Real-world application of data-driven and intelligent systems (AI) is increasing in the private and public sector as well as in society at large. Many organizations transform as a consequence of increased AI implementation. The consequences of such transformations may include new recruitment plans, procurement of additional IT, changes in existing positions and roles, new business models, as well as new policies and regulations. However, it is unclear how this transformation varies across different types of organizations. We study the effects of bottom-up approaches, such as pilot projects and mentoring to specific groups within organizations, and aim to explore how such approaches can complement the top-down approach of strategic AI implementation. Our context is the public sector. Our goal is to acquire an improved understanding of how and when AI transformation occurs in the public sector, which are the consequences, and which strategies are fruitful or detrimental to the organization. We aim to study public sector organizations in Sweden, Norway, New Zealand, Germany, and The Netherlands to learn about potential similarities and differences with regard to AI transformation. 

This paper introduces a novel deep learning architecture, named DIGITNET, and a large-scale digit dataset, named DIDA, to detect and recognize handwritten digits in historical document images written in the nineteen century. To generate the DIDA dataset, digit images are collected from 100,000 Swedish handwritten historical document images, which were written by different priests with different handwriting styles. This dataset contains three sub-datasets including single digit, large-scale bounding box annotated multi-digit, and digit string with 250,000, 25,000, and 200,000 samples in Red-Green-Blue (RGB) color spaces, respectively. Moreover, DIDA is used to train the DIGITNET network, which consists of two deep learning architectures, called DIGITNET-dect and DIGITNET-rec, respectively, to isolate digits and recognize digit strings in historical handwritten documents. In DIGITNET-dect architecture, to extract features from digits, three residual units where each residual unit has three convolution neural network structures are used and then a detection strategy based on You Look Only Once (YOLO) algorithm is employed to detect handwritten digits at two different scales. In DIGITNET-rec, the detected isolated digits are passed through 3 different designed Convolutional Neural Network (CNN) architectures and then the classification results of three different CNNs are combined using a voting scheme to recognize digit strings. The proposed model is also trained with various existing handwritten digit datasets and then validated over historical handwritten digit strings. The experimental results show that the proposed architecture trained with DIDA (publicly available from: https://didadataset.github.io/DIDA/) outperforms the state-of-the-art methods.

Energy consumption reduction has been an increasing trend in machine learning over the past few years due to its socio-ecological importance. In new challenging areas such as edge computing, energy consumption and predictive accuracy are key variables during algorithm design and implementation. State-of-the-art ensemble stream mining algorithms are able to create highly accurate predictions at a substantial energy cost. This paper introduces the nmin adaptation method to ensembles of Hoeffding tree algorithms, to further reduce their energy consumption without sacrificing accuracy. We also present extensive theoretical energy models of such algorithms, detailing their energy patterns and how nmin adaptation affects their energy consumption. We have evaluated the energy efficiency and accuracy of the nmin adaptation method on five different ensembles of Hoeffding trees under 11 publicly available datasets. The results show that we are able to reduce the energy consumption significantly, by 21% on average, affecting accuracy by less than one percent on average.

Recently machine learning researchers are designing algorithms that can run in embedded and mobile devices, which introduces additional constraints compared to traditional algorithm design approaches. One of these constraints is energy consumption, which directly translates to battery capacity for these devices. Streaming algorithms, such as the Very Fast Decision Tree (VFDT), are designed to run in such devices due to their high velocity and low memory requirements. However, they have not been designed with an energy efficiency focus. This paper addresses this challenge by presenting the nmin adaptation method, which reduces the energy consumption of the VFDT algorithm with only minor effects on accuracy. nmin adaptation allows the algorithm to grow faster in those branches where there is more confidence to create a split, and delays the split on the less confident branches. This removes unnecessary computations related to checking for splits but maintains similar levels of accuracy. We have conducted extensive experiments on 29 public datasets, showing that the VFDT with nmin adaptation consumes up to 31% less energy than the original VFDT, and up to 96% less energy than the CVFDT (VFDT adapted for concept drift scenarios), trading off up to 1.7 percent of accuracy.

Score calculation and performance analysis of shooting targets is an important aspect in the development of sports shooting ability. An image-based automatic scoring algorithm would provide automation of this procedure and digital visualization of the result. Existing solutions are able to detect hits with high precision. However, these methods are either too expensive or adapted to unrealistic use cases where high quality paper targets are photographed in very favorable environments. Usually, precision pistol shooting is performed outdoors and bullet holes are covered with stickers between shooting rounds. The targets are reused until they are destroyed. This paper introduces the first generation of an image-based method for automatic hit detection adapted to realistic shooting conditions. It relies solely on available image processing techniques. The proposed algorithm detects hits with 40 percent detection rate in low-quality targets, reaching 88 percent detection rate in targets of higher quality.

Recommendation systems in fashion are used to provide recommendations to users on clothing items, matching styles, and size or fit. These recommendations are generated based on user actions such as ratings, reviews or general interaction with a seller. There is an increased adoption of implicit feedback in models aimed at providing recommendations in fashion. This paper aims to understand the nature of implicit user feedback in fashion recommendation systems by following guidelines to group user actions. Categories of user actions that characterize implicit feedback are examination, retention, reference, and annotation. Each category describes a specific set of actions a user takes. It is observed that fashion recommendations using implicit user feedback mostly rely on retention as a user action to provide recommendations.

This paper compares three different word image representations as base for label free sample selection for word spotting in historical handwritten documents. These representations are a temporal pyramid representation based on pixel counts, a graph based representation, and a pyramidal histogram of characters (PHOC) representation predicted by a PHOCNet trained on synthetic data. We show that the PHOC representation can help to reduce the amount of required training samples by up to 69% depending on the dataset, if it is learned iteratively in an active learning like fashion. While this works for larger datasets containing about 1,700 images, for smaller datasets with 100 images, we find that the temporal pyramid and the graph representation perform better.

Involving humans in the learning process of a machine learning algorithm can have many advantages ranging from establishing trust into a particular model to added personalization capabilities to reducing labeling efforts. While these approaches are commonly summarized under the term interactive machine learning (iML), no unambiguous definition of iML exists to clearly define this area of research. In this position paper, we discuss the shortcomings of current definitions of iML and propose and define the term guided machine learning (gML) as an alternative.

Finding experts in academics is an important practical problem, e.g. recruiting reviewersfor reviewing conference, journal or project submissions, partner matching for researchproposals, finding relevant M. Sc. or Ph. D. supervisors etc. In this work, we discuss anexpertise recommender system that is built on data extracted from the Blekinge Instituteof Technology (BTH) instance of the institutional repository system DiVA (DigitalScientific Archive). DiVA is a publication and archiving platform for research publicationsand student essays used by 46 publicly funded universities and authorities in Sweden andthe rest of the Nordic countries (www.diva-portal.org). The DiVA classification system isbased on the Swedish Higher Education Authority (UKÄ) and the Statistic Sweden's (SCB)three levels classification system. Using the classification terms associated with studentM. Sc. and B. Sc. theses published in the DiVA platform, we have developed a prototypesystem which can be used to identify and recommend subject thesis supervisors in academy.