In this article, we propose a novel ensemble technique with a multi-scheme weighting based on a technique called coopetitive soft gating. This technique combines both, ensemble member competition and cooperation, in order to maximize the overall forecasting accuracy of the ensemble. The proposed algorithm combines the ideas of multiple ensemble paradigms (power forecasting model ensemble, weather forecasting model ensemble, and lagged ensemble) in a hierarchical structure. The technique is designed to be used in a flexible manner on single and multiple weather forecasting models, and for a variety of lead times. We compare the technique to other power forecasting models and ensemble techniques with a flexible number of weather forecasting models, which can have the same, or varying forecasting horizons. It is shown that the model is able to outperform those models on a number of publicly available data sets. The article closes with a discussion of properties of the proposed model which are relevant in its application. Keywords: Ensemble techniques, Power forecasting, Multi model ensembles, Combining forecasts, Model selection, Time series, Data mining 1. Introduction During the past decade, there has been a tremendous growth of the installed capacity of various forms of renewable energy generation. Wind turbines and photovoltaic powerplants contribute substantially to the new mix of energy, which consists of both nonrenewable and renewable energy power plants. Most renewable energy sources have intermittent generation characteristics, i.e., the amount of generated power highly depends on the weather situation and it cannot be regulated the way it is possible with traditional power plants. In order to guarantee grid stability, the power generation and load in the grid have to be balanced, as the intermediate storage of electrical energy is both inefficient and expensive. Intelligent Embedded Systems Homepage: http://www. Depending on the forecasting horizon, the forecast is of interest to different actors in the field, e.g., network operators, power plant operators, or electricity traders. Having an accurate power forecast, the technical and financial risks for all market participants can be reduced. The power forecasting process typically takes place in two steps: 1. A meteorological forecast for the desired area (the location of the renewable energy power plant) is computed. This forecast is called numerical weather prediction (NWP). In this article, we focus on the second step of the forecasting process, i.e., we assume the NWP as given.

Accurate demand forecasts can help on-line retail organizations better plan their supply-chain processes. The challenge, however, is the large number of associative factors that result in large, non-stationary shifts in demand, which traditional time series and regression approaches fail to model. In this paper, we propose a Neural Network architecture called AR-MDN, that simultaneously models associative factors, time-series trends and the variance in the demand. We first identify several causal features and use a combination of feature embeddings, MLP and LSTM to represent them. We then model the output density as a learned mixture of Gaussian distributions. The AR-MDN can be trained end-to-end without the need for additional supervision. We experiment on a dataset of an year's worth of data over tens-of-thousands of products from Flipkart. The proposed architecture yields a significant improvement in forecasting accuracy when compared with existing alternatives.

This farm in Arkansas may soon be the most scientifically advanced farm in the world. There's a farm in Arkansas growing soybeans, corn, and rice that is aiming to be the most scientifically advanced farm in the world. Soil samples are run through powerful machines to have their microbes genetically sequenced, drones are flying overhead taking hyperspectral images of the crops, and soon supercomputers will be crunching the massive volumes of data collected. Scientists at the Department of Energy's Lawrence Berkeley National Laboratory (Berkeley Lab), working with the University of Arkansas and Glennoe Farms, hope this project, which brings together molecular biology, biogeochemistry, environmental sensing technologies, and machine learning, will revolutionize agriculture and create sustainable farming practices that benefit both the environment and farms. If successful, they envision being able to reduce the need for chemical fertilizers and enhance soil carbon uptake, thus improving the long-term viability of the land, while at the same time increasing crop yields.

Tyler Hamilton works with cleantech companies from across Canada as an adviser with the non-profit MaRS Discovery District in Toronto. "We're ingesting hundreds of terabytes of satellite imagery data, and we've got thousands of sensors in fields, all providing us information on a per-second, per-minute or daily basis," says Kevin Grant, the company's chief technology officer. That information, he says, is gathered from thousands of farms across Canada, the United States and overseas, and includes everything from soil and weather conditions to the performance of farm equipment and evidence of pests and rot. Sometimes it comes from Farmers Edge's own weather stations and in-field sensors; sometimes from third-party databases. The end game is to increase crop yield and lower costs by helping farmers make better decisions about when and exactly where to plant, irrigate, fertilize and harvest, and by allowing for the pro-active maintenance of equipment.

