Search

Despite deep learning being state of the art for data-driven model predictions, its application in ecology is currently subject to two important constraints: (i) deep-learning methods are powerful in data-rich regimes, but in ecology data are typically sparse; and (ii) deep-learning models are black-box methods and inferring the processes they represent are non-trivial to elicit. Process-based (= mechanistic) models are not constrained by data sparsity or unclear processes and are thus important for building up our ecological knowledge and transfer to applications. In this work, we combine process-based models and neural networks into process-informed neural networks (PINNs), which incorporate the process knowledge directly into the neural network structure. In a systematic evaluation of spatial and temporal prediction tasks for C-fluxes in temperate forests, we show the ability of five different types of PINNs (i) to outperform process-based models and neural networks, especially in data-sparse regimes with high-transfer task and (ii) to inform on mis- or undetected processes., (© 2024 The Author(s). Ecology Letters published by John Wiley & Sons Ltd.)

Machine learning techniques, particularly deep convolutional neural networks (ConvNets), are increasingly being used to automate clinical EEG analysis, with the potential to reduce the clinical burden and improve patient care. However, further research is required before they can be used in clinical settings, particularly regarding the impact of the number of training samples and model parameters on their testing error. To address this, we present a comprehensive study of the empirical scaling behaviour of ConvNets for EEG pathology classification. We analysed the testing error with increasing the training samples and model size for four different ConvNet architectures. The focus of our experiments is width scaling, and we have increased the number of parameters to up to 1.8 million. Our evaluation was based on two publicly available datasets: the Temple University Hospital (TUH) Abnormal EEG Corpus and the TUH Abnormal Expansion Balanced EEG Corpus, which together contain 10,707 training samples. The results show that the testing error follows a saturating power-law with both model and dataset size. This pattern is consistent across different datasets and ConvNet architectures. Furthermore, empirically observed accuracies saturate at 85%-87%, which may be due to an imperfect inter-rater agreement on the clinical labels. The empirical scaling behaviour of the test performance with dataset and model size has significant implications for deep EEG pathology classification research and practice. Our findings highlight the potential of deep ConvNets for high-performance EEG pathology classification, and the identified scaling relationships provide valuable recommendations for the advancement of automated EEG diagnostics., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could appear to have influenced the work reported in this paper., (Copyright © 2024 The Author(s). Published by Elsevier Ltd.. All rights reserved.)

Food systems are the primary cause of biodiversity loss globally. Biodiversity and specifically, the role that wild, forest and neglected and underutilised species (NUS) foods might play in diet quality is gaining increased attention. The narrow focus on producing affordable staples for dietary energy has contributed to largely homogenous and unhealthy diets. To date, evidence to quantify the nutritional contribution of these biodiverse foods is limited. A scoping review was conducted to document the methods used to quantify the contribution of wild, forest and NUS foods. We found 37 relevant articles from 22 different countries, mainly from Africa (45%), the Americas (19%), and Asia (10%). There were 114 different classifications used for the foods, 73% of these were specifically related to wild or forest foods. Most dietary assessments were completed using a single day qualitative or quantitative 24 h open recall ( n  = 23), or a food frequency questionnaire ( n  = 12). There were 18 different diet related indicators used, mainly nutrient adequacy ( n  = 9) and dietary diversity scores ( n  = 9). Often, no specific nutritionally validated diet metric was used. There were 16 studies that presented results (semi) quantitatively to measure the contribution of wild, forest or NUS foods to dietary intakes. Of these, 38% were aggregated together with broader classifications of 'traditional' or 'local' foods, without definitions provided meaning it was not possible to determine if or to what extend wild, forest of NUS foods were included (or not). In almost all studies there was insufficient detail on the magnitude of the associations between wild, forest or NUS foods and dietary energy or nutrient intakes or the (qualitative) diet recall methodologies that were used inhibited the quantification of the contribution of these foods to diets. A set of six recommendations are put forward to strengthen the evidence on the contribution of wild, NUS, and forest foods to human diets., Competing Interests: The reviewer GK declared a past co-authorship and collaboration with the authors CL, GH-C, JR, and JB to the handling editor. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest., (Copyright © 2023 Raneri, Boedecker, Fallas Conejo, Muir, Hanley-Cook and Lachat.)

