Publicação
Functional programming for explainable AI
| Resumo: | Neural Networks, increasingly used in Artificial Intelligence, are computational devices inspired by existing bio logical neural systems, seeking to be capable of learning how to perform tasks and recognize complex patterns. Internally, Neural Network programs are made up of several small structures called neurons (by analogy with Biology) which are responsible for handling input values in order to determine what the output values should be. The fact that these programs are organized hierarchically makes it plausible applying compositional patterns, often associated with functional programming, in order to obtain more refined neuronal networks, and whose understanding would be easier. This dissertation intends to focus more on the possibility of using the high compositionality presented in functional languages, namely Haskell, in order to make Neural Network programming better structured and elegant, facilitating not only the creation but also the understanding of what goes on inside these systems, which are sometimes seen as black boxes, due to the great lack of knowledge about how they work. |
|---|---|
| Autores principais: | Esteves, Gonçalo José Azevedo |
| Assunto: | Artificial Intelligence Neural networks Compositionality Functional programming Haskell Dissertação de mestrado Programação funcional Composicionalidade Inteligência Artificial Redes neuronais |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Neural Networks, increasingly used in Artificial Intelligence, are computational devices inspired by existing bio logical neural systems, seeking to be capable of learning how to perform tasks and recognize complex patterns. Internally, Neural Network programs are made up of several small structures called neurons (by analogy with Biology) which are responsible for handling input values in order to determine what the output values should be. The fact that these programs are organized hierarchically makes it plausible applying compositional patterns, often associated with functional programming, in order to obtain more refined neuronal networks, and whose understanding would be easier. This dissertation intends to focus more on the possibility of using the high compositionality presented in functional languages, namely Haskell, in order to make Neural Network programming better structured and elegant, facilitating not only the creation but also the understanding of what goes on inside these systems, which are sometimes seen as black boxes, due to the great lack of knowledge about how they work. |
|---|