Artificial Stream Research Articles

The impact of repeated pyrethroid pulses on aquatic communities

Pesticides are considered to be one of the main causes of the decline in macroinvertebrate biodiversity in small streams. In particular, pyrethroids are detected in agricultural surface waters worldwide and pose a high risk to aquatic invertebrates. Due to their knock-down effect, even short pyrethroid exposure pulses can have significant short- and long-term effects on macroinvertebrate communities. Therefore, it is necessary to consider more realistic exposure scenarios for the environmental risk assessment of pyrethroids and, consequently, to obtain more realistic effect data by using multi-stressor test systems.In an experimental setup with artificial indoor streams (AIS), four pyrethroid pulses simulated the exposure scenario of heavy rainfall events. Effects of these 12 h-exposures at different concentrations of deltamethrin (0.64 ng/L, 4 ng/L, 16 ng/L, 64 ng/L) with intervening recovery periods of six days were assessed on an aquatic community consisting of Gammarus pulex, Ephemera danica, Lumbriculus variegatus and Potamopyrgus antipodarum with various lethal and sub-lethal endpoints.The mortality rate of G. pulex significantly increased with increasing deltamethrin concentrations, whereas the mean number of offspring significantly decreased (NOECoffspring: 16 ng/L, LOECoffspring: 64 ng/L). The biomass of L. variegatus decreased with increasing deltamethrin concentrations (NOECdry weight: 16 ng/L, LOECdry weight: 64 ng/L).The findings of this study clearly demonstrate that 12 h-deltamethrin pulses at environmentally relevant concentrations adversely affect an aquatic community. Based on the results of this study a RAC value of 5.33 ng/L is assumed, which is below the concentrations measured in rivers of up to 58.8 ng/L. Unacceptable effects on the entire freshwater environment can therefore not be ruled out. The experimental AIS approach is a useful tool for assessing the effects of repeated pulse exposures that occur during surface runoff events.

Change detection and adaptation in multi-target regression on data streams

AbstractAn essential characteristic of data streams is the possibility of occurrence of concept drift, i.e., change in the distribution of the data in the stream over time. The capability to detect and adapt to changes in data stream mining methods is thus a necessity. While methods for multi-target prediction on data streams have recently appeared, they have largely remained without such capability. In this paper, we propose novel methods for change detection and adaptation in the context of incremental online learning of decision trees for multi-target regression. One of the approaches we propose is ensemble based, while the other uses the Page–Hinckley test. We perform an extensive evaluation of the proposed methods on real-world and artificial data streams and show their effectiveness. We also demonstrate their utility on a case study from spacecraft operations, where cosmic events can cause change and demand an appropriate and timely positioning of the space craft.

Cobalt effects on prokaryotic communities of river biofilms: Impact on their colonization kinetics, structure and functions

Although cobalt (Co) plays a significant role in the transition to low-carbon technologies, its environmental impact remains largely unknown. This study examines Co impacts on the prokaryotic communities within river biofilms to evaluate their potential use as bioindicators of Co contamination. To this end, biofilms were cultivated in artificial streams enriched with different environmental Co concentrations (0.1, 0.5, and 1 μM Co) over 28 days and examined for prokaryotic abundance and diversity via quantitative PCR and DNA-metabarcoding every 7 days. The prokaryotic community's resilience was further investigated after an additional 35 days without Co contamination. The prokaryotic communities were affected by 0.5 and 1 μM Co from the onset of biofilm colonization. The biofilm biomass was comparable between treatments, but the community composition differed. Control biofilms were dominated by Cyanobacteria and Planctomycetes, whereas Bacteroidetes dominated the Co-contaminated biofilms. Potential functional redundancy was observed through the implementation of carbon fixation alternatives by non-photosynthetic prokaryotes in biofilms exposed to high Co concentrations. No structural resilience was observed in the biofilms after 35 days without Co contamination. Measuring the prokaryotic community structural response using molecular approaches appears to be a promising method for assessing shifts in water quality owing to Co contamination.

