| 1. |
- Andersson, Patrick, et al.
(författare)
-
Exploration strategies for homeostatic agents
- 2019
-
Ingår i: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). - Cham : Springer International Publishing. - 1611-3349 .- 0302-9743. ; 11654 LNAI, s. 178-187
-
Konferensbidrag (refereegranskat)abstract
- This paper evaluates two new strategies for investigating artificial animals called animats. Animats are homeostatic agents with the objective of keeping their internal variables as close to optimal as possible. Steps towards the optimal are rewarded and steps away punished. Using reinforcement learning for exploration and decision making, the animats can consider predetermined optimal/acceptable levels in light of current levels, giving them greater flexibility for exploration and better survival chances. This paper considers the resulting strategies as evaluated in a range of environments, showing them to outperform common reinforcement learning, where internal variables are not taken into consideration.
|
|
| 2. |
- Balkenius, Christian, et al.
(författare)
-
Elements of cognition for general intelligence
- 2023
-
Ingår i: Artificial General Intelligence : 16th International Conference, AGI 2023, Stockholm, Sweden, June 16–19, 2023, Proceedings - 16th International Conference, AGI 2023, Stockholm, Sweden, June 16–19, 2023, Proceedings. - 0302-9743 .- 2945-9141 .- 1611-3349 .- 2945-9133. - 9783031334696 - 9783031334689 ; 13921, s. 11-20
-
Konferensbidrag (refereegranskat)abstract
- What can artificial intelligence learn from the cognitive sciences? We review some fundamental aspects of how human cognition works and relate it to different brain structures and their function. A central theme is that cognition is very different from how it is envisioned in classical artificial intelligence which offers a novel path toward intelligent systems that in many ways is both simpler and more attainable. We also argue that artificial intelligent systems takes more than a single silver bullet. It requires a large number of interacting subsystem that are coupled to both the body and to the environment. We argue for an approach to artificial general intelligence based on a faithful reproduction of known brain processes in a system-level model that incorporates a large number of components modelled after the human brain.
|
|
| 3. |
|
|
| 4. |
- Engström, Fredrik, 1977, et al.
(författare)
-
Generating Comprehensible Explanations in Description Logic
- 2014
-
Ingår i: 27th International Workshop on Description Logics. Vienna, Austria, July 17-20, 2014. - 1613-0073.
-
Konferensbidrag (refereegranskat)abstract
- We propose a method for generating comprehensible explanations in description logic. Such explanations could be of potential use for e.g. engineers, doctors, and users of the semantic web. Users commonly need to understand why a logical statement follows from a set of hypotheses. Then, automatically generated explanations that are easily understandable could be of help. A proof system for description logic that can be used for generating comprehensible explanations is proposed. Similar systems have been proposed for propositional logic [30] and first-order logic [28].
|
|
| 5. |
- Hernandez-Orallo, J., et al.
(författare)
-
A New AI Evaluation Cosmos: Ready to Play the Game?
- 2017
-
Ingår i: AI Magazine. - : Wiley. - 0738-4602 .- 2371-9621. ; 38:3, s. 66-69
-
Tidskriftsartikel (refereegranskat)abstract
- We report on a series of new platforms and events dealing with AI evaluation that may change the way in which AI systems are compared and their progress is measured. The introduction of a more diverse and challenging set of tasks in these platforms can feed AI research in the years to come, shaping the notion of success and the directions of the field. However, the playground of tasks and challenges presented there may misdirect the field without some meaningful structure and systematic guidelines for its organization and use. Anticipating this issue, we also report on several initiatives and workshops that are putting the focus on analyzing the similarity and dependencies between tasks, their difficulty, what capabilities they really measure and ultimately on elaborating new concepts and tools that can arrange tasks and benchmarks into a meaningful taxonomy.
|
|
| 6. |
- Johannesson, Louise, et al.
