Featured Content
System-Level Cognitive Modeling with Ikaros
The goal of the Ikaros project is to develop an open infrastructure for system-level cognitive modelling. This includes system-level models of the brain, but also cognitive models that are not directly motivated by brain function. The article describes the background for the project and the
details of the Ikaros system.
What's New
The Missing Link Between Memory and Reinforcement Learning
December 10, 2020
In a new publication, Balkenius, Tjøstheim, Johansson, Wallin and Gädenfors extend their earlier model of memory processing with a decision making mechanism. They describe how this memory mechanism can support decision making when the alternatives cannot be evaluated based on immediate sensory information alone. Instead we first imagine, and then evaluate a possible future that will result from choosing one of the alternatives. The model accumulates evidence over time, whether that information comes from the sequential attention to different sensory properties or from internal simulation of the consequences of making a particular choice. The authors show how the new model explains both simple immediate choices, choices that depend on multiple sensory factors and complicated selections between alternatives that require forward looking simulations based on episodic and semantic memory structures. In this framework, vicarious trial and error is explained as an internal simulation that accumulates evidence for a particular choice. It is argued that a system like this forms the “missing link” between more traditional ideas of semantic and episodic memory, and the associative nature of reinforcement learning.
Balkenius, C., Tjøstheim, T. A., Johansson, B., Wallin, A., & Gärdenfors, P. (2020). The Missing Link Between Memory and Reinforcement Learning. Frontiers in Psychology, 11, 3446.
A Computational Model of Trust-, Pupil-, and Motivation Dynamics
September 20, 2019
In a new publication, Tjøstheim, Johansson and Balkenius argue that machines may benefit from being able to explicitly build or withdraw trust with specific humans. The latter is relevant in situations where the integrity of an autonomous system is compromised, or if humans display untrustworthy behaviour towards the system. Examples of systems that could benefit might be delivery robots, maintenance robots, or autonomous taxis. This work contributes by presenting a biologically plausible model of unconditional trust dynamics, which simulates trust building based on familiarity, but which can be modulated by painful and gentle touch. The model displays interactive behaviour by being able to realistically control pupil dynamics, as well as determine approach and avoidance motivation.
Tjøstheim, T. A., Johansson, B., & Balkenius, C. (2019). A Computational Model of Trust-, Pupil-, and Motivation Dynamics. In HAI 2019. ACM.
Cumulative inhibition in neural networks
November 03, 2018
In a new publication, Tjøstheim and Balkenius show how a multi-resolution network can model the development of acuity and coarse-to-fine processing in the mammalian visual cortex. The network adapts to input statistics in an unsupervised manner, and learns a coarse-to-fine representation by using cumulative inhibition of nodes within a network layer. We show that a system of such layers can represent input by hierarchically composing larger parts from smaller components. It can also model aspects of top-down processes, such as image regeneration.
Tjøstheim, T. A. & Balkenius, C. (2018). Cumulative inhibition in neural networks, Cognitive Processing, 1-16.
Resources
Logged Starts (v1, v2)
Logged Starts (v3)
Recent Commits
Ikaros can be extended with modules that implements models of different brain regions or communicates with external devices such as robots.
A number of models of haptic perception have used Ikaros to control different robot hands.