Prediction Mistakes Research Articles

Identifying Prediction Mistakes in Observational Data

Identifying Prediction Mistakes in Observational Data

Estimating the Potential of Biogas Yield from Anaerobic Co-digestion of Organic Waste with Ensemble Machine learning

The biogas has the potential to serve as a substitute as fossil derived fuel. The present study investigates anaerobic co-digestion of organic waste for predictive modelling. Several co-digestion studies were performed with different pH, solid concentration, temperature, and co-digestion ratios. A water displacement apparatus was used to test biogas yield, and data was collected meticulously. Linear regression (LR) and Random Forest (RF) based models were built with Python-based tools and tested using statistical measures for the prediction of biogas yield. The LR showed a strong linear association, with R and R2 values of 0.9892 and 0.9785, respectively. However, RF surpassed LR, with higher R and R2 values of 0.9919 and 0.9826, respectively. Furthermore, RF had lower MSE and MAE values, indicating higher prediction accuracy and precision. RF consistently scored well in tests, demonstrating its ability to capture complicated relationships while minimizing prediction mistakes. Taylor's diagrams further demonstrated RF's excellent performance during both the training and testing periods. Overall, RF emerges as the optimum model for reliably estimating biogas output in anaerobic co-digestion systems, with important implications for waste-to-energy processes.

Nonempirical fractal permeability model and experimental verification of hydrate-bearing sediments based on heterogeneous distribution of particles

The one of the factors limiting production of gas-water is affected by sediments' permeability. Evaluating the normalized permeability (Kn) of sediments containing hydrates is necessary for analyzing the saturation (decomposition level) and flow capacity. The mistakes in Kn prediction are frequently brought about by the assumption such as homogenous particles distribution and ignoring microscopic pore structure. To address the issue, a fractal-based nonempirical prediction model of Kn was developed by modeling, parameter acquisition using SEM and Analyzer, experiment validation, error comparison with published data, and sensitivity analysis for evolution, respectively. Assuming that sediment skeleton particles were misaligned spherical particles of equal diameter. The heterogeneity of sediments was described by the vertical and horizontal distance ratio (m) and offset angle (θ) of the particles. This led to a novel approach for determining average tortuosity (τav,0) and tortuosity fractal dimension (Dτ,0). The viability of the model was confirmed by permeability measurement experiments. Additionally, using the measuring function of SEM, the maximum pore diameter of sediments was obtained. Furthermore, a comparison with earlier data demonstrated the effect. Considering heterogeneity, the average accuracy of Geometric Mean Variance in Kn has increased to 9.66%. Finally, each parameter's impact on prediction was examined. At m = 1.357, Dτ,0 grows symmetrically. Kn,GC varies 15 times more with an increase in Dτ,0 than Kn,PF. All findings indicate that the new model is more likely to be forecasted Kn for a variety of systems, including clayey-silt, sand, and silty-sand system, and especially in reflecting the heterogeneity.

Thermally efficient gas turbine with pressure drop-based automated filter cleaning and optimized fuel control system

Optimizing elements like fuel specifications and inlet–outlet conditions is essential for enhancing gas turbine performance. Prior research optimized filter house pressure drop using statistical methods, but these methods had drawbacks related to measurement errors. Additionally, with changing air conditions, periodic-based cleaning for self-cleaning pulse filters at the turbine inlet proved unsuccessful, making the procedure difficult and time-consuming. A novel “Thermally Efficient Gas Turbine with Pressure Drop-Based Automated Filter Cleaning and Optimized Fuel Control System” has been suggested to improve gas turbine performance. To handle pressure drop effectively, this cutting-edge system adds a predictive Apriori-based parallel auto-cleaning pulse filter, with automatic cleaning activated at a predetermined pressure difference. An adaptive neuro-fuzzy inference system-multi-layer perceptron integrated simulated annealing-Levenberg Marquardt algorithm has been created to handle mistakes in prediction and improve turbine parameters for greater energy efficiency. This complete strategy considerably increases energy efficiency while also improving air pressure drop prediction. A 200 MW power output, 70 % energy efficiency, 1.3 kg/kW-h fuel consumption, a small residual error (−0.4 mbar), and a pressure difference of only 3 mbar are just a few of the model's excellent performance indicators.

