ClaimsYann LeCun

“Each stage in a ConvNets is composed of a filter bank, some non-linearities, and feature pooling layers.”

Yann LeCun, co-author of paper 2010 Convolutional Networks and Applications in Vision (yann.lecun.com) archived copy

Factual AI capabilities

Each stage of a ConvNet is composed of a filter bank, some non-linearities, and feature pooling layers.

Resolved true The record shows it happened or was so.

Factual: Statements about the present or the past, checked against the record: did they hold up?

The evidence is the person's or their organisation's own records.

What would count as true or false

Judged as of 2010, as a description of standard ConvNet architecture. True: the ConvNet literature of that time (for example, LeCun, Kavukcuoglu & Farabet 2010; Jarrett et al. 2009) describes each stage of a typical ConvNet as a filter bank (convolution layer), followed by one or more non-linearities (such as tanh, rectification or normalisation) and a feature pooling/subsampling layer. Mostly true: this matches the standard description, but some stages or architectures omit or reorder a component (for example, there is no pooling in the final stage). Mixed: the description fits roughly half of the common architectures of the time. Mostly false: only a minority of ConvNet stages follow this structure. False: ConvNet stages are not generally made of these components.

Set before the research began.

Evidence (4)

  1. E1 [PDF] What is the Best Multi-Stage Architecture for Object Recognition? Courant Institute of Mathematical Sciences, New York University (PDF hosted on yann.lecun.com) Own record

    “In many recent object recognition systems, feature extraction stages are generally composed of a filter bank, a non-linear transformation, and some sort of feature pooling layer.”

  2. E2 [PDF] What is the Best Multi-Stage Architecture for Object Recognition? Courant Institute of Mathematical Sciences, New York University (PDF hosted on yann.lecun.com) Own record

    “This is the basic building block of traditional convolutional networks, alternating tanh-squashed filter banks with average down-sampling layers [14, 10]. A complete convolutional network would have several sequences of “FCSG - PA” followed by by a linear classifier.”

  3. E3 [PDF] What is the Best Multi-Stage Architecture for Object Recognition? Courant Institute of Mathematical Sciences, New York University (PDF hosted on yann.lecun.com) Own record

    “Several authors have proposed systems with two stages of learned feature extractors, each of which comprises filter banks, non-linearities, and pooling.”

  4. E4 [PDF] What is the Best Multi-Stage Architecture for Object Recognition? Courant Institute of Mathematical Sciences, New York University (PDF hosted on yann.lecun.com) Own record

    “FCSG − PM This is also a typical building block of convolutional networks, as well as the basis of the HMAX and other architectures [28, 25], which alternate tanh-squashed filter banks with max-pooling layers.”

Reasoning

The resolution criteria name Jarrett et al. 2009 as reference literature. In that paper, feature extraction stages are described as generally made of a filter bank, a non-linear transformation and a feature pooling layer (E1). The same paper calls tanh-squashed filter banks followed by average or max pooling the basic, typical building block of convolutional networks, with several such stages followed by a classifier (E2, E4). It also says other authors' two-stage systems each comprise filter banks, non-linearities and pooling (E3). These sources match the claimed description of a standard ConvNet stage, but all of them come from a single paper hosted on the subject's website and marked as not independent of the subject.

Cite this claim

Yann LeCun, “Each stage of a ConvNet is composed of a filter bank, some non-linearities, and feature pooling layers.” (said 2010). Verdict: resolved true. Reliability API, claim c_45f18100ad01, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_45f18100ad01

This claim is in dataset release v1.1.2, with the same verdict.

Reply to this claim

For Yann LeCun, or someone speaking for them: a correction or counter-evidence. Replies are read before anything is shown; an approved reply shows your name, role, date, statement and link. Your email address is never shown and is deleted 30 days after the decision (privacy).

More factual claims by Yann LeCun

  1. Mostly true Factual 18 March 2024
    “For starters, transformers are equivariant to permutations.”

    Transformers are equivariant to permutations, as ConvNets are equivariant to translations.

  2. Mostly true Factual 14 March 2024
    “If it were the case, we would have AI systems that could teach themselves to drive a car in 20 hours of practice, like any 17 year-old.”

    There are currently no AI systems that can teach themselves to drive a car in about 20 hours of practice, as a 17-year-old can.

  3. Mostly true Factual 8 March 2024
    “The architecture is good, the architecture of the encoder is good, but the fact that you train the system to reconstruct images does not lead it to produce to long, good generic features of images.”

    Training an encoder to reconstruct images does not lead it to produce good generic image features, though the encoder architecture is good.

  4. Mostly true Factual 8 March 2024
    “LLMs can do none of those or they can only do them in a very primitive way and they don’t really understand the physical world.”

    Current LLMs can do none of understanding the physical world, persistent memory, reasoning and planning, or only in a very primitive way.

  5. Mostly true Factual 8 March 2024
    “No, there’s one thing that autoregressive LLMs or that LLMs in general, not just the autoregressive one, but including the bird style bidirectional ones, are exploiting and its self supervised…”

    LLMs, including BERT-style bidirectional ones, exploit self-supervised learning, and LeCun has been a very strong advocate of self-supervised learning for many years.

All of them

Verdicts are made automatically; how verdicts are made.