I detta arbete presenteras en ny teori som försöker förklara hur man ytterligare kan förbättra instruktioner av processer. Teorin försöker bygga en bro mellan metoder för att stödja bearbetningen i arbetsminnet och hur perceptionen fungerar genom att utgå från arbetsminnets roll vid inlärning.

Curiosity is thought to be an intrinsically motivated driving force for seeking information. Thus, the opportunity for an information gain (IG) should instil curiosity in humans and result in information gathering actions. To investigate if, and how, information acts as an intrinsic reward, a search task was set in a context of blurred background images which could be revealed by iterative clicking. The search task was designed such that it prevented efficient IG about the underlying images. Participants therefore had to trade between clicking regions with high search target probability or high expected image content information. Image content IG was established from “information-maps” based on participants exploration with the intention of understanding (1) the main theme of the image and (2) how interesting the image might appear to others. Note that IG is in this thesis not identical with the information theoretic concept of information gain, the quantities are however probably related. It was hypothesised that participants would be distracted by visually informative regions and that images independently rated as more interesting would yield higher image based IG. It was also hypothesised that image based IG would increase as a function of time. Results show that participants sometimes explored images driven by curiosity, and that there was considerable individual variation in which images participants were curious about. Independent interest ratings did not account for image based IG. The level of IG increased over trials, interestingly without affecting participants’ performance on the visual search task designed to prevent IG. Results support that IG is rewarding as participants learned to optimize IG over trials without compromising performance on the extrinsically motivated search; managing to both keep the cake and eat it.

Visual search is the perceptual task of locating a target in a visual environment. Due to applications in areas like search and rescue, surveillance, and home assistance, it is of great interest to automate visual search. An autonomous system can potentially search more efficiently than a manually controlled one and has the advantages of reduced risk and cost of labor. In many environments, there is structure that can be utilized to find targets quicker. However, manually designing search algorithms that properly utilize structure to search efficiently is not trivial. Different environments may exhibit vastly different characteristics, and visual cues may be difficult to pick up. A learning system has the advantage of being applicable to any environment where there is a sufficient number of samples to learn from. In this thesis, we investigate how an agent that learns to search can be implemented with deep reinforcement learning. Our approach jointly learns control of visual attention, recognition, and localization from a set of sample search scenarios. A recurrent convolutional neural network takes an image of the visible region and the agent's position as input. Its outputs indicate whether a target is visible and control where the agent looks next. The recurrent step serves as a memory that lets the agent utilize features of the explored environment when searching. We compare two memory architectures: an LSTM, and a spatial memory that remembers structured visual information. Through experimentation in three simulated environments, we find that the spatial memory architecture achieves superior search performance. It also searches more efficiently than a set of baselines that do not utilize the appearance of the environment and achieves similar performance to that of a human searcher. Finally, the spatial memory scales to larger search spaces and is better at generalizing from a limited number of training samples.