• Corpus ID: 232092757

Learners' languages

@article{Spivak2021LearnersL,
  title={Learners' languages},
  author={David I. Spivak},
  journal={ArXiv},
  year={2021},
  volume={abs/2103.01189}
}
In “Backprop as functor”, the authors show that the fundamental elements of deep learning—gradient descent and backpropagation—can be conceptualized as a strong monoidal functor Para(Euc) → Learn from the category of parameterized Euclidean spaces to that of learners, a category developed explicitly to capture parameter update and backpropagation. It was soon realized that there is an isomorphism Learn Para(SLens), where SLens is the symmetric monoidal category of simple lenses as used in… 

Figures from this paper

Compositional Active Inference II: Polynomial Dynamics. Approximate Inference Doctrines

TLDR
The compositional theory of active inference is developed by introducing activity, functorially relating statistical games to the dynamical systems which play them, using the new notion of approximate inference doctrine, which produces systems which additionally optimize the parameters which determine their predictions.

Dynamic categories, dynamic operads: From deep learning to prediction markets

TLDR
The idea that natural organized systems adapt to internal and external pressures and this seems to happens all the way down is motivated by wanting to think clearly about this idea and so the idea is elaborated extensively in the introduction.

Backpropagation of Spirit: Hegelian Recollection and Human-A.I. Abductive Communities

This article examines types of abductive inference in Hegelian philosophy and machine learning from a formal comparative perspective and argues that Robert Brandom’s recent reconstruction of the

Data-driven dynamics and collective decision-making

In the original project, AFOSR grant FA9550-20-1-0348, we proposed a mathematical formalism for modeling, designing, and simulating very general sorts of interactions [Spi20]. The mathematical

References

SHOWING 1-8 OF 8 REFERENCES

Poly: An abundant categorical setting for mode-dependent dynamics

Dynamical systems---by which we mean machines that take time-varying input, change their state, and produce output---can be wired together to form more complex systems. Previous work has shown how to

An Invitation to Applied Category Theory

Category theory is unmatched in its ability to organize and layer abstractions and to find commonalities between structures of all sorts. No longer the exclusive preserve of pure mathematicians, it

Sheaves in geometry and logic: a first introduction to topos theory

This text presents topos theory as it has developed from the study of sheaves. Sheaves arose in geometry as coefficients for cohomology and as descriptions of the functions appropriate to various

Backprop as Functor: A compositional perspective on supervised learning

TLDR
A key contribution is the notion of request function, which provides a structural perspective on backpropagation, giving a broad generalisation of neural networks and linking it with structures from bidirectional programming and open games.

Compositional Game Theory

TLDR
The concept of coutility is introduced, which is the utility generated by an open game and returned to its environment and it is shown that a variety of games can be faithfully represented as open games in the sense of having the same Nash equilibria and off-equilibrium best responses.

Polynomial functors and polynomial monads

Abstract We study polynomial functors over locally cartesian closed categories. After setting up the basic theory, we show how polynomial functors assemble into a double category, in fact a framed

Introduction to Coalgebra

  • (cit. on p
  • 2017