Streambed immobilization controls the transport of antibiotic resistance genes in flowing water

Antibiotic resistance is a serious global health issue, resulting in at least 1.2 million deaths in 2019. The environment is a potentially important reservoir of antibiotic resistance; however, the fate of Antibiotic Resistance Genes (ARGs) in the environment remains poorly characterized. One important environmental source of ARGs is manure used as a soil amendment. ARGs from manure may then enter nearby flowing waterbodies, where the factors governing their downstream transport remain unknown. To address this, we conducted experiments by spiking cattle manure in an artificial stream to estimate removal rates (k; m−1) for three ARGs (mefA, tetQ, and tetW) and a ruminant fecal marker (bacR). We then used a Stochastic Mobile-Immobile Model (SMIM) to separate the overall removal into two components, rs, and rh, corresponding to immobilizations in the surface (i.e., water column) and subsurface (i.e., streambed), respectively. Finally, we applied the SMIM across four model streams to predict the downstream travel distance of ARGs and bacR. Our results showed measurable removal for all targets in all experimental replicates (n = 3) and no differences were found in the removal rates among replicates for any target (ANCOVA; p > 0.05). We found that the removal of bacR was significantly lower than tetW (p < 0.05) and slightly lower than mefA (p = 0.088), while tetQ removal was slightly different from tetW’s (p = 0.072). We also found that rh values were orders of magnitude larger than rs for ARGs and bacR (t-test; p < 0.05). These findings suggest that ARGs and bacR are being removed from the water column through immobilization reactions occurring in the streambed. Additionally, we predicted that the 90 % removal (or D90) of targets occurs within the first 500 m in all model streams except in a slow-flow pastoral stream, which required 1400 m of downstream transport for 90 % removal. Our findings and model stand out as promising tools to predict the fate of ARGs in streams and will contribute to improving and managing agricultural practices that employ animal manure.

3D Multi‐Fluid MHD Simulation of the Early Time Behavior and Cross‐Field Propagation of an Artificial Plasma Cloud in the Bottom Side Ionosphere

AbstractThe relative motion of an artificial plasma stream in an ambient plasma background transverse to the geomagnetic field produces a polarization current and an drift in the direction of the injected photoionized neutral cloud. The polarization current couples the ion cloud momentum to the ambient plasma causing the stream to brake or “skid” before coming to rest. A multi‐fluid five‐moment resistive magnetohydrodynamic (MHD) model has been previously developed and is used in this study to simulate an artificial barium cloud released at 8 km/s in the lower ionosphere transverse to the magnetic field. The MHD model’s governing equations are used to derive 2D analytical solutions to the equations of motion that capture oscillatory behavior relating the cloud ion cyclotron frequency, artificial and ambient plasma density, initial release velocity, and photoionization rate. The analytical solutions are compared to the MHD results for time t &lt; 1 s and previous 1D momentum coupling models. It is shown that motion in the cloud release direction is consistent with previous models while motion in the polarization direction is consistent with numerical results.

New sensitive tools to characterize meta-metabolome response to short- and long-term cobalt exposure in dynamic river biofilm communities

