Skip to content
 

Lame Statistics Patents

Manoel Galdino wrote in a comment off-topic on another post (which I erased):

I know you commented before about patents on statistical methods. Did you know this patent (http://www.archpatent.com/patents/8032473)? Do you have any comment on patents that don’t describe mathematically how it works and how and if they’re any different from previous methods? And what about the lack of scientific validation of the claims in such a method?

The patent in question, “US 8032473: “Generalized reduced error logistic regression method,” begins with the following “claim”:

A system for machine learning comprising: a computer including a computer-readable medium having software stored thereon that, when executed by said computer, performs a method comprising the steps of being trained to learn a logistic regression match to a target class variable so to exhibit classification learning by which: an estimated error in each variable’s moment in the logistic regression be modeled and reduced through constraints that require that the expected extreme error be inversely related to a t-value for that variable; an estimated error in each variable’s moment in the logistic regression be modeled and reduced through constraints that require that the probability of positive and negative estimated errors be substantially equal across all variable moments; where there is substantially no bias in the probability of positive or negative estimated errors across even versus odd polynomial moments; and, an estimated error in each variable’s moment in the logistic regression is constrained by a scaling that is not the sum of t-values across all variables but instead is substantially twice that sum so to reflect both positive and negative expected errors whereby when this substantially twice sum value is divided by the t-value for any variable, it yields a large expected error for small t-values and a small expected error for large t-values.

I have no idea what Andrew’s take on patents is, but my own experience in computer science is that you can patent just about anything with enough patience. There’s no “scientific validation” component to the patent process in the sense that you need peer-reviewed citations (not that you should trust peer review any more than patent-office review). There’s supposed to be a novelty component in that you’re only supposed to be able to patent something that’s not obvious to someone “skilled in the art.” The problem is that they don’t assume much “skill” in this judgment given the obvious things people patent. One of my personal favorites is US 6192338, an AT&T patent involving connecting a speech recognizer to a database over the network.

3 Comments

  1. Rahul says:

    Lame-Statistics Patents or Lame Statistics-Patents?

    • I meant the second, but it could clearly be used either way.

      Of course, there’s no hyphen for either in English typesetting. Copy editors will have you hyphenate noun compounds if and only if they’re (a) used as nominal modifiers and (b) not “very common”. So “towel-rack designer” and “designer of towel racks” are kosher, but “designer of towel-racks” is not (but “designer of towel-rack fasteners” is — go figure).

  2. Fernando says:

    Here is an idea for Bayesians out there: patent p-values, refuse to give out licenses, and sue anyone who uses them.