Cycling of biologic or targeted synthetic disease modifying antirheumatic drugs (b/tsDMARDs) in rheumatoid arthritis (RA) patients due to non-response is a problem preventing and delaying disease control. We aimed to assess and validate treatment response of b/tsDMARDs among clusters of RA patients identified by deep learning. We clustered RA patients clusters at first-time b/tsDMARD (cohort entry) in the Swiss Clinical Quality Management in Rheumatic Diseases registry (SCQM) [1999-2018]. We performed comparative effectiveness analyses of b/tsDMARDs (ref. adalimumab) using Cox proportional hazard regression. Within 15 months, we assessed b/tsDMARD stop due to non-response, and separately a ≥20% reduction in DAS28-esr as a response proxy. We validated results through stratified analyses according to most distinctive patient characteristics of clusters. Clusters comprised between 362 and 1481 patients (3516 unique patients). Stratified (validation) analyses confirmed comparative effectiveness results among clusters: Patients with ≥2 conventional synthetic DMARDs and prednisone at b/tsDMARD initiation, male patients, as well as patients with a lower disease burden responded better to tocilizumab than to adalimumab (hazard ratio [HR] 5.46, 95% confidence interval [CI] [1.76-16.94], and HR 8.44 [3.43-20.74], and HR 3.64 [2.04-6.49], respectively). Furthermore, seronegative women without use of prednisone at b/tsDMARD initiation as well as seropositive women with a higher disease burden and longer disease duration had a higher risk of non-response with golimumab (HR 2.36 [1.03-5.40] and HR 5.27 [2.10-13.21], respectively) than with adalimumab. Our results suggest that RA patient clusters identified by deep learning may have different responses to first-line b/tsDMARD. Thus, it may suggest optimal first-line b/tsDMARD for certain RA patients, which is a step forward towards personalizing treatment. However, further research in other cohorts is needed to verify our results., Competing Interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: TH has received research grants and honorariums from Pfizer, AbbVie, Novartis, and Janssen, and does consultancy for Eli Lilly, Janssen, and Menarini; no other relationships or activities that could appear to have influenced the submitted work., (Copyright: © 2023 Kalweit et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.)

Automated clinical EEG analysis using machine learning (ML) methods is a growing EEG research area. Previous studies on binary EEG pathology decoding have mainly used the Temple University Hospital (TUH) Abnormal EEG Corpus (TUAB) which contains approximately 3,000 manually labelled EEG recordings. To evaluate and eventually even improve the generalisation performance of machine learning methods for EEG pathology, decoding larger, publicly available datasets is required. A number of studies addressed the automatic labelling of large open-source datasets as an approach to create new datasets for EEG pathology decoding, but little is known about the extent to which training on larger, automatically labelled dataset affects decoding performances of established deep neural networks. In this study, we automatically created additional pathology labels for the Temple University Hospital (TUH) EEG Corpus (TUEG) based on the medical reports using a rule-based text classifier. We generated a dataset of 15,300 newly labelled recordings, which we call the TUH Abnormal Expansion EEG Corpus (TUABEX), and which is five times larger than the TUAB. Since the TUABEX contains more pathological (75%) than non-pathological (25%) recordings, we then selected a balanced subset of 8,879 recordings, the TUH Abnormal Expansion Balanced EEG Corpus (TUABEXB). To investigate how training on a larger, automatically labelled dataset affects the decoding performance of deep neural networks, we applied four established deep convolutional neural networks (ConvNets) to the task of pathological versus non-pathological classification and compared the performance of each architecture after training on different datasets. The results show that training on the automatically labelled TUABEXB dataset rather than training on the manually labelled TUAB dataset increases accuracies on TUABEXB and even for TUAB itself for some architectures. We argue that automatically labelling of large open-source datasets can be used to efficiently utilise the massive amount of EEG data stored in clinical archives. We make the proposed TUABEXB available open source and thus offer a new dataset for EEG machine learning research., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2023. Published by Elsevier Inc.)

