NIPS 2011 - NIPS 2011 Workshop on Interpretable Decoding of Higher Cognitive States from Neural Data

NIPS 2011 Workshop on Interpretable Decoding of Higher Cognitive States from Neural Data

Dec 16 -17, 2011, Granada, Spain

Overview

Over recent years, machine learning methods have become a crucial analytical tool in cognitive neuroscience (see reviews by Formisano et al., 2008; Pereira et al., 2009). Decoding techniques have dramatically increased the sensitivity of experiments, and so also the subtlety of cognitive questions that can be asked. At the same time the mental phenomena being studied are moving beyond lower-level perceptual and motor processes which are directly grounded in external measurable realities.

Decoding higher cognition and interpreting the learned behaviour of the classifiers used pose unique challenges, as these psychological states are complex, fast-changing and often ill-defined. Contemporary machine learning methods deal well with the small numbers of cases, and high numbers of co-linear dimensions typical of neural data, and are generally optimized to maximize classification performance, rather than to enable meaningful interpretation of the features they learn from. And indeed recent work has succeeded to decode psychological phenomena including visual object recognition (e.g. Kriegeskorte et al., 2008; Connolly et al., 2011), perceptual interpretation of sounds (Staeren et al., 2009), lexical semantics (Mitchell et al., 2008; Simanova et al., 2010; Devereux et al., 2010; Murphy et al., 2011), decision making during game playing (Xiang et al., 2009) and the process of mental arithmetic (Anderson et al., 2008). But for the cognitive scientists who use these methods, the primary question is often not "how much" but rather "how" and "why" the patterns of neural activity identified by a machine learning algorithm encode particular cognitive processes.

The aim of this workshop is therefore to 1) discuss the achievements and problems of the decoding of high-level cognitive states, and 2) explore the use of machine learning methodologies and other computational models that enable such cognitive interpretation of neural recordings of different modalities. Advances in this field require close collaboration between machine learning experts, neuroscientists and cognitive scientists. Thus, this workshop is highly interdisciplinary and will aim to attract submissions also from outside the existing NIPS community. By stimulating discussions among experts in the different fields, the workshop seeks to generate novel insights and new directions for research.

Topics of interest

The field requires techniques that are capable of taking advantage of spatially distributed patterns in the brain, that are separated in space but coordinated in their activity. Methods should also be sensitive to the fine-grained temporal patterns of multiple processes - which may proceed in a serial fashion, overlapping or in parallel with each-other, or in multiple passes with bidirectional information flows. Different recording modalities have distinctive advantages: fMRI provides very fine millimetre-level localisation in the brain but poor temporal resolution, while EEG and MEG have millisecond temporal resolution at the cost of spatial resolution. Ideally machine learning methods would be able to meaningfully combine complementary information from these different neuroimaging techniques (see e.g. De Martino et al., 2010). Moreover, as the processes underlying higher cognition are so complex, methods should be able to disentangle even tightly linked and confounded subprocesses. Finally, general use algorithms that could induce latent dimensions from neural data, and so reveal the "hidden" psychological states, would be a dramatic advance on current hypothesis-driven analytical paradigms. Originality of approach is encouraged and submissions on any related methodological approach are welcomed, such as:

- Interpreting spatial and temporal location of selected features and their weights
- Discovering "hidden" or "latent" cognitive representations
- Disentangling confounded processes and representations
- Comparing or combining data from recording modalities (e.g. fMRI, EEG, structural MRI, DTI, MEG, NIRS, EcOG, single cell recordings)
- Fuzzy and partial classifications
- Unaligned or incommensurate feature spaces and data representation

As noted above, the complexity of higher cognition poses challenges. To take language comprehension as an example, speech is received at 3-4 words; acoustic, semantic and syntactic processing can occur in parallel; and the form of underlying representations (sentence structures, conceptual descriptions) remains controversial. We welcome submissions dealing with any high-level cognitive functions that exhibit similar complexity, for instance:

- Knowledge representation and concepts
- Language and communication
- Understanding visual and auditory experience
- Memory and learning
- Reasoning and problem solving
- Decision making and executive control

Submissions

Authors are invited to submit full papers on original, unpublished work in the topic area of this workshop via the NIPS 2011 submission site at https://cmt.research.microsoft.com/NIPS2011/Defaul.... Submissions should be formatted using the NIPS 2011 stylefiles, with blind review and not exceeding 8 pages plus an extra page for references. Author and submission information can be found at http://nips.cc/PaperInformation/AuthorSubmissionIn.... The stylefiles are available at http://nips.cc/PaperInformation/StyleFiles. Each submission will be reviewed at least by two members of the programme committee. Accepted papers will be published in the workshop proceedings. Dual submissions to the main NIPS 2011 conference and this workshop are allowed; if you submit to the main session, indicate this when you submit to the workshop. If your paper is accepted for the main session, you should withdraw your paper from the workshop upon notification by the main session.

Important Dates

- Aug 30, 2011: Call for papers
- Sep 23, 2011: Deadline for submission of workshop papers
- Oct 15, 2011: Notification of acceptance
- Oct 31, 2011: Camera-ready papers due
- Dec 16 - 17, 2011: Workshop date

Organisers

The organizing committee are researchers who are all directly involved in machine learning of higher cognitive states, and have previous experience running similarly themed interdisciplinary workshops, including the NAACL Workshop on Computational Neurolinguistics (2010), ICCS Symposium on Neural Decoding of Higher Cognitive States (2010), the CAOS Special Session on Computational Approaches to the Neuroscience of Concepts (2010).