Archaeological site segmentation of ancient city walls based on deep learning and LiDAR remote sensing

Ancient city walls, one of the most notable distinguishing features of Chinese ancient cities, are military defenses constructed of rammed earth. The ancient city walls have considerable study value because they served as the city's boundary and a symbol of power. However, as a result of natural erosion and human activities, many sites have been ruined. Existing optical remote sensing technologies, LiDAR point cloud processing algorithms, and deep learning methods are inadequate for the extraction and segmentation of ancient city wall sites. The novel semantic segmentation method for ancient city wall sites is described in this paper that extracts sites at the pixel level from LiDAR remote sensing data based on deep learning. To begin, the point cloud data collected by airborne laser scanning is processed into DEM data, and the distribution of ancient city walls in the study area is obtained through archaeological survey and expert interpretation. The dataset for deep learning semantic segmentation is then generated using image cropping and data augmentation techniques. Third, implement a U-Net semantic segmentation framework for microtopographic archaeological sites, and predict ancient city wall sites in the testing region after model training. Finally, the deep learning results are optimized using the connected component analysis method, and prediction mistakes such as holes and noise are removed. Taking Jinancheng, the capital city of the Chu kingdom, as an example, the proposed method process can identify and extract the ancient city wall sites at the pixel level, where the evaluation metrics reach 94.12% (Precision) and 81.38% (IoU). The experiment results are excellent due to improvement strategies in the dataset generation, model training, and post-processing steps. Thus, this study is significant for the current survey and protection of ancient city wall sites. The source code will be freely available at https://github.com/wshunli/Open-CHAI-CityWalls.

Bet-hedging via dispersal aids the evolution of plastic responses to unreliable cues.

Adaptive plasticity is expected to evolve when informative cues predict environmental variation. However, plastic responses can be maladaptive even when those cues are informative, if prediction mistakes are shared across members of a generation. These fitness costs can constrain the evolution of plasticity when initial plastic mutants use of cues of only moderate reliability. Here, we model the barriers to the evolution of plasticity produced by these constraints and show that dispersal across a metapopulation can overcome them. Constraints are also lessened, though not eliminated, when plastic responses are free to evolve gradually and in concert with increased reliability. Each of these factors be viewed as a form of bet-hedging: by lessening correlations in the fates of relatives, dispersal acts as diversifying bet-hedging, while producing submaximal responses to a cue can be understood as a conservative bet-hedging strategy. While poor information may constrain the evolution of plasticity, the opportunity for bet-hedging may predict when that constraint can be overcome.

Impact of using a predictive neural network of multi-term zenith angle function on energy management of solar-harvesting sensor nodes

Abstract Using the Neural Networks to predict solar harvestable energy would contribute to prolonging the duration of the effective operation and thus less consumption in solar-harvesting sensor nodes. The NNs with higher prediction accuracy have the longest effective operation. Till now, the NNs that use the zenith angle function as input have been utilized with only two terms. This paper shows the advantages of using a multi-term zenith angle function on the energy management in the nodes. To this end, this paper considers two, three, and four terms for the function of the zenith angle. The results showed that the case of four terms has the lowest prediction mistakes on average (0.83%) compared to (2.13% and 1.75%) for the cases of two and three terms, respectively. This is followed by a reduction in energy consumption in favor of four terms case. For one month simulation period with hourly prediction, the sensor node worked at the higher consumption mode (M2) in the case of four terms 4 hours less than three terms and 7 hours less than two terms case. Thus, increasing the number of terms in the zenith angle function leads to higher accuracy and less energy consumption.

Subpopulation-specific machine learning prognosis for underrepresented patients with double prioritized bias correction

BackgroundMany clinical datasets are intrinsically imbalanced, dominated by overwhelming majority groups. Off-the-shelf machine learning models that optimize the prognosis of majority patient types (e.g., healthy class) may cause substantial errors on the minority prediction class (e.g., disease class) and demographic subgroups (e.g., Black or young patients). In the typical one-machine-learning-model-fits-all paradigm, racial and age disparities are likely to exist, but unreported. In addition, some widely used whole-population metrics give misleading results.MethodsWe design a double prioritized (DP) bias correction technique to mitigate representational biases in machine learning-based prognosis. Our method trains customized machine learning models for specific ethnicity or age groups, a substantial departure from the one-model-predicts-all convention. We compare with other sampling and reweighting techniques in mortality and cancer survivability prediction tasks.ResultsWe first provide empirical evidence showing various prediction deficiencies in a typical machine learning setting without bias correction. For example, missed death cases are 3.14 times higher than missed survival cases for mortality prediction. Then, we show DP consistently boosts the minority class recall for underrepresented groups, by up to 38.0%. DP also reduces relative disparities across race and age groups, e.g., up to 88.0% better than the 8 existing sampling solutions in terms of the relative disparity of minority class recall. Cross-race and cross-age-group evaluation also suggests the need for subpopulation-specific machine learning models.ConclusionsBiases exist in the widely accepted one-machine-learning-model-fits-all-population approach. We invent a bias correction method that produces specialized machine learning prognostication models for underrepresented racial and age groups. This technique may reduce potentially life-threatening prediction mistakes for minority populations.

