Papers by Louk Smalbil
The purpose of this paper is to understand what delimits computation. The classical notion of com... more The purpose of this paper is to understand what delimits computation. The classical notion of computation formulated in the Church-Turing Thesis seems to suggest that computability is a mathematical concept. Physicists David Deutsch points out that there is a physical assertion to this claim. We will discuss this claim and its consequences. Moreover, we will discuss what happens when the boundaries of what it computable get stretched too far. In that respect, we will look at hypercompuation. Lastly, I will discuss a paper by Andrew Hodges in which he criticises hypercomputation. In the end, we will answer the question as to what extent the limits of computation are physical and how the mathematical aspects of computation as developed by Church and Turing relate to the concept's physical limitations.
Identifying causal e ects from observational data is a challenging task. Nevertheless, the burgeo... more Identifying causal e ects from observational data is a challenging task. Nevertheless, the burgeoning field of causal discovery has, over the last two decades, produced a wide variety of algorithms aimed at recovering the causal substructures from data. Because of the complexity of the problems, however, these algorithms require a lot of strong assumptions such as causal faithfulness or the absence of hidden variables. For this reason, recent research has tried to overcome these limiting assumptions by exploiting the power of generative neural networks in the context of causal learning. Promising research has been done in regards to recovering causal directed acyclic graphs (DAGs) from observational data using generative adversarial networks. Other studies sought to exploit the power of variational autoencoders (VAEs) to find the per-variable e ect (i.e. does X cause Y). However, VAEs have not been used to uncover entire causal models. In this study, we will contribute to this field of research by demonstrating that a VAEbased architecture can be used to learn causal DAGs. In order to do so, we propose Causal Discovery Variational Auto-encoder (CDVAE), an architecture tasked with recovering DAGs from observational data. We will show that our method often performs significantly better than existing non-neural network based approaches. Moreover, we show that a latent variable model is a robust methodology even in the case of hidden variables.
Proceedings of the 12th International Conference on Natural Language Generation
A prominent strand of work in formal semantics investigates the ways in which human languages qua... more A prominent strand of work in formal semantics investigates the ways in which human languages quantify over the elements of a set, as when we say "All A are B", "All except two A are B", "Only a few of the A are B" and so on. Our aim is to build Natural Language Generation algorithms that mimic humans' use of quantified expressions. To inform these algorithms, we conducted on a series of elicitation experiments in which human speakers were asked to perform a linguistic task that invites the use of quantified expressions. We discuss how these experiments were conducted and what corpora they gave rise to. We conduct an informal analysis of the corpora, and offer an initial assessment of the challenges that these corpora pose for Natural Language Generation. The dataset is available at: https: //github.com/a-quei/qtuna.
Thesis Chapters by Louk Smalbil
In the Image of Men: Towards a Philosophy of Machine Learning, 2020
The field of machine learning has seen some major developments in the past decades. As a field wi... more The field of machine learning has seen some major developments in the past decades. As a field with such a large impact on today’s society, technologies and scientific methodologies, there is surprisingly little philosophical inquiry as to what ML is, what it could do and how ML as a technology relates to us as humans. In this thesis, therefore, the aim is to outline the philosophical aspects of ML. Moreover, in doing so, an outline of the philosophical underpinnings of ML is given in which not only the inferential and inductive issues are taken into consideration, but also the issues stemming from the data and the way the data is shaped by humans. It is argued that even though ML is indeed a veritable process of knowledge-making, it is di↵erent from knowledge-making as done by humans. As our algorithms are products of human knowledge-making and therefore a derivation, ML should be seen as a second-order knowledge-maker.
Drafts by Louk Smalbil
Davidson's Formal Hermeneutics: An inquiry into the reconcilability of Davidson’s early and later works, 2017
''… I often suspect that all the really breath-taking views for which Davidson has become famous ... more ''… I often suspect that all the really breath-taking views for which Davidson has become famous can be both defended and understood without reference to, or knowledge of, the project of developing a Tarskian truth-theory for a natural language.'' 3
Teaching Documents by Louk Smalbil
The Limits of Computation, 2020
The purpose of this paper is to understand what delimits computation. The classical notion of com... more The purpose of this paper is to understand what delimits computation. The classical notion of computation formulated in the Church-Turing Thesis seems to suggest that computability is a mathematical concept. Physicist David Deutsch points out that there is a physical assertion to this claim. We will discuss this claim and its consequences. Moreover, we will discuss what happens when the boundaries of what is computable get stretched too far. In that respect, we will look at hypercomputation. Lastly, we will discuss a paper by Andrew Hodges in which he criticises hypercomputation. In the end, we will answer the question as to what extent the limits of computation are physical and how the mathematical aspects of computation as developed by Church and Turing relate to the concept's physical limitations.
Uploads
Papers by Louk Smalbil
Thesis Chapters by Louk Smalbil
Drafts by Louk Smalbil
Teaching Documents by Louk Smalbil