Prof. Lara will deliver a keynote at the upcoming The 7th Annual Henry Taub TCE Conference held at Technion, Haifa, Israel.
Prof. Lara Dolecek’s comprehensive survey on channel coding methods for non-volatile memories will appear in an upcoming issue of the Proceedings of the IEEE. Paper is
L. Dolecek and Y. Cassuto, “Coding for non-volatile memories: recent advances and new opportunities”.
Proceedings of the IEEE is the most highly-cited general interest journal in electrical engineering and computer science: link here
Prof. Lara Dolecek will be giving an invited talk at Caltech on some of our recent research early next month.
Prof. Lara Dolecek presented a tutorial on coding for error-efficient computing at the SNF/EPFL-sponsored International Exploratory Workshop on
Theory and Practice for Error-Efficient Computing Systems.
Workshop Info: http://people.csail.mit.edu/psm/tpeecs-2017/
We had 2 oral presentations at the Non-Volatile Memories Workshop:
Ahmed Hareedy presented the work on “Non-Binary LDPC Code Optimization for Modern Storage Systems,” by A. Hareedy, C. Lanka, and L. Dolecek.
Shahroze Kabir presented the work on “Modeling the effects of radiation induced soft errors on LDPC decoding,” by F. Sala, C. Schoeny, S. Kabir, D. Divsalar, and L. Dolecek.
Prof. Lara gives an invited talk at ITA on our group’s research on reliable inference under noisy computations and implications on radiation-hardened LDPC decoders.
Ahmed Hareedy has been selected to present his PhD research in the esteemed Graduation Day session at ITA 2017. For more information, see here http://ita.ucsd.edu/workshop/17/talks.php#talk2
Prof. Lara will give a plenary talk on coding for memories at the upcoming annual meeting of the American Mathematical Society to be held in the first week of January in Atlanta, GA.
LORIS has two papers at Asilomar Conference on Signals, Systems, and Computers. The first paper is an invited contribution on statistical inference on noisy hardware. It discusses LDPC decoders in Flash memories used in high radiation environments. The second paper is on combinatorial characterization of finite length spatially coupled codes with applications to magnetic recording applications.
- F. Sala, C. Schoeny, S. Kabir, D. Divsalar, and L. Dolecek, “Flash Memories in High Radiation Environments: LDPC Decoder Study,” in Proc. IEEE Asilomar Conference on Signals, Systems, and Computers, Nov. 2016 (invited).
- H. Esfahanizadeh, A. Hareedy, and L. Dolecek, “Spatially Coupled LDPC Codes Optimized for 1-D Magnetic Recording Channels,” in Proc. IEEE Asilomar Conference on Signals, Systems, and Computers, Nov. 2016.
In collaboration with Prof. Van den Broeck (CS, UCLA), we have an exciting new research on using coding methods to make ML algorithms more robust. Initial results from this work were presented at 2016 Allerton Conference.
|Robust Channel Coding Strategies for Machine Learning Data (I)|
|Van den Broeck, Guy||UCLA|
|Dolecek, Lara||Univ. of California, Los Angeles|
Keywords: Coding Theory, Data Analytics, Machine Learning and Learning Theory
Abstract: Two important recent trends are the proliferation of learning algorithms along with the massive increase of data stored on unreliable storage mediums. These trends impact each other; noisy data can have an undesirable effect on the results provided by learning algorithms. Although traditional tools exist to improve the reliability of data storage devices, these tools operate at a different abstraction level and therefore ignore the data application, leading to an inefficient use of resources. In this paper we propose taking the operation of learning algorithms into account when deciding how to best protect data. Specifically, we examine several learning algorithms that operate on data that is stored on noisy mediums and protected by error-correcting codes with a limited budget of redundancy; we develop a principled way to allocate resources so that the harm on the output of the learning algorithm is minimized.