- Kai-min Kevin Chang, Language Technologies Institute & Centre for Cognitive Brain Imaging, Carnegie Mellon University
- Anna Korhonen, Computer Laboratory & Research Centre for English and Applied Linguistics, University of Cambridge
- Brian Murphy, Computation, Language and Interaction Group, Centre for Mind/Brain Sciences, University of Trento
- Irina Simanova, Max Planck Institute for Psycholinguistics & Donders Institute for Brain, Cognition and Behaviour, Nijmegen

Invited speakers

- Elia Formisano, Universiteit Maastricht, Netherlands
- Francisco Pereira, Princeton University, USA (provisional)

Programme committee

The preliminary programme comittee listing is given below, and includes leading researchers in a range of fields covering machine learning, neuroscience and wider cognitive sciences:

- John Anderson, Carnegie Mellon University, USA
- Yi Chen, Max-Planck Institute for Human Cognitive and Brain Sciences Leipzig, Germany
- Mark Cohen, University of California Los Angeles, USA
- Kevyn Collins-Thompson, Microsoft Research, USA
- Andy Connolly, Dartmouth College, USA
- Jack Gallant, University of California Berkeley, USA
- Marcel van Gerven, Radboud University Nijmegen, Netherlands
- Michael Hanke, Dartmouth College, USA
- Jim Haxby, Dartmouth College, USA & University of Trento, Italy
- Tom Heskes, Radboud University Nijmegen, Netherlands
- Mark Johnson, Macquarie University, Australia
- Marius Peelen, University of Trento, Italy
- Francisco Pereira, Princeton University, USA
- Russ Poldrack, University of Texas Austin, USA
- Dean Pomerleau, Intel Labs Pittsburgh, USA
- Diego Sona, Fondazione Bruno Kessler, Italy

References

- Anderson, J. R., Carter, C. S., Fincham, J. M., Qin,. Y., Ravizza, S. M., and Rosenberg-Lee, M. (2008). Using fMRI to Test Models of Complex Cognition. Cognitive Science, 32, 1323-1348.
- Connolly, A. C., Guntupalli, J. S., Gors, J., Hanke, M., Halchenko, Y. O., Wu, Y., Abdi, H. and Haxby, J. V. (Submitted). Representation of biological classes in the human brain.
- De Martino F., Valente G., de Borst A. W., Esposito F., Roebroeck A., Goebel R., Formisano E. (2010). Multimodal imaging: an evaluation of univariate and multivariate
methods for simultaneous EEG/fMRI. Magn Reson Imaging. 28(8), 1104-12.
- Devereux, B., Kelly, C., and Korhonen, A. (2010). Using fMRI Activation to Conceptual Stimuli to Evaluate Methods for Extracting Conceptual Representations from Corpora. Proceedings of the NAACL-HLT Workshop on Computational Neurolinguistics.
- Formisano E., De Martino F., Valente G. (2008). Multivariate analysis of fMRI time series: classification and regression of brain responses using machine learning. Magn Reson Imaging, 26(7), 921-34.
- Kriegeskorte, N., Mur, M., Ruff, D., Kiani, R., Bodurka, J., Esteky, H., Tanaka, K., and Bandettini, P. (2008). Matching categorical object representations in inferior temporal cortex of man and monkey. Neuron, 60(6), 1126-1141.
- Mitchell, T. M., Shinkareva, S. V., Carlson, A., Chang, K. M., Malave, V. L., Mason, R. A., and Just, M. A. (2008). Predicting Human Brain Activity Associated with the Meanings of Nouns. Science, 320, 1191-1195.
- Murphy, B., Poesio, M., Bovolo, F., Bruzzone, L., Dalponte, M., and Lakany, H. (2011). EEG decoding of semantic category reveals distributed representations for single concepts. Brain and Language, 117, 12-22.
- Pereira F., Mitchell T., Botvinick M. (2009). Machine learning classifiers and fMRI: a tutorial overview. Neuroimage. 45(1 Suppl) S199-209.
- Simanova, I., Van Gerven, M., Oostenveld, R., and Hagoort, P. (2010). Identifying object categories from event-related EEG: Toward decoding of conceptual representations. Plos One, 512, E14465.
- Staeren N., Renvall H., De Martino F., Goebel R., Formisano E. (2009). Sound categories are represented as distributed patterns in the human auditory cortex. Curr Biol, 19(6), 498-502.
- Xiang, J. and Chen, J. and Zhou, H. and Qin, Y. and Li, K. and Zhong, N. 2009: Using SVM to predict high-level cognition from fMRI data: a case study of 4* 4 Sudoku solving. Brain Informatics, 171-181.

Links

- NIPS 2011 website: http://nips.cc/Conferences/2011/
- Workshop website: https://sites.google.com/site/decodehighcogstate
- Call for Papers: https://sites.google.com/site/decodehighcogstate/c...
(Please feel free to distribute the CFP to all the interested persons and groups.)

--
Kai-min Kevin Chang
http://www.cs.cmu.edu/~kkchang/