Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
78 72 83 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
64 43 83 -
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
Insufficient record: 0 promises resolved, 5 needed for a score.
All types combined: 77 (likely 71 to 82, high confidence), from 84 claims: 79 factual claims, 5 predictions. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Yann LeCun's claims
55 claims
“The ConvNet architecture was loosely inspired by that of the ventral pathway in the visual cortex”
Yann LeCun, paper author 2019 ISSCC 2019 Plenary: Deep Learning Hardware: Past, Present and Future (yann.lecun.com)
The ConvNet architecture was loosely inspired by the ventral pathway in the visual cortex.
Mostly true True in substance, wrong in a detail.
What would count as true or false
True if the historical and technical literature shows that the ConvNet architecture (local receptive fields, shared weights, alternating feature-extraction and pooling/subsampling layers) took its inspiration, directly or through the Neocognitron, from Hubel and Wiesel's simple/complex cell findings and the hierarchical ventral visual stream. The claim says only 'loosely inspired', so close biological fidelity is not required. Mostly true if the inspiration is documented but traces mainly to early visual cortex (V1 simple/complex cells) rather than the whole ventral pathway, which is a minor detail. Mostly false if any biological link is incidental or added afterwards, with the design actually coming from other sources such as signal-processing convolution. False if the evidence shows that visual cortex played no part in the architecture's design.
Evidence
- Neocognitron Scholarpedia (article by Kunihiko Fukushima) · 19 January 2007
“S-cells work as feature-extracting cells. They resemble simple cells of the primary visual cortex in their response.”
- Gradient-based learning applied to document recognition HAL (open archive copy of Proceedings of the IEEE, 1998, 86(11), pp. 2278-2324) · 1998 Own record
“The convolution/subsampling combination, inspired by Hubel and Wiesel’s notions of “simple” and “complex” cells, was implemented in Fukushima’s Neocognitron [32],”
- [PDF] Convolutional Neural Networks as a Model of the Visual System arXiv (review by Grace W. Lindsay, Gatsby Computational Unit, UCL; author's final version of a Journal of Cognitive Neuroscience article)
“The convolutional neural network as we know it today comes from the field of computer vision yet the inspiration from the work of of Hubel and Wiesel is clearly visible in it (Figure 1) [10].”
- Convolutional neural network - Wikipedia Wikipedia · 31 August 2013
“Their paper replaced multiplication with convolution in time, inherently providing shift invariance, motivated by and connecting more directly to the signal-processing concept of a filter, and demonstrated it on a speech recognition task.”
Reasoning
An independent review, E3, says the inspiration from Hubel and Wiesel's work is "clearly visible" in the convolutional neural network. Fukushima's own account, E1, says the feature-extracting S-cells of the Neocognitron resemble simple cells of the primary visual cortex in their response. LeCun's 1998 paper, E2, traces the convolution/subsampling combination to Hubel and Wiesel's simple and complex cells by way of the Neocognitron. E4 notes a parallel line of work that took its motivation from signal-processing filters, and the documented inspiration points mainly to the simple and complex cells of early visual cortex (V1) rather than the whole ventral pathway, so the verdict is mostly true rather than true.
Cite this claim
Yann LeCun, “The ConvNet architecture was loosely inspired by the ventral pathway in the visual cortex.” (said 2019). Verdict: mostly true. Reliability API, claim c_b397b8df531f, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_b397b8df531f
Claims without a verdict
- Unresolvable 110
- Not yet due 3
- Not researched 309
- Not checkable 669
Is a verdict here wrong, or do you speak for Yann LeCun? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.