Reverse derivative categories

The reverse derivative is a fundamental operation in machine learning and automatic differentiation. This paper gives a direct axiomatization of a category with a reverse derivative operation, in a similar style to that given by Cartesian differential categories for a forward derivative. Intriguingly, a category with a reverse derivative also has a forward derivative, but the converse is not true. In fact, we show explicitly what a forward derivative is missing: a reverse derivative is equivalent to a forward derivative with a dagger structure on its subcategory of linear maps. Furthermore, we show that these linear maps form an additively enriched category with dagger biproducts.

[1]  J.R.B. Cockett,et al.  DIFFERENTIAL RESTRICTION CATEGORIES , 2012, 1208.4068.

[2]  Richard Blute,et al.  CARTESIAN DIFFERENTIAL CATEGORIES , 2009 .

[3]  Bruce Christianson A Leibniz Notation for Automatic Differentiation , 2012 .

[4]  Sally Popkorn,et al.  A Handbook of Categorical Algebra , 2009 .

[5]  J. Robin B. Cockett,et al.  Differential categories , 2006, Mathematical Structures in Computer Science.

[6]  The dual fibration in elementary terms , 2015, 1501.01947.

[7]  R. Blute,et al.  CARTESIAN DIFFERENTIAL STORAGE CATEGORIES , 2014, 1405.6973.

[8]  J. Robin B. Cockett,et al.  Differential Structure, Tangent Structure, and SDG , 2014, Appl. Categorical Struct..

[9]  Conal Elliott,et al.  The simple essence of automatic differentiation , 2018, Proc. ACM Program. Lang..

[10]  Andreas Griewank,et al.  Who Invented the Reverse Mode of Differentiation , 2012 .

[11]  Jean-Simon Lemay A Tangent Category Alternative to the Fa\`a di Bruno Construction , 2018 .

[12]  S. Linnainmaa Taylor expansion of the accumulated rounding error , 1976 .

[13]  J. Robin B. Cockett,et al.  Restriction categories I: categories of partial maps , 2002, Theor. Comput. Sci..

[14]  Bart Jacobs,et al.  Categorical Logic and Type Theory , 2001, Studies in logic and the foundations of mathematics.

[15]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.

[16]  Peter Selinger,et al.  Dagger Compact Closed Categories and Completely Positive Maps: (Extended Abstract) , 2007, QPL.