Untargeted metabolomics is a non-a priori analysis of biomolecules that characterizes the metabolome variations induced by short- and long-term exposures to stressors. Even if the metabolite annotation remains lacunar due to database gaps, the global metabolomic fingerprint allows for trend analyses of dose-response curves for hundreds of cellular metabolites. Analysis of dose/time-response curve trends (biphasic or monotonic) of untargeted metabolomic features would thus allow the use of all the chemical signals obtained in order to determine stress levels (defense or damage) in organisms. To develop this approach in a context of time-dependent microbial community changes, mature river biofilms were exposed for 1 month to four cobalt (Co) concentrations (from background concentration to 1 × 10−6 M) in an open system of artificial streams. The meta-metabolomic response of biofilms was compared against a multitude of biological parameters (including bioaccumulation, biomass, chlorophyll a content, composition and structure of prokaryotic and eukaryotic communities) monitored at set exposure times (from 1 h to 28 d). Cobalt exposure induced extremely rapid responses of the meta-metabolome, with time range inducing defense responses (TRIDeR) of around 10 s, and time range inducing damage responses (TRIDaR) of several hours. Even in biofilms whose structure had been altered by Co bioaccumulation (reduced biomass, chlorophyll a contents and changes in the composition and diversity of prokaryotic and eukaryotic communities), concentration range inducing defense responses (CRIDeR) with similar initiation thresholds (1.41 ± 0.77 × 10−10 M Co2+ added in the exposure medium) were set up at the meta-metabolome level at every time point. In contrast, the concentration range inducing damage responses (CRIDaR) initiation thresholds increased by 10 times in long-term Co exposed biofilms. The present study demonstrates that defense and damage responses of biofilm meta-metabolome exposed to Co are rapidly and sustainably impacted, even within tolerant and resistant microbial communities.

Exposure and recovery: The effect of different dilution factors of treated and untreated metal mining effluent on freshwater biofilm function and structure

Abandoned mines generate effluents rich in heavy metals, and these contaminants are released uncontrolled into the nearby aquatic ecosystems, causing severe pollution. However, no real solution exists, leaving a legacy of global pollution. In this study, the efficiency of the treatment technologies in reducing the ecological impacts of mining effluents to freshwater ecosystems with different dilution capacities was tested using biofilm communities as biological indicators. The functional and structural recovery capacity of biofilm communities after 21 days of exposure was assessed. With this aim, we sampled aquatic biofilms from a pristine stream and exposed them to treated (T) and untreated (U) metal mining effluent from Frongoch abandoned mine (Mid Wales, UK). Additionally, we simulated two different flow conditions for the receiving stream: high dilution (HD) and low dilution (LD). After exposure, the artificial streams were filled with artificial water for 14 days to assess the biofilm recovery. Unexposed biofilm served as control for biofilm responses (functional and structural) measured throughout time. During the exposure, short term effects on biofilm functioning (photosynthetic efficiency, nutrient uptake) were observed in T-LD, U-HD, and U-LD, whereas long term effects (community composition, chl-a, and diatom metrics) were observed on the structure of all biofilms exposed to the treated and untreated mining effluent. On the other hand, metal accumulation occurred in biofilms exposed to the mining effluents. However, a functional recovery was observed for all treatments, except in the U-LD in which biofilm structure did not present a significant recovery after the exposure period. The results presented here highlight the need to consider the dilution capacity of the receiving stream to assess the real efficiency of treatment technologies applied to mining effluents to mitigate the ecological impact on freshwater ecosystems.

Smoclust: synthetic minority oversampling based on stream clustering for evolving data streams

Many real-world data stream applications not only suffer from concept drift but also class imbalance. Yet, very few existing studies investigated this joint challenge. Data difficulty factors, which have been shown to be key challenges in class imbalanced data streams, are not taken into account by existing approaches when learning class imbalanced data streams. In this work, we propose a drift adaptable oversampling strategy to synthesise minority class examples based on stream clustering. The motivation is that stream clustering methods continuously update themselves to reflect the characteristics of the current underlying concept, including data difficulty factors. This nature can potentially be used to compress past information without caching data in the memory explicitly. Based on the compressed information, synthetic examples can be created within the region that recently generated new minority class examples. Experiments with artificial and real-world data streams show that the proposed approach can handle concept drift involving different minority class decomposition better than existing approaches, especially when the data stream is severely class imbalanced and presenting high proportions of safe and borderline minority class examples.

