Exploiting structure in data is crucial for the success of many techniques in neuroscience, machine learning, signal processing, and statistics.  In this context, the fact that data of interest can be modeled via sparsity has been proven extremely valuable. As a consequence, numerous algorithms either aiming at learning sparse representations of data, or exploiting sparse representations in applications have been proposed within the machine learning and signal processing communities over the last few years.

This workshop aims at highlighting the differences, commonalities, advantages and disadvantages of the analysis and synthesis data models. The workshop will provide a venue for discussing pros and cons of the two approaches in terms of scalability, ease of learning, and most importantly, applicability to problems in machine learning such as classification, recognition, data completion, source separation, etc. The targeted group of participants ranges from researchers in machine learning and signal processing to mathematicians. All participants of the workshop will gain a deeper understanding of the duality of the two approaches for modeling data and a clear view of which model suits best for certain applications. Moreover, further research directions will be identified that adress the issue of usability of the analysis operator approach for problems arising in Machine Learning as well as important theoretical questions related to the connection of the two fraternal twins in sparse modeling.

Topics of Interest

  • Dictionary Learning 
  • Analysis Operator Learning 
  • Joint Learning and Classification/Recognition 
  • Task Oriented Learning of Sparse Representations 
  • Theory of Analysis Operators and Dictionaries 
  • Sparse Models for Data Completion 
  • Multimodal Dictionary/Analysis Operator Learning 
  • Optimization for Learning Dictionaries and Analysis Operators

General Information

Invited Speakers

  • Michael Elad Technion, Tel Aviv, Israel
  • Yann LeCun New York University, New York, USA
  • Lawrence Carin Duke University, Durham NC, USA 
  • Yi Ma Microsoft Research Asia, Beijing, China 
  • Bruno A. Olshausen UC Berkeley,  Berkeley, USA