Background: There is a current need for better understanding the impact of nutrition-sensitive agriculture interventions. This study is based on a community-based participatory project that diversified diets of women and children by making use of local food biodiversity. This retrospective impact pathway analysis aims at explaining why and how impact was reached., Objectives: This study aimed to understand how a nutrition-sensitive agriculture project improved people's diets by analyzing the pathways from agriculture to nutrition. It also aimed to test theoretical pathways by comparing the documented pathways with those from a widely used framework from the literature., Methods: A qualitative study was conducted in 2019 through 10 semistructured focus group discussions with community members engaging in the project and 5 key informant interviews with local authorities that worked with these communities during the project. Summative content analysis was used to identify pathways through which the project affected diets of beneficiaries. The defined pathways were compared with the pathways of the widely used Tackling the Agriculture-Nutrition Disconnect in India (TANDI) framework from the literature., Results: Out of the agriculture-nutrition pathways that are presented in the literature, 3 were found in the responses: 1 ) food from own production; 2 ) income from sale of foods produced; and 3 ) women's empowerment through access to and control over resources. In addition, 5 other pathways were identified and indicated spillover effects from the intervention to the control participants, increased nutrition knowledge, improved health, savings, and empowerment and harmony in the household., Conclusions: Pathway analysis in nutrition-sensitive agriculture can provide valuable understanding on how and why dietary improvements have been achieved in an intervention. The approach can hence be instrumental in addressing the current demand within the field on understanding the progress and impact of interventions. Pathway analysis also helps to address knowledge gaps regarding theoretical frameworks, as in the present study, concerning women empowerment pathways., (© The Author(s) 2021. Published by Oxford University Press on behalf of the American Society for Nutrition.)

Background: Deep neural networks learn from former experiences on a large scale and can be used to predict future disease activity as potential clinical decision support. AdaptiveNet is a novel adaptive recurrent neural network optimized to deal with heterogeneous and missing clinical data., Objective: We investigate AdaptiveNet for the prediction of individual disease activity in patients from a rheumatoid arthritis (RA) registry., Methods: Demographic and disease characteristics from over 9500 patients and 65.000 visits from the Swiss Quality Management (SCQM) database were used to train and evaluate the network. Patient characteristics, clinical and patient reported outcomes, laboratory values and medication were used as input features. DAS28-BSR served as a target to predict active RA and future numeric individual disease activity by classification and regression., Results: AdaptiveNet predicted active disease defined as DAS28-BSR >2.6 at the next visit with an overall accuracy of 75.6% (SD +- 0.7%) and a sensitivity and specificity of 84.2% (SD +- 1.6%) and 61.5% (SD +- 3.6%), respectively. Prediction performance was significantly higher in patients with a disease duration >3 years and positive rheumatoid factor. Regression allowed forecasting individual DAS28-BSR values with a mean squared error (MSE) of 0.9 (SD +- 0.05). This corresponds to a 8% deviation between estimated and real DAS28-BSR values. Compared to linear regression, random forest and support vector machines, AdaptiveNet showed an increased performance of over 7% in MSE. Medication played a minor role in the prediction of RA disease activity., Conclusion: AdaptiveNet has a superior capacity to predict numeric RA disease activity compared to classical machine learning architectures. All investigated models had limitations in low specificity., Competing Interests: The authors have no competing interest.