An investigation into the use of riverine mesocosms to analyse the effect of flow velocity and recipient textiles on forensic fibre persistence studies

Textile fibre evidence can provide important activity level information in criminal cases. To date, very few studies have investigated fibre persistence on fabrics exposed to aquatic conditions, even though items of evidence and victim’s bodies can regularly be found in aquatic environments. This lack of research on whether fibres (and other trace evidence) persist on evidence submerged in water, has shown to impact practice as it is reported that crime scene examiners do not attempt to recover this evidence, due to the belief that it would not be present. The dynamic nature of aquatic environments mean that the studies are difficult to conduct in situ and variables, such as water flow rate are not possible to control and thought to be difficult to monitor. To address these challenges, artificial streams (also known as mesocosms) were employed in this study to investigate the persistence rate of polyester fibres on different fabric types (Woollen/nylon mix carpet, 100% polyester fleece, and 95% polyester/5% elastane sports vest) for a four week exposure time (1, 8, 24, 48, 120, 168, 264, 336, 504 and 672 hrs). The effect of water flow rate on the persistence of fibres was investigated by conducting the experiment with two flow velocities; 'high' (∼2.75 L/s) or 'low' (∼0.7 L/s). Significant differences between textile type were seen at 504 hrs under low flow conditions and 8, 24, 168 and 264 hrs under high flow conditions. When comparing flow velocities, a significant difference was seen at 1 hr exposure for the fleece textile only, indicating that the two flow rates used in this study do not significantly affect fibre persistence. Initial loss rates were highest for the first hour of submergence for the carpet, fleece and sports vest. Fibre persistence rates were highest on the carpet, followed by fleece and then sports vest. Persistence rates remained mostly constant after 24 hrs for all textiles but with redistribution of fibres between textiles being seen after this exposure time. The use of artificial flumes in this study provided a balance between realistic experimentation and a controlled study; key experimental variables could be continously and safely monitored. This study provides the first fibre persistence data in river type environments and proposes a new method for testing persistence in aquatic environments. This approach is not limited to fibres evidence and could be employed for other evidence such as glass, pollen, fingerprints and DNA.

Numerical Investigation of the Optimum Angle for Open Channel Junction

Numerous natural and artificial streams, including those for irrigation ditches, wastewater treatment facilities, and conveyance structures for fish movement, have open channel confluences. The flow dynamics at and around the junction are intricate; in particular, immediately downstream of the junction, the flow creates a zone of separation on the inner wall along with secondary recirculation patterns. The structure of this complicated flow depends on several factors, including the flow rates in both channels, the angle of confluence, the geometry of the channels, including the longitudinal slope and bed discordance, the roughness of the boundary, and the intensity of the turbulence. It also has a significant impact on bed erosion, bank scouring, etc. The objective of the current work is to calculate the velocity profile and the separation zone dimensions for four angles (30o, 45o, 60o, and 75o) through the simulation process, and the best angle using a three-dimensional model. This work gives a detailed application of the numerical solution (Finite Volume) via Flow 3D software. Results for two flow discharge ratios, q*=0.250 and q*=0.750 were shown; the numerical model and the experimental results agreed well. The findings are consistent with past research and demonstrate how the main channel flow pattern is affected by changes in the channel crossing angle, as well as how greater separation zones are produced in the main channel when the flow discharge ratio q* (main channel flow divided by total flow) is smaller. Analysis revealed that the separation zone's smallest diameter will be at the 75ocrossing angle. Doi: 10.28991/CEJ-2023-09-05-07 Full Text: PDF

Eukaryotes contribute more than bacteria to the recovery of freshwater ecosystem functions under different drought durations.

Global climate change mostly impacts river ecosystems by affecting microbial biodiversity and ecological functions. Considering the high functional redundancy of microorganisms, the unknown relationship between biodiversity and ecosystem functions obstructs river ecological research, especially under the influence of increasing weather extremes, such as in intermittent rivers and ephemeral streams (IRES). Herein, dry-wet alternation experiments were conducted in artificial stream channels for 25 and 90 days of drought, both followed by 20 days of rewetting. The dynamic recovery of microbial biodiversity and ecosystem functions (represented by ecosystem metabolism and denitrification rate) were determined to analyse biodiversity-ecosystem-function (BEF) relationships after different drought durations. There was a significant difference between bacterial and eukaryotic biodiversity recovery after drought. Eukaryotic biodiversity was more sensitive to drought duration than bacterial, and the eukaryotic network was more stable under dry-wet alternations. Based on the establishment of partial least squares path models, we found that eukaryotic biodiversity has a stronger effect on ecosystem functions than bacteria after long-term drought. Indeed, this work represents a significant step forward for further research on the ecosystem functions of IRES, especially emphasizing the importance of eukaryotic biodiversity in the BEF relationship.