Adversarial Learning from Crowds

Learning from Crowds (LFC) seeks to induce a high-quality classifier from training instances, which are linked to a range of possible noisy annotations from crowdsourcing workers under their various levels of skills and their own preconditions. Recent studies on LFC focus on designing new methods to improve the performance of the classifier trained from crowdsourced labeled data. To this day, however, there remain under-explored security aspects of LFC systems. In this work, we seek to bridge this gap. We first show that LFC models are vulnerable to adversarial examples---small changes to input data can cause classifiers to make prediction mistakes. Second, we propose an approach, A-LFC for training a robust classifier from crowdsourced labeled data. Our empirical results on three real-world datasets show that the proposed approach can substantially improve the performance of the trained classifier even with the existence of adversarial examples. On average, A-LFC has 10.05% and 11.34% higher test robustness than the state-of-the-art in the white-box and black-box attack settings, respectively.

Learning-Based Synthesis of Social Laws in STRIPS

In a multi-agent environment, each agent must take into account not only the actions it must perform to achieve its goals, but also the behavior of other agents in the system, which usually requires some sort of coordination between the agents. One way to avoid the complexity of centralized planning and online negotiation between agents is to design an artificial social system. This system enacts a social law that restricts the behavior of the agents. A robust social law enables the agents to reach their goals while keeping them from interfering with each other. However, the problem of efficient synthesis of such laws is computationally hard, and previously proposed search techniques do not scale well. In this paper, we propose the use of graph neural networks to predict social laws from a graph-based representation of multi-agent systems. However, as this prediction can be wrong, we use heuristic search to correct possible mistakes in the network's prediction ensuring that the produced social law is indeed robust. Our empirical evaluation shows that this approach beat the previous state-of-the-art in social law synthesis, and that is can learn from an imperfect expert, even in the presence of noise.

A note on the price of bandit feedback for mistake-bounded online learning

The standard model and the bandit model are two generalizations of the mistake-bound model to online multiclass classification. In both models the learner guesses a classification in each round, but in the standard model the learner receives the correct classification after each guess, while in the bandit model the learner is only told whether or not their guess is correct in each round. For any set F of multiclass classifiers, define optstd(F) and optbandit(F) to be the optimal worst-case number of prediction mistakes in the standard and bandit models respectively.Long (Theoretical Computer Science, 2020) claimed that for all M>2 and infinitely many k, there exists a set F of functions from a set X to a set Y of size k such that optstd(F)=M and optbandit(F)≥(1−o(1))(|Y|ln⁡|Y|)optstd(F). The proof of this result depended on the following lemma, which is false e.g. for all prime p≥5, s=1 (the all 1 vector), t=2 (the all 2 vector), and all z.Lemma: Fix n≥2 and prime p, and let u be chosen uniformly at random from {0,…,p−1}n. For any s,t∈{1,…,p−1}n with s≠t and for any z∈{0,…,p−1}, we have Pr⁡(t⋅u=zmodp | s⋅u=zmodp)=1p.We show that this lemma is false precisely when s and t are multiples of each other mod p. Then using a new lemma, we fix Long's proof.

Do children learn from their prediction mistakes? A registered report evaluating error-based theories of language acquisition.

Error-based theories of language acquisition suggest that children, like adults, continuously make and evaluate predictions in order to reach an adult-like state of language use. However, while these theories have become extremely influential, their central claim—that unpredictable input leads to higher rates of lasting change in linguistic representations—has scarcely been tested. We designed a prime surprisal-based intervention study to assess this claim. As predicted, both 5- to 6-year-old children (n = 72) and adults (n = 72) showed a pre- to post-test shift towards producing the dative syntactic structure they were exposed to in surprising sentences. The effect was significant in both age groups together, and in the child group separately when participants with ceiling performance in the pre-test were excluded. Secondary predictions were not upheld: we found no verb-based learning effects and there was only reliable evidence for immediate prime surprisal effects in the adult, but not in the child group. To our knowledge, this is the first published study demonstrating enhanced learning rates for the same syntactic structure when it appeared in surprising as opposed to predictable contexts, thus providing crucial support for error-based theories of language acquisition.