Machine learning (ML) methods have the potential to automate clinical EEG analysis. They can be categorized into feature-based (with handcrafted features), and end-to-end approaches (with learned features). Previous studies on EEG pathology decoding have typically analyzed a limited number of features, decoders, or both. For a I) more elaborate feature-based EEG analysis, and II) in-depth comparisons of both approaches, here we first develop a comprehensive feature-based framework, and then compare this framework to state-of-the-art end-to-end methods. To this aim, we apply the proposed feature-based framework and deep neural networks including an EEG-optimized temporal convolutional network (TCN) to the task of pathological versus non-pathological EEG classification. For a robust comparison, we chose the Temple University Hospital (TUH) Abnormal EEG Corpus (v2.0.0), which contains approximately 3000 EEG recordings. The results demonstrate that the proposed feature-based decoding framework can achieve accuracies on the same level as state-of-the-art deep neural networks. We find accuracies across both approaches in an astonishingly narrow range from 81 to 86%. Moreover, visualizations and analyses indicated that both approaches used similar aspects of the data, e.g., delta and theta band power at temporal electrode locations. We argue that the accuracies of current binary EEG pathology decoders could saturate near 90% due to the imperfect inter-rater agreement of the clinical labels, and that such decoders are already clinically useful, such as in areas where clinical EEG experts are rare. We make the proposed feature-based framework available open source and thus offer a new tool for EEG machine learning research., Competing Interests: Declaration of competing interest The authors declare no competing financial interests., (Copyright © 2020. Published by Elsevier Inc.)

Wild plant species are often excellent sources of micronutrients and have the potential to promote healthy living, yet they are under-exploited. Distribution of micronutrient powders as diet supplements can play an effective role in reducing micronutrient deficiencies among infants and young children. However, assessing their effects in ensuring a nutritious diet at low cost have been limited. This study assessed the impact of including wild plant species and micronutrient powders in modeled optimized lowest-cost diets for women and children in rural Kenya. Market surveys, focus group discussions in six villages and a 24-h dietary intake recall were used to collect data that were subsequently entered in the cost of diet linear programming tool to model lowest-cost nutritious diets for women and children in Turkana County, Kenya. Three wild vegetables, three wild fruits, and micronutrient powder were added to the models to assess their impact on the cost and the nutrient adequacy of the diets. A locally adapted cost optimized nutritious diet without any intervention costs between 50 and 119 Kenyan shillings (KES) daily ($0.5 to $1.2) for children between 6 and 23 months and 173 to 305 KES ($1.8 to $2.9) for women. Addition of the three wild vegetables resulted in cost reductions between 30 and 71% as well as making up for iron and zinc gaps. The micronutrient powder had an insignificant effect on diet cost and filling nutrient gaps. Edible wild plant species, specifically wild vegetables, can reduce diet costs in considerable proportions while filling nutrient gaps year-round. However, affordability of a nutritious diet remains a major challenge in Turkana County, irrespective of the wealth group., (Copyright © 2020 Sarfo, Keding, Boedecker, Pawelzik and Termote.)

Machine learning as a field of artificial intelligence is increasingly applied in medicine to assist patients and physicians. Growing datasets provide a sound basis with which to apply machine learning methods that learn from previous experiences. This review explains the basics of machine learning and its subfields of supervised learning, unsupervised learning, reinforcement learning and deep learning. We provide an overview of current machine learning applications in rheumatology, mainly supervised learning methods for e-diagnosis, disease detection and medical image analysis. In the future, machine learning will be likely to assist rheumatologists in predicting the course of the disease and identifying important disease factors. Even more interestingly, machine learning will probably be able to make treatment propositions and estimate their expected benefit (e.g. by reinforcement learning). Thus, in future, shared decision-making will not only include the patient's opinion and the rheumatologist's empirical and evidence-based experience, but it will also be influenced by machine-learned evidence., (© The Author(s) 2020. Published by Oxford University Press on behalf of the British Society for Rheumatology.)