Effects of combining flow intermittency and exposure to emerging contaminants on the composition and metabolic response of streambed biofilm bacterial communities

Freshwater ecosystems are characterised by the co-occurrence of stressors that simultaneously affect the biota. Among these, flow intermittency and chemical pollution severely impair the diversity and functioning of streambed bacterial communities. Using an artificial streams mesocosm facility, this study examined how desiccation and pollution caused by emerging contaminants affect the composition of stream biofilm bacterial communities, their metabolic profiles, and interactions with their environment. Through an integrative analysis of the composition of biofilm communities, characterization of their metabolome and composition of the dissolved organic matter, we found strong genotype-to-phenotype interconnections. The strongest correlation was found between the composition and metabolism of the bacterial community, both of which were influenced by incubation time and desiccation. Unexpectedly, no effect of the emerging contaminants was observed, which was due to the low concentration of the emerging contaminants and the dominant impact of desiccation. However, biofilm bacterial communities modified the chemical composition of their environment under the effect of pollution. Considering the tentatively identified classes of metabolites, we hypothesised that the biofilm response to desiccation was mainly intracellular while the response to chemical pollution was extracellular. The present study demonstrates that metabolite and dissolved organic matter profiling may be effectively integrated with compositional analysis of stream biofilm communities to yield a more complete picture of changes in response to stressors.

Periphyton Phosphorus Uptake in Response to Dynamic Concentrations in Streams: Assimilation and Changes to Intracellular Speciation.

Effective modeling and management of phosphorus (P) losses from landscapes to receiving waterbodies requires an adequate understanding of P retention and remobilization along the terrestrial-aquatic continuum. Within aquatic ecosystems, the stream periphyton can transiently store bioavailable P through uptake and incorporation into biomass during subscouring and baseflow conditions. However, the capacity of stream periphyton to respond to dynamic P concentrations, which are ubiquitous in streams, is largely unknown. Our study used artificial streams to impose short periods (48 h) of high SRP concentration on stream periphyton acclimated to P scarcity. We examined periphyton P content and speciation through nuclear magnetic resonance spectroscopy to elucidate the intracellular storage and transformation of P taken up across a gradient of transiently elevated SRP availabilities. Our study demonstrates that the stream periphyton not only takes up significant quantities of P following a 48-h high P pulse but also sustains supplemental growth over extended periods of time (10 days), following the reestablishment of P scarcity by efficiently assimilating P stored as polyphosphates into functional biomass (i.e., phospho-monoesters and phospho-diesters). Although P uptake and intracellular storage approached an upper limit across the experimentally imposed SRP pulse gradient, our findings demonstrate the previously underappreciated extent to which the periphyton can modulate the timing and magnitude of P delivery from streams. Further elucidating these intricacies in the transient storage potential of periphyton highlights opportunities to enhance the predictive capacity of watershed nutrient models and potentially improve watershed P management.

