| 1. |
- Hernández-Orozco, S, et al.
(författare)
-
Algorithmic Probability-Guided Machine Learning on Non-Differentiable Spaces
- 2020
-
Ingår i: Frontiers in artificial intelligence. - : Frontiers Media SA. - 2624-8212. ; 3, s. 567356-
-
Tidskriftsartikel (refereegranskat)abstract
- We show how complexity theory can be introduced in machine learning to help bring together apparently disparate areas of current research. We show that this model-driven approach may require less training data and can potentially be more generalizable as it shows greater resilience to random attacks. In an algorithmic space the order of its element is given by its algorithmic probability, which arises naturally from computable processes. We investigate the shape of a discrete algorithmic space when performing regression or classification using a loss function parametrized by algorithmic complexity, demonstrating that the property of differentiation is not required to achieve results similar to those obtained using differentiable programming approaches such as deep learning. In doing so we use examples which enable the two approaches to be compared (small, given the computational power required for estimations of algorithmic complexity). We find and report that 1) machine learning can successfully be performed on a non-smooth surface using algorithmic complexity; 2) that solutions can be found using an algorithmic-probability classifier, establishing a bridge between a fundamentally discrete theory of computability and a fundamentally continuous mathematical theory of optimization methods; 3) a formulation of an algorithmically directed search technique in non-smooth manifolds can be defined and conducted; 4) exploitation techniques and numerical methods for algorithmic search to navigate these discrete non-differentiable spaces can be performed; in application of the (a) identification of generative rules from data observations; (b) solutions to image classification problems more resilient against pixel attacks compared to neural networks; (c) identification of equation parameters from a small data-set in the presence of noise in continuous ODE system problem, (d) classification of Boolean NK networks by (1) network topology, (2) underlying Boolean function, and (3) number of incoming edges.
|
|
| 2. |
- Laitinen, Mikko, 1973-, et al.
(författare)
-
Size matters : digital social networks and language change
- 2020
-
Ingår i: Frontiers in Artificial Intelligence. - : Frontiers Media S.A.. - 2624-8212. ; 3, s. 1-15
-
Tidskriftsartikel (refereegranskat)abstract
- Social networks play a role in language variation and change, and the social network theory has offered a powerful tool in modeling innovation diffusion. Networks are characterized by ties of varying strength which influence how novel information is accessed. It is widely held that weak-ties promote change, whereas strong ties lead to norm-enforcing communities that resist change. However, the model is primarily suited to investigate small ego networks, and its predictive power remains to be tested in large digital networks of mobile individuals. This article revisits the social network model in sociolinguistics and investigates network size as a crucial component in the theory. We specifically concentrate on whether the distinction between weak and strong ties levels in large networks over 100 nodes. The article presents two computational methods that can handle large and messy social media data and render them usable for analyzing networks, thus expanding the empirical and methodological basis from small-scale ethnographic observations. The first method aims to uncover broad quantitative patterns in data and utilizes a cohort-based approach to network size. The second is an algorithm-based approach that uses mutual interaction parameters on Twitter. Our results gained from both methods suggest that network size plays a role, and that the distinction between weak ties and slightly stronger ties levels out once the network size grows beyond roughly 120 nodes. This finding is closely similar to the findings in other fields of the study of social networks and calls for new research avenues in computational sociolinguistics.
|
|