Appropriate robot behavior during human-robot interaction is a key part in the development of human-compliant assistive robotic systems. This study poses the question of how to continuously evaluate the quality of robotic behavior in a hybrid brain-computer interfacing (BCI) task, combining brain and non-brain signals, and how to use the collected information to adapt the robot's behavior accordingly. To this aim, we developed a rating system compatible with EEG recordings, requiring the users to execute only small movements with their thumb on a wireless controller to rate the robot's behavior on a continuous scale. The ratings were recorded together with dry EEG, respiration, ECG, and robotic joint angles in ROS. Pilot experiments were conducted with three users that had different levels of previous experience with robots. The results demonstrate the feasibility to obtain continuous rating data that give insight into the subjective user perception during direct human-robot interaction. The rating data suggests differences in subjective perception for users with no, moderate, or substantial previous robot experience. Furthermore, a variety of regression techniques, including deep CNNs, allowed us to predict the subjective ratings. Performance was better when using the position of the robotic hand than when using EEG, ECG, or respiration. A consistent advantage of features expected to be related to a motor bias could not be found. Across-user predictions showed that the models most likely learned a combination of general and individual features across-users. A transfer of pre-trained regressor to a new user was especially accurate in users with more experience. For future research, studies with more participants will be needed to evaluate the methodology for its use in practice. Data and code to reproduce this study are available at https://github.com/TNTLFreiburg/NiceBot., (Copyright © 2019 Fiederer, Völker, Schirrmeister, Burgard, Boedecker and Ball.)

Due to their limited access to the external productive inputs and the dependency on rain-fed agricultural production, small scale farmers in sub-Saharan Africa have continued to face undernutrition despite the significant advancements in agriculture. They however often live in areas endowed with high agrobiodiversity which could contribute, if explored, to improved diets and nutrition. Few studies have linked the contribution of agrobiodiversity to the micronutrient adequacy of the diets of young children among smallholder farmers. The study explored this relationship and contributes to the growing body of literature linking agrobiodiversity to nutrition of young children. Two cross-sectional surveys were conducted as part of baseline assessment for an intervention study, one in the lean and a second in the plenty season in Vihiga county, Kenya. Household level interviews were administered to 634 households with children 12-23 months. Agrobiodiversity was defined as the number of crop species cultivated or harvested from the wild and the number of livestock maintained by the household across two agricultural seasons. Dietary data were collected using two-non-consecutive quantitative 24-hour recalls and analyzed using Lucille software. Diet quality was assessed using dietary diversity score based on seven food groups and mean probability of micronutrient adequacy computed for eleven micronutrients. A total of 80 species were maintained or harvested from the wild by the households. Mean household species richness was 9.9 ± 4.3. One in every four children did not meet the minimum dietary diversity score. The average mean probability of micronutrient adequacy was 68.11 ± 16.08 in plenty season compared to 56.37± 19.5% in the lean season. Iron, zinc and calcium were most limiting micronutrients in the diet, with less than 30% average probability of adequacy in both seasons. Household agrobiodiversity was positively associated with both dietary diversity score (r = 0.09, p = 0.029) and micronutrient adequacy (r = 0.15, p<0.000) in the pooled sample. One unit increase in species diversity was associated with 12.7% improvement in micronutrient adequacy. Despite the rich agrobiodiversity in the study area the diets were low in diversity and there is an unrealized opportunity to improve micronutrient intake through greater promotion and consumption of locally available agrobiodiversity., Competing Interests: The authors have declared that no competing interests exist.