Removal of Reactive Blue 19 dye from synthetic wastewater using UV/H2O2 and UV/Cl advanced oxidation processes

AbstractThe textile and dyeing industries are among the largest water‐consuming and polluting industries in the world. The most important feature of the textile dyeing industry wastewater is its color, due to the use of colored materials. Most of these dye compounds are resistant to conventional purification methods and their biodegradation is very low through secondary purification processes, resulting in incomplete removal. Therefore, selecting the optimal method to remove these color compounds is essential. In this study, we studied the removal of an organic dye contaminant (Reactive Blue dye 19 [RB19]) using advanced oxidation processes (AOPs). For this purpose, ultraviolet (UV) mercury lamps with a wavelength of 254 nm and a voltage of W16 inside a reactor were used as an energy source. The experiments were performed in a collimated beam reactor inside a dark chamber. Two oxidizers, sodium hypochlorite (NaOCl) and hydrogen peroxide (H2O2), were used to remove RB19 from the artificial sewage stream. Removal of RB19 with a concentration of 20 mg/L with variable pH (5, 7, and 9), oxidant concentrations (5, 10, and 20 mg/L), and time (5, 10, 15, and 30 min) were investigated during the processes of photolysis, chemical oxidation (by H2O2 and NaOCl), and UV/NaOCl and UV/H2O2 AOPs. The photolysis process did not remove the RB19. The highest removal efficiencies of RB19 by chemical oxidation processes with NaOCl and H2O2, UV/NaOCl, and UV/H2O2 at optimal conditions (pH = 5, [oxidant] = 20 mg/L, RB19 = 20 mg/L, and radiation intensity of 1005 mJ/cm2) were 64.49%, 0.88%, 99.7%, and 13.31%, respectively. These results indicate that the hydroxyl radical was produced, under optimum conditions, more in the acidic medium; thus, the RB19 removal efficiency was higher in the acidic medium. The combination of UV rays with oxidants resulted in the production of more hydroxyl radicals and increased removal efficiency.

Artificial intelligence in medical practice: current status and future perspectives

Artificial intelligence in medical practice: current status and future perspectives

Small mammals in natural and agricultural lands of Slobozhanshchyna (eastern Ukraine): results of a five-year-long survey

The study of the species composition, biological and ecological features of small mammals was carried out for five years (2018–2022) in the surroundings of the village of Rozsohuvata (Kharkiv Oblast, Ukraine). During the study period, nine species of mouse-like rodents and three species of insectivorans were detected. The re-search area is represented by a ravine-beam system, agrocoenoses, field-protecting forest strips, riparian vege-tation around artificial reservoirs and streams, dry meadows, and pastures. Trappings of small mammals were carried out on thirty lines of traps. During the entire period of trapping, the dominant species was Sylvaemus uralensis and in some seasons Sylvaemus sylvaticus dominated. Starting from 2018, the relative abundance increased in autumn, reached a peak in 2020 and declined in the following years. In general, 2019 was the year of maximum abundance and most species were recorded in the summer of this year. The greatest number of species was found in the ecotone between riparian vegetation and field (10 species), as well as in the area of the multi-grass steppe in the beam (8 species). The eurytopic species include Sylvaemus uralensis and Syl-vaemus sylvaticus, whereas the stenotopic species is Mus spicilegus. According to the obtained abundance scores, only one species—Myodes glareolus—occurs randomly, whereas the rare species here are Mus spicile-gus, Sorex minutus, and Crocidura suaveolens. Six species are non-abundant: Apodemus agrarius, Sylvaemus tauricus, Mus musculus, Microtus levis, Cricetulus migratorius, and Sorex araneus. Only Sylvaemus sylvaticus belongs to common species, and only Sylvaemus uralensis is abundant. It should be mentioned that Cri-cetulus migratorius, which usually belongs to the category of random or rare species, turned out to be non-abundant in the study region, and it was recorded regularly in almost half of the studied biotopes. Myodes glareolus, which usually prefers oak forest, was extremely rare. The abundance of Apodemus agrarius, which usually belongs to common species in river valleys, was also low. Sylvaemus showed greater ecological valence, dominated in abundance and inhabited most biotopes, except for Sylvaemus tauricus, which began to occur only in the last two years and mainly in forest strips.