Wasserstein Loss With Alternative Reinforcement Learning for Severity-Aware Semantic Segmentation

Semantic segmentation is important for many real-world systems, e.g., autonomous vehicles, which predict the class of each pixel. Recently, deep networks achieved significant progress w.r.t. the mean Intersection-over Union (mIoU) with the cross-entropy loss. However, the cross entropy loss can essentially ignore the difference of severity for an autonomous car with different wrong prediction mistakes. For example, predicting the car to the road is much more servery than recognize it as the bus. Targeting for this difficulty, we develop a Wasserstein training framework to explore the inter-class correlation by defining its ground metric as misclassification severity. The ground metric of Wasserstein distance can be pre-defined following the experience on a specific task. From the optimization perspective, we further propose to set the ground metric as an increasing function of the pre-defined ground metric. Furthermore, an adaptively learning scheme of the ground matrix is proposed to utilize the high-fidelity CARLA simulator. Specifically, we follow a reinforcement alternative learning scheme. The experiments on both CamVid and Cityscapes datasets evidenced the effectiveness of our Wasserstein loss. The SegNet, ENet, FCN and Deeplab networks can be adapted following a plug in manner. We achieve significant improves on the predefined important classes, and much longer continuous play time in our simulator.

Uncertainty‐assisted deep vision structural health monitoring

AbstractComputer vision leveraging deep learning has achieved significant success in the last decade. Despite the promising performance of the existing deep vision inspection models, the extent of models’ reliability remains unknown. Structural health monitoring (SHM) is a crucial task for the safety and sustainability of structures, and thus, prediction mistakes can have fatal outcomes. In this paper, we use Bayesian inference for deep vision SHM models where uncertainty can be quantified using the Monte Carlo dropout sampling. Three independent case studies for cracks, local damage identification, and bridge component detection are investigated using Bayesian inference. Aside from better prediction results, the two uncertainty metrics, variations in softmax probability and entropy, are shown to have good correlations with misclassifications. However, modifying the decision or triggering human intervention can be challenging based on raw uncertainty outputs. Therefore, the concept of surrogate models is proposed to develop the models for uncertainty‐assisted segmentation and prediction quality tagging. The former refines the segmentation mask and the latter is used to trigger human interventions. The proposed framework can be applied to future deep vision SHM frameworks to incorporate model uncertainty in the inspection processes.

Non-word Attributes’ Efficiency in Text Mining Authorship Prediction

Abstract Literature scripts can be compared to paintings, in an artistic way as well as in the perspective of financial value, whereas the value of these scripts rise and fall depending on their author’s popularity. Authors’ scripts represent a specific style of writing that can be measured and compared using a text mining field called Stylometric. Stylometric analysis depends on some features called authorship attributes, and these attributes or features can be used in special algorithms and methods to reach that aim. Generally, each method selected in the Stylometric field uses a variety of attributes to reach higher prediction accuracy. The aim of this research is to improve the accuracy of authorship prediction in literary works based on the artistic writing style of the authors. To achieve that, a new set of attributes will be used with the Stylometric Authorship Balanced Attribution method, which was chosen in this research among several other machine language methods because of its delicateness in authorship prediction projects. The attributes that have been used by most of the researchers were word frequencies (single word, pair of words, or trio of words), which led to some prediction mistakes. In this research, a new set of attributes is used to decrease these mistakes. These proposed non-word attributes are named sentence length, special characters, and punctuation symbols. The results obtained by using these proposed attributes were excellent.