Our study assessed the effectiveness of a community-based participatory approach in increasing micronutrient adequacy of diets of women and young children through agricultural activities and nutrition education in Vihiga County, Western Kenya. Outcome indicators include the mean dietary diversity score (DDS), the percentage of women and children reaching minimum dietary diversity (MDD), and micronutrient adequacy (mean adequacy ratio). The project consisted of(a) a diagnostic survey covering agrobiodiversity and nutrition, (b) participatory development of activities to improve nutrition, (c) a baseline survey covering dietary intakes, (d) participatory implementation of the developed activities, and (e) an endline survey covering dietary intakes. The diagnostic survey was conducted in 10 sublocations of Vihiga County, which were pair-matched and split into five intervention and five control sublocations. The intervention sublocations developed activities towards improving nutrition. Before implementation, a baseline survey collected the dietary intake data of 330 women-child pairs in the intervention and control sublocations. To support the activities, communities received agriculture and nutrition training. After 1 year of implementation, an endline survey collected dietary intake data from 444 women-child pairs in the intervention and control sublocations. Impact was assessed using the difference-in-difference technique. Highly significant positive impacts on children's mean DDS (treatment effect = 0.7, p < 0.001) and on the share of children reaching MDD (treatment effect = 0.2, p < 0.001) were shown. Higher dietary diversity can be explained by the development of subsistence and income-generating pathways and increased nutrition knowledge. Participatory farm diversification and nutrition education were shown to significantly increase dietary diversity of young children in Western Kenya., (© 2019 The Authors. Maternal and Child Nutrition Published by John Wiley & Sons, Ltd.)

Smallholder farmers dependent on rain-fed agriculture experience seasonal variations in food and nutrient availability occasioned by seasonality of production patterns. This results in periods of nutrient abundance in the plenty seasons followed closely by periods of nutrient inadequacies and malnutrition. This pattern contributes to a cycle of deteriorating health and nutrition status and deprives children of their ability to realize full developmental potential. This study investigates the role of caregiver's nutritional knowledge and attitudes in mediating effects of seasonality on children's diets. Repeated cross-sectional surveys were conducted on 151 randomly selected households in the plenty and lean seasons to collect dietary data using two non-consecutive quantitative 24-hr recalls and caregiver's nutritional knowledge and attitudes assessed using interviewer administered questionnaire. Sixty-five percent of the caregivers had attained a primary level education or less. There was a positive modest correlation between caregivers' nutritional knowledge and their attitudes (r = 0.3, P < 0.000, α = 0.01). Children's mean adequacy ratio was significantly higher in the plenty season than in the lean season (0.84 vs. 0.80, P < 0.000). A two-block hierarchical regression to predict the seasonal changes in dietary quality of children using caregiver's nutritional knowledge and attitude scores while controlling for the effect of sociodemographics and mean adequacy ratio at first season (plenty) found that caregiver's nutritional knowledge (ß = -0.007, SE = 0.003, P = 0.027, 95% CI [-0.013, -0.001] ŋ 2  = 0.034) but not attitudes had significant contribution to the prediction. Maternal nutritional knowledge mediates seasonal variation in child nutrient intakes., (© 2018 The Authors. Maternal and Child Nutrition Published by John Wiley & Sons, Ltd.)

We present for the first time a μW-power convolutional neural network for seizure detection running on a low-power microcontroller. On a dataset of 22 patients a median sensitivity of 100% is achieved. With a false positive rate of 20.7 fp/h and a short detection delay of 3.4 s it is suitable for the application in an implantable closed-loop device.

Driven by clinical needs and progress in neurotechnology, targeted interaction with neuronal networks is of increasing importance. Yet, the dynamics of interaction between intrinsic ongoing activity in neuronal networks and their response to stimulation is unknown. Nonetheless, electrical stimulation of the brain is increasingly explored as a therapeutic strategy and as a means to artificially inject information into neural circuits. Strategies using regular or event-triggered fixed stimuli discount the influence of ongoing neuronal activity on the stimulation outcome and are therefore not optimal to induce specific responses reliably. Yet, without suitable mechanistic models, it is hardly possible to optimize such interactions, in particular when desired response features are network-dependent and are initially unknown. In this proof-of-principle study, we present an experimental paradigm using reinforcement-learning (RL) to optimize stimulus settings autonomously and evaluate the learned control strategy using phenomenological models. We asked how to (1) capture the interaction of ongoing network activity, electrical stimulation and evoked responses in a quantifiable 'state' to formulate a well-posed control problem, (2) find the optimal state for stimulation, and (3) evaluate the quality of the solution found. Electrical stimulation of generic neuronal networks grown from rat cortical tissue in vitro evoked bursts of action potentials (responses). We show that the dynamic interplay of their magnitudes and the probability to be intercepted by spontaneous events defines a trade-off scenario with a network-specific unique optimal latency maximizing stimulus efficacy. An RL controller was set to find this optimum autonomously. Across networks, stimulation efficacy increased in 90% of the sessions after learning and learned latencies strongly agreed with those predicted from open-loop experiments. Our results show that autonomous techniques can exploit quantitative relationships underlying activity-response interaction in biological neuronal networks to choose optimal actions. Simple phenomenological models can be useful to validate the quality of the resulting controllers.