Rearranging objects on a tabletop surface by means of nonprehensile manipulation is a task which requires skillful interaction with the physical world. Usually, this is achieved by precisely modeling physical properties of the objects, robot, and the environment for explicit planning. In contrast, as explicitly modeling the physical environment is not always feasible and involves various uncertainties, we learn a nonprehensile rearrangement strategy with deep reinforcement learning based on only visual feedback. For this, we model the task with rewards and train a deep Q-network. Our potential field-based heuristic exploration strategy reduces the amount of collisions which lead to suboptimal outcomes and we actively balance the training set to avoid bias towards poor examples. Our training process leads to quicker learning and better performance on the task as compared to uniform exploration and standard experience replay. We demonstrate empirical evidence from simulation that our method leads to a success rate of 85%, show that our system can cope with sudden changes of the environment, and compare our performance with human level performance.

Crowdsourcing is a strategy to categorize data through the contribution of many individuals. A wide range of theoretical and algorithmic contributions are based on the model of Dawid and Skene [1]. Recently it was shown in [2,3] that, in certain regimes, belief propagation is asymptotically optimal for data generated from the Dawid-Skene model. This paper is motivated by this recent progress. We analyze the dense limit of the Dawid-Skene model. It is shown that it belongs to a larger class of low-rank matrix estimation problems for which it is possible to express the asymptotic, Bayes-optimal, performance in a simple closed form. In the dense limit the mapping to a low-rank matrix estimation problem provides an approximate message passing algorithm that solves the problem algorithmically. We identify the regions where the algorithm efficiently computes the Bayes-optimal estimates. Our analysis refines the results of [2,3] about optimality of message passing algorithms by characterizing regions of parameters where these algorithms do not match the Bayes-optimal performance. We further study numerically the performance of approximate message passing, derived in the dense limit, on sparse instances and carry out experiments on a real world dataset.

For the last eight months, farms near Kisarazu City in Japan have been home to a horrifying robot wolf. But don't worry, it wasn't created to terrorize local residents (although, from the looks of the thing, it probably did). Its official name is "Super Monster Wolf," and engineers designed it to stop animals from eating farmers' crops. In truth, the story of the robowolf is more than a little sad. As Motherboard reports, wolves went extinct in Japan in the early 1800s.

Large-scale deep neural networks are both memory intensive and computation-intensive, thereby posing stringent requirements on the computing platforms. Hardware accelerations of deep neural networks have been extensively investigated in both industry and academia. Specific forms of binary neural networks (BNNs) and stochastic computing based neural networks (SCNNs) are particularly appealing to hardware implementations since they can be implemented almost entirely with binary operations. Despite the obvious advantages in hardware implementation, these approximate computing techniques are questioned by researchers in terms of accuracy and universal applicability. Also it is important to understand the relative pros and cons of SCNNs and BNNs in theory and in actual hardware implementations. In order to address these concerns, in this paper we prove that the "ideal" SCNNs and BNNs satisfy the universal approximation property with probability 1 (due to the stochastic behavior). The proof is conducted by first proving the property for SCNNs from the strong law of large numbers, and then using SCNNs as a "bridge" to prove for BNNs. Based on the universal approximation property, we further prove that SCNNs and BNNs exhibit the same energy complexity. In other words, they have the same asymptotic energy consumption with the growing of network size. We also provide a detailed analysis of the pros and cons of SCNNs and BNNs for hardware implementations and conclude that SCNNs are more suitable for hardware.

A new deep-learning algorithm studies aerial photographs after fires to identify damage. How it works: From satellite images taken before and after the California wildfires of 2017, researchers created a data set of buildings that were either damaged or left unscathed. The results: They tweaked a pre-trained ImageNet neural network and got it to spot damaged buildings with an accuracy of up to 85 percent. Why it matters: After a disaster, pinpointing the hardest-hit areas could save lives and help with relief efforts. The researchers also released the data set to the public, which could improve other research that requires satellite images, like conservation and developmental aid work.

FRANKFURT (Reuters) - Utilities can increase their efficiency by using more artificial intelligence (AI) technology, such as software to predict demand swings in the power grid or to control home appliances, consultancy Roland Berger said. European utilities could achieve efficiency gains of up to a fifth over the next five years using such technology, it said, adding that less than a quarter of firms had a strategy to do this. Power firms across Europe, which previously depended on coal or gas-fired power plants, are having to adapt to the expanding use of renewable power sources and facing a profit squeeze as wholesale electricity prices have fallen. "Companies need to respond to this change and come up with new business models," Torsten Henzelmann, partner at Roland Berger, said."To The rise of renewables, such as solar and wind that provide intermittent supply, has increased the need for intelligent IT systems to balance demand and supply swings as companies seek to meet energy and carbon emissions targets, the consultancy said.