An Online Multiview Learning Algorithm for PolSAR Data Real-Time Classification

Polarimetric synthetic aperture radar (PolSAR) data are sequentially acquired and usually large scale. Fast and accurate classification is particularly important for their applications. By introducing online learning, the PolSAR system can learn a classification model incrementally from a stream of instances, which is of high efficiency for newly arrived samples processing, strong adaptability for a dynamically changing environment, and excellent scalability for rapidly increasing data. In this paper, we propose an Online Multi-view Passive-Aggressive learning algorithm, named OMPA, for PolSAR data real-time classification. The polarimetric, color, and texture features are extracted to characterize PolSAR data, and each type of features corresponds to one view. In order to exploit the consistency and complementary property of these views, we give a new optimization model that ensembles the classifiers of multiple distinct views and enforces the agreement between each predictor and the combined predictor. The corresponding algorithms for both binary and multiclass classification tasks are derived, and the update steps have analytical solutions. In addition, we rigorously derive a bound on the number of prediction mistakes of the method. The proposed OMPA algorithm is evaluated on two real PolSAR datasets for built-up areas extraction and land cover classification, respectively. Experimental results demonstrate that OMPA consistently maintains a smaller mistake rate with low time cost and achieves about 1% and 2% accuracy improvements on the datasets, respectively, compared with the best results of the previously known online single-view and multiview learning methods.

Goodness of fit and model selection criteria for time series models

Time series is an important part of decision making and several results are constructed on estimates of forthcoming events. In the Time series process, since, for the coming actions include probability; the predictions are frequently not perfect. The objectives of the Time series are to decrease the prediction error; to create predictions that are rarely improper and that have minor prediction mistakes. In selecting a suitable Time series model, the researcher wants to be responsive that numerous altered models may have comparable properties. A good model will fit the data well. The goodness of fit recovers as additional limits are involved in the model. This arises a problem with the ARMA (p, q) models, because p and q take low values. It should be noted that the best model fit need not imply to provide best forecasts. There exists several model selection criteria that trade off a decrease in the sum of squares of the errors for a more sparing model. A goodness of fit criterion for ARMA (p, q) model and modified selection criteria for Time series models have been suggested in the present study.

Strategizing With Others’ Misperceptions of Luck in Extreme Performances

Research suggests that people tend to be fooled by randomness and mistake luck for skill, particularly when evaluating extreme performances. I argue that these predictable mistakes can be translated into a source of competitive advantage: informed managers can exploit others’ misperceptions of luck, such as by arbitraging in strategic factor markets. I then discuss limits to this arbitrage strategy due to social constraints: stakeholders may not be able to accurately evaluate performances and may disapprove atypical strategic activities, suggesting that only strategists who are less sensitive to this “lemon problem” can take advantage of the resulting arbitrage opportunities. I conclude with a flowchart about when strategists should pursue this alternative source of strategic opportunity that turns the well- known biases on their head.

The Fragmented Regulation of Investment Advice: A Call for Harmonization

Decades of short-term thinking and regulatory fixes created the bewilderingly complex statutory and regulatory structures governing the giving of personalized investment advice to retail customers. Although deeply flawed, the current systems remain entrenched because of the difficulties inherent in making radical alterations. Importantly, the current patchwork systems do not seem to serve retail customers particularly well. Retail customers tend to make predictable and costly mistakes in allocating their assets. Some of this occurs because many investors lack basic financial literacy. A recent study released by the staff of the Securities and Exchange Commission (the “Commission”) on financial literacy among investors (the “Literacy Study”) highlights some frightening findings. The Literacy Study documents that many investors struggle to protect themselves against fraud and do not understand basic concepts such as diversification, investment costs, inflation, or compound interest. Certain subgroups, “including women, African-Americans, Hispanics, the oldest segment of the elderly population, and those who are poorly educated, have an even greater lack of investment knowledge than the general population.” Much has been said about the role of financial literacy in protecting investors. Investors may also be protected through legislation and regulation. Following the Great Depression, the federal securities laws were enacted for precisely this purpose. But over time, the distinctions that once existed between the different professionals offering different forms of investment advice have disappeared, blurring the lines between existing regulatory structures.

Professionally biased: Misestimations of driving speed, journey time and time-savings among taxi and car drivers

AbstractPeople make systematic and predictable mistakes regarding estimations of average speed and journey time. In addition, people have been shown to commit a time-saving bias by underestimating the time that can be saved when increasing from a low speed and overestimating the time that can be saved when increasing from a relatively high speed. These misestimations have been shown to relate to biases in judgments of the speed required to arrive at a specific time and to choosing unduly high speed. Professional drivers, such as taxi drivers, might be less susceptible to these biases due to their increased driving experience. In the current study, we interviewed taxi drivers about a journey they were currently making and examined their estimations of journey time, average speed and time savings. Compared to a group of non-professional car drivers, taxi drivers showed the same considerable misestimations of driving speed, journey time and time savings as non-professionals. However, overestimations of time savings among taxi drivers were smaller than those made by car drivers. We discuss the practical significance of these findings.