(författare)
-
Basic language learning in artificial animals
- 2019
-
Ingår i: Advances in Intelligent Systems and Computing. - Cham : Springer International Publishing. - 2194-5365 .- 2194-5357. - 9783319993157 ; 848, s. 155-161
-
Konferensbidrag (refereegranskat)abstract
- We explore a general architecture for artificial animals, or animats, that develops over time. The architecture combines reinforcement learning, dynamic concept formation, and homeostatic decision-making aimed at need satisfaction. We show that this architecture, which contains no ad hoc features for language processing, is capable of basic language learning of three kinds: (i) learning to reproduce phonemes that are perceived in the environment via motor babbling; (ii) learning to reproduce sequences of phonemes corresponding to spoken words perceived in the environment; and (iii) learning to ground the semantics of spoken words in sensory experience by associating spoken words (e.g. the word “cold”) to sensory experience (e.g. the activity of a sensor for cold temperature) and vice versa.
|
|
| 7. |
|
|
| 8. |
- Johansson, Anton, 1993, et al.
(författare)
-
Exact spectral norm regularization for neural networks
- 2022
-
Tidskriftsartikel (övrigt vetenskapligt/konstnärligt)abstract
- We pursue a line of research that seeks to regularize the spectral norm of the Jacobian of the input-output mapping for deep neural networks. While previous work rely on upper bounding techniques, we provide a scheme that targets the exact spectral norm. We showcase that our algorithm achieves an improved generalization performance compared to previous spectral regularization techniques while simultaneously maintaining a strong safeguard against natural and adversarial noise. Moreover, we further explore some previous reasoning concerning the strong adversarial protection that Jacobian regularization provides and show that it can be misleading.
|
|
| 9. |
- Johansson, Anton, 1993, et al.
(författare)
-
Improved Spectral Norm Regularization forNeural Networks
- 2023
-
Ingår i: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). - 0302-9743 .- 1611-3349.
-
Konferensbidrag (refereegranskat)abstract
- We improve on a line of research that seeks to regularize the spectral norm of the Jacobian of the input-output mapping for deep neural networks. While previous work rely on upper bounding techniques, we propose a scheme that targets the exact spectral norm. We evaluate this regularization method empirically with respect to its generalization performance and robustness. Our results demonstrate that this improved spectral regularization scheme outperforms L2-regularization as well as the previously used upper bounding technique. Moreover, our results suggest that exact spectral norm regularization and exact Frobenius norm regularization have comparable performance. We analyze these empirical findings in the light of the mathematical relations that hold between the spectral and the Frobenius norms. Lastly, in light of our evaluation we revisit an argument concerning the strong adversarial protection that Jacobian regularization provides and show that it can be misleading. In summary, we propose a new regularization method and contribute to the practical and theoretical understanding of when one regularization method should be preferred over another.
|
|
| 10. |
- Johansson, Anton, 1993, et al.
(författare)
-
Slope and Generalization Properties of Neural Networks
- 2022
-
Ingår i: 34th Workshop of the Swedish Artificial Intelligence Society, SAIS 2022.
-
Konferensbidrag (refereegranskat)abstract
- Neural networks are very successful tools in for example advanced classification. From a statistical point of view, fitting a neural network may be seen as a kind of regression, where we seek a function from the input space to a space of classification probabilities that follows the 'general' shape of the data, but avoids overfitting by avoiding memorization of individual data points. In statistics, this can be done by controlling the geometric complexity of the regression function. We propose to do something similar when fitting neural networks by controlling the slope of the network.After defining the slope and discussing some of its theoretical properties, we go on to show empirically in examples, using ReLU networks, that the distribution of the slope of a well-Trained neural network classifier is generally independent of the width of the layers in a fully connected network, and that the mean of the distribution only has a weak dependence on the model architecture in general. We discuss possible applications of the slope concept, such as using it as a part of the loss function or stopping criterion during network training, or ranking data sets in terms of their complexity.
|
|
