A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation
While many models of biological object recognition share a common set of ‘‘broad-stroke’’ properties, the performance of any one model depends strongly on the choice of parameters in a particular instantiation of that model—e.g., the number of units per layer, the size of pooling kernels, exponent...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | en_US |
| Published: |
Public Library of Science
2010
|
| Online Access: | http://hdl.handle.net/1721.1/55377 https://orcid.org/0000-0002-1592-5896 https://orcid.org/0000-0002-2189-9743 |
| _version_ | 1826204656823435264 |
|---|---|
| author | Cox, David D. Pinto, Nicolas Doukhan, David DiCarlo, James |
| author2 | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences |
| author_facet | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences Cox, David D. Pinto, Nicolas Doukhan, David DiCarlo, James |
| author_sort | Cox, David D. |
| collection | MIT |
| description | While many models of biological object recognition share a common set of ‘‘broad-stroke’’ properties, the performance of
any one model depends strongly on the choice of parameters in a particular instantiation of that model—e.g., the number
of units per layer, the size of pooling kernels, exponents in normalization operations, etc. Since the number of such
parameters (explicit or implicit) is typically large and the computational cost of evaluating one particular parameter set is
high, the space of possible model instantiations goes largely unexplored. Thus, when a model fails to approach the abilities
of biological visual systems, we are left uncertain whether this failure is because we are missing a fundamental idea or
because the correct ‘‘parts’’ have not been tuned correctly, assembled at sufficient scale, or provided with enough training.
Here, we present a high-throughput approach to the exploration of such parameter sets, leveraging recent advances in
stream processing hardware (high-end NVIDIA graphic cards and the PlayStation 3’s IBM Cell Processor). In analogy to highthroughput
screening approaches in molecular biology and genetics, we explored thousands of potential network
architectures and parameter instantiations, screening those that show promising object recognition performance for further
analysis. We show that this approach can yield significant, reproducible gains in performance across an array of basic object
recognition tasks, consistently outperforming a variety of state-of-the-art purpose-built vision systems from the literature.
As the scale of available computational power continues to expand, we argue that this approach has the potential to greatly
accelerate progress in both artificial vision and our understanding of the computational underpinning of biological vision. |
| first_indexed | 2024-09-23T12:58:54Z |
| format | Article |
| id | mit-1721.1/55377 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T12:58:54Z |
| publishDate | 2010 |
| publisher | Public Library of Science |
| record_format | dspace |
| spelling | mit-1721.1/553772022-09-28T11:18:02Z A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation Cox, David D. Pinto, Nicolas Doukhan, David DiCarlo, James Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences McGovern Institute for Brain Research at MIT DiCarlo, James Cox, David D. Pinto, Nicolas Doukhan, David DiCarlo, James While many models of biological object recognition share a common set of ‘‘broad-stroke’’ properties, the performance of any one model depends strongly on the choice of parameters in a particular instantiation of that model—e.g., the number of units per layer, the size of pooling kernels, exponents in normalization operations, etc. Since the number of such parameters (explicit or implicit) is typically large and the computational cost of evaluating one particular parameter set is high, the space of possible model instantiations goes largely unexplored. Thus, when a model fails to approach the abilities of biological visual systems, we are left uncertain whether this failure is because we are missing a fundamental idea or because the correct ‘‘parts’’ have not been tuned correctly, assembled at sufficient scale, or provided with enough training. Here, we present a high-throughput approach to the exploration of such parameter sets, leveraging recent advances in stream processing hardware (high-end NVIDIA graphic cards and the PlayStation 3’s IBM Cell Processor). In analogy to highthroughput screening approaches in molecular biology and genetics, we explored thousands of potential network architectures and parameter instantiations, screening those that show promising object recognition performance for further analysis. We show that this approach can yield significant, reproducible gains in performance across an array of basic object recognition tasks, consistently outperforming a variety of state-of-the-art purpose-built vision systems from the literature. As the scale of available computational power continues to expand, we argue that this approach has the potential to greatly accelerate progress in both artificial vision and our understanding of the computational underpinning of biological vision. Dr. Gerald Burnett and Marjorie Burnett McKnight Endowment for Neuroscience Rowland Institute at Harvard National Institutes of Health (U.S.) (NEI R01EY014970) 2010-06-03T15:20:08Z 2010-06-03T15:20:08Z 2009-11 2009-06 Article http://purl.org/eprint/type/JournalArticle 1553-7358 553-734X http://hdl.handle.net/1721.1/55377 Pinto, Nicolas et al. “A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation.” PLoS Comput Biol 5.11 (2009): e1000579. © 2009 Pinto et al. https://orcid.org/0000-0002-1592-5896 https://orcid.org/0000-0002-2189-9743 en_US http://dx.doi.org/10.1371/journal.pcbi.1000579 PLoS Computational Biology Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. application/pdf Public Library of Science PLoS |
| spellingShingle | Cox, David D. Pinto, Nicolas Doukhan, David DiCarlo, James A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation |
| title | A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation |
| title_full | A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation |
| title_fullStr | A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation |
| title_full_unstemmed | A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation |
| title_short | A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual Representation |
| title_sort | high throughput screening approach to discovering good forms of biologically inspired visual representation |
| url | http://hdl.handle.net/1721.1/55377 https://orcid.org/0000-0002-1592-5896 https://orcid.org/0000-0002-2189-9743 |
| work_keys_str_mv | AT coxdavidd ahighthroughputscreeningapproachtodiscoveringgoodformsofbiologicallyinspiredvisualrepresentation AT pintonicolas ahighthroughputscreeningapproachtodiscoveringgoodformsofbiologicallyinspiredvisualrepresentation AT doukhandavid ahighthroughputscreeningapproachtodiscoveringgoodformsofbiologicallyinspiredvisualrepresentation AT dicarlojames ahighthroughputscreeningapproachtodiscoveringgoodformsofbiologicallyinspiredvisualrepresentation AT coxdavidd highthroughputscreeningapproachtodiscoveringgoodformsofbiologicallyinspiredvisualrepresentation AT pintonicolas highthroughputscreeningapproachtodiscoveringgoodformsofbiologicallyinspiredvisualrepresentation AT doukhandavid highthroughputscreeningapproachtodiscoveringgoodformsofbiologicallyinspiredvisualrepresentation AT dicarlojames highthroughputscreeningapproachtodiscoveringgoodformsofbiologicallyinspiredvisualrepresentation |