifiers.

http://arxiv.org/abs/1402.7001

Post has attachment
I am PC member for the Workshop on Interactive Language Learning, Visualization, and Interfaces. The workshop aims to attract a mix of researchers from information visualization, natural language processing, and machine learning. The workshop will he held in conjunction with ACL on June 27, 2014. The deadline for papers is March 21, 2014. This workshop sounds like it will be a lot of fun, so I encourage all of you to submit!

http://nlp.stanford.edu/events/illvi2014/index.html

Post has shared content
My Google Techtalk on t-SNE is now available on Youtube.
Visualizing High Dimensional Data 

Data visualization tools are often used to communicate complex information in a more intuitive way. By representing high dimensional data (e.g. objects with 4 or more variables) by two-dimensional points, in such a way that similar objects are represented by nearby points and dissimilar objects are represented by distant points, scientists can gain intuition by examining the presence of structure and clustering in plots of the data.

Unfortunately, the majority of current visualization techniques can only be used to inspect a limited number of variables of interest simultaneously, and are not suitable for big data (http://goo.gl/DFFbr) that contains more than two or three dimensions.

Recently, Google hosted Delft University of Technology Assistant Professor +Laurens van der Maaten (http://goo.gl/O8EdtY), who gave a talk titled Visualizing Data using t-SNE, in which he describes a technique for the visualization of high dimensional data sets, developed in collaboration with Google Distinguished Researcher +Geoffrey Hinton.

Built on the concept of Nonlinear Dimensionality Reduction (http://goo.gl/aEvOAT), t-Distributed Stochastic Neighbor Embedding (t-SNE) is a new technique to embed high dimensional objects in a two dimensional map that produces substantially better results than alternative techniques, providing a new way to visualize data from computer vision and bioinformatics domains.

To learn more, watch the video below. For access to t-SNE, visit Dr. van der Maaten’s web page, linked above.

Post has attachment
I'm looking for a talented and motivated PhD student (fully funded). Please forward to potential candidates.

https://www.academictransfer.com/employer/TUD/vacancy/18819/lang/en/

Post has attachment
Owners of a Parrot AR.Drone: Please play this crowdsourcing game and help Guido de Croon at ESA to develop vision-based techniques for automatic docking of spacecraft.

http://www.esa.int/Our_Activities/Technology/Smartphone_app_turns_home_drone_into_spacecraft
Wait while more posts are being loaded