A common assumption in psychology, economics, and other fields holds that higher performance will result if extrinsic rewards (such as money) are offered as an incentive. While this principle seems to work well for tasks that require the execution of the same sequence of steps over and over, with little uncertainty about the process, in other cases, especially where creative problem solving is required due to the difficulty in finding the optimal sequence of actions, external rewards can actually be detrimental to task performance. Furthermore, they have the potential to undermine intrinsic motivation to do an otherwise interesting activity. In this work, we extend a computational model of the dorsomedial and dorsolateral striatal reinforcement learning systems to account for the effects of extrinsic and intrinsic rewards. The model assumes that the brain employs both a goal-directed and a habitual learning system, and competition between both is based on the trade-off between the cost of the reasoning process and value of information. The goal-directed system elicits internal rewards when its models of the environment improve, while the habitual system, being model-free, does not. Our results account for the phenomena that initial extrinsic reward leads to reduced activity after extinction compared to the case without any initial extrinsic rewards, and that performance in complex task settings drops when higher external rewards are promised. We also test the hypothesis that external rewards bias the competition in favor of the computationally efficient, but cruder and less flexible habitual system, which can negatively influence intrinsic motivation and task performance in the class of tasks we consider.

We investigate information processing in randomly connected recurrent neural networks. It has been shown previously that the computational capabilities of these networks are maximized when the recurrent layer is close to the border between a stable and an unstable dynamics regime, the so called edge of chaos. The reasons, however, for this maximized performance are not completely understood. We adopt an information-theoretical framework and are for the first time able to quantify the computational capabilities between elements of these networks directly as they undergo the phase transition to chaos. Specifically, we present evidence that both information transfer and storage in the recurrent layer are maximized close to this phase transition, providing an explanation for why guiding the recurrent layer toward the edge of chaos is computationally useful. As a consequence, our study suggests self-organized ways of improving performance in recurrent neural networks, driven by input data. Moreover, the networks we study share important features with biological systems such as feedback connections and online computation on input streams. A key example is the cerebral cortex, which was shown to also operate close to the edge of chaos. Consequently, the behavior of model systems as studied here is likely to shed light on reasons why biological systems are tuned into this specific regime.

Reservoir computing (RC) is a recent paradigm in the field of recurrent neural networks. Networks in RC have a sparsely and randomly connected fixed hidden layer, and only output connections are trained. RC networks have recently received increased attention as a mathematical model for generic neural microcircuits to investigate and explain computations in neocortical columns. Applied to specific tasks, their fixed random connectivity, however, leads to significant variation in performance. Few problem-specific optimization procedures are known, which would be important for engineering applications, but also in order to understand how networks in biology are shaped to be optimally adapted to requirements of their environment. We study a general network initialization method using permutation matrices and derive a new unsupervised learning rule based on intrinsic plasticity (IP). The IP-based learning uses only local learning, and its aim is to improve network performance in a self-organized way. Using three different benchmarks, we show that networks with permutation matrices for the reservoir connectivity have much more persistent memory than the other methods but are also able to perform highly nonlinear mappings. We also show that IP-based on sigmoid transfer functions is limited concerning the output distributions that can be achieved.