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
110 claims · page 2 of 2
“Thus we conclude that there exists at least one point in HP1 at which the loss is less than at all points in HP2.”
Yann LeCun et al., tutorial authors 2006 A Tutorial on Energy-Based Learning (yann.lecun.com)
There exists at least one point in HP1 at which the loss is less than at all points in HP2.
Unresolvable The public record could not settle it.
What would count as true or false
Evaluate the claim as a mathematical conclusion in the 2006 source paper, where HP1 and HP2 are the regions or sets it defines. TRUE: under the paper's stated assumptions and its definition of the loss (negative log-likelihood, if that is the loss the paper uses), the existence of at least one point in HP1 with loss lower than at every point in HP2 follows validly. MOSTLY TRUE: the conclusion is valid, but the proposition misnames a minor detail (e.g., the loss is a closely related one rather than exactly the negative log-likelihood). MOSTLY FALSE: the conclusion holds only under further assumptions the paper does not state, or holds for a weaker statement (e.g., less than or equal, or less than some but not all points in HP2). FALSE: the reasoning is invalid or a counterexample within the paper's setting is shown.
Evidence
- Factor Graphs for Relational Regression New York University, Department of Computer Science (PhD dissertation by Sumit Prakash Chopra) · January 2009 Own record
“Let the two half planes EC + m < EI and EC + m ≥ EI be denoted by HP1 and HP2 respectively.”
- Factor Graphs for Relational Regression New York University, Department of Computer Science (PhD dissertation by Sumit Prakash Chopra) · January 2009 Own record
“Lnll(W, Y i, X i) = E(W, Y i, X i) + Fβ(W, Y, X i).”
- Wake-Sleep Energy Based Models for Continual Learning CVF Open Access (CVPR 2024 Workshops, CLVision) · 2024
“Therefore it can be concluded that there exists at least one point in P1 at which the loss is less than at all points in P2.”
Reasoning
E2 defines HP1 (EC + m < EI) and HP2 (EC + m ≥ EI), and E3 gives the formula for the negative log-likelihood loss in a 2009 dissertation. Both come from a 2009 dissertation that is not independent of the subject, and neither shows the argument that leads to the conclusion. E4 is an independent paper that restates the same conclusion for its own regions P1 and P2. It does not show the derivation or the assumptions it depends on, or whether those assumptions hold for the negative log-likelihood loss. The evidence therefore does not establish whether the conclusion follows validly under the 2006 paper's assumptions, as the criteria require.
Cite this claim
Yann LeCun, “There exists at least one point in HP1 at which the loss is less than at all points in HP2.” (said 2006). Verdict: unresolvable. Reliability API, claim c_ea7fbc4d08d2, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_ea7fbc4d08d2
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.