Persistence of low pathogenic avian influenza virus in artificial streams mimicking natural conditions of waterfowl habitats in the Mediterranean climate

Avian influenza viruses (AIVs) can affect wildlife, poultry, and humans, so a One Health perspective is needed to optimize mitigation strategies. Migratory waterfowl globally spread AIVs over long distances. Therefore, the study of AIV persistence in waterfowl staging and breeding areas is key to understanding their transmission dynamics and optimizing management strategies. Here, we used artificial streams mimicking natural conditions of waterfowl habitats in the Mediterranean climate (day/night cycles of photosynthetic active radiation and temperature, low water velocity, and similar microbiome to lowland rivers and stagnant water bodies) and then manipulated temperature and sediment presence (i.e., 10–13 °C vs. 16–18 °C, and presence vs. absence of sediments). An H1N1 low pathogenic AIV (LPAIV) strain was spiked in the streams, and water and sediment samples were collected at different time points until 14 days post-spike to quantify viral RNA and detect infectious particles. Viral RNA was detected until the end of the experiment in both water and sediment samples. In water samples, we observed a significant combined effect of temperature and sediments in viral decay, with higher viral genome loads in colder streams without sediments. In sediment samples, we didn't observe any significant effect of temperature. In contrast to prior laboratory-controlled studies that detect longer persistence times, infectious H1N1 LPAIV was isolated in water samples till 2 days post-spike, and none beyond. Infectious H1N1 LPAIV wasn't isolated from any sediment sample. Our results suggest that slow flowing freshwater surface waters may provide conditions facilitating bird-to-bird transmission for a short period when water temperature are between 10 and 18 °C, though persistence for extended periods (e.g., weeks or months) may be less likely. We hypothesize that experiments simulating real environments, like the one described here, provide a more realistic approach for assessing environmental persistence of AIVs.

Combined effect of stream drying and nutrient enrichment on macroinvertebrate community: experimental study from artificial stream mesocosms

Combined effect of stream drying and nutrient enrichment on macroinvertebrate community: experimental study from artificial stream mesocosms

Online AutoML: an adaptive AutoML framework for online learning

Automated Machine Learning (AutoML) has been used successfully in settings where the learning task is assumed to be static. In many real-world scenarios, however, the data distribution will evolve over time, and it is yet to be shown whether AutoML techniques can effectively design online pipelines in dynamic environments. This study aims to automate pipeline design for online learning while continuously adapting to data drift. For this purpose, we design an adaptive Online Automated Machine Learning (OAML) system, searching the complete pipeline configuration space of online learners, including preprocessing algorithms and ensembling techniques. This system combines the inherent adaptation capabilities of online learners with the fast automated pipeline (re)optimization capabilities of AutoML. Focusing on optimization techniques that can adapt to evolving objectives, we evaluate asynchronous genetic programming and asynchronous successive halving to optimize these pipelines continually. We experiment on real and artificial data streams with varying types of concept drift to test the performance and adaptation capabilities of the proposed system. The results confirm the utility of OAML over popular online learning algorithms and underscore the benefits of continuous pipeline redesign in the presence of data drift.

Automated Design of Salient Object Detection Algorithms with Brain Programming

Despite recent improvements in computer vision, artificial visual systems’ design is still daunting since an explanation of visual computing algorithms remains elusive. Salient object detection is one problem that is still open due to the difficulty of understanding the brain’s inner workings. Progress in this research area follows the traditional path of hand-made designs using neuroscience knowledge or, more recently, deep learning, a particular branch of machine learning. Recently, a different approach based on genetic programming appeared to enhance handcrafted techniques following two different strategies. The first method follows the idea of combining previous hand-made methods through genetic programming and fuzzy logic. The second approach improves the inner computational structures of basic hand-made models through artificial evolution. This research proposes expanding the artificial dorsal stream using a recent proposal based on symbolic learning to solve salient object detection problems following the second technique. This approach applies the fusion of visual saliency and image segmentation algorithms as a template. The proposed methodology discovers several critical structures in the template through artificial evolution. We present results on a benchmark designed by experts with outstanding results in an extensive comparison with the state of the art, including classical methods and deep learning approaches to highlight the importance of symbolic learning in visual saliency.