Search

Smoke taint in wine has become a critical issue in the wine industry due to its significant negative impact on wine quality. Data-driven approaches including univariate analysis and predictive modeling are applied to a data set containing concentrations of 20 VOCs in 48 grape samples and 56 corresponding wine samples with a taster-evaluated smoke taint index. The resulting models for predicting the smoke taint index of wines are highly predictive when using as inputs VOC concentrations after log conversion in both grapes and wines (Pearson Correlation Coefficient PCC = 0.82; R2 = 0.68) and less so when only grape VOCs are used (Pearson Correlation Coefficient PCC = 0.76; R2 = 0.56), and the classification models also show the capacity for detecting smoke-tainted wines using both wine and grape VOC concentrations (Recall = 0.76; Precision = 0.92; F1 = 0.82) or using only grape VOC concentrations (Recall = 0.74; Precision = 0.92; F1 = 0.80). The performance of the predictive model shows the possibility of predicting the smoke taint index of the wine and grape samples before fermentation. The corresponding code of data analysis and predictive modeling of smoke taint in wine is available in the Github repository (https://github.com/IBPA/smoke_taint_prediction).

We propose here a set of new methods to directly detect light mass dark matter through its scattering with abundant atmospheric muons or accelerator beams. Firstly, we plan to use the free cosmic-ray muons interacting with dark matter in a volume surrounded by tracking detectors, to trace possible interaction between dark matter and muons. Secondly, we will interface our device with domestic or international muon beams. Due to much larger muon intensity and focused beam, we anticipate the detector can be made further compact and the resulting sensitivity on dark matter searches will be improved. Furthermore, we will measure precisely directional distributions of cosmic-ray muons, either at mountain or sea level, and the differences may reveal possible information of dark matter distributed near the earth. Specifically, our methods can have advantages over `exotic' dark matters which are either muon-philic or slowed down due to some mechanism, and sensitivity on dark matter and muon scattering cross section can reach as low as microbarn level., Comment: Added a few sentences to highlight that our methods can have advantages over exotic dark matters which are either muon-philic or slowed down due to some mechanism

In this paper, we present a framework for convolutional coded Poisson receivers (CCPRs) that incorporates spatially coupled methods into the architecture of coded Poisson receivers (CPRs). We use density evolution equations to track the packet decoding process with the successive interference cancellation (SIC) technique. We derive outer bounds for the stability region of CPRs when the underlying channel can be modeled by a $\phi$-ALOHA receiver. The stability region is the set of loads that every packet can be successfully received with a probability of 1. Our outer bounds extend those of the spatially-coupled Irregular Repetition Slotted ALOHA (IRSA) protocol and apply to channel models with multiple traffic classes. For CCPRs with a single class of users, the stability region is reduced to an interval. Therefore, it can be characterized by a percolation threshold. We study the potential threshold by the potential function of the base CPR used for constructing a CCPR. In addition, we prove that the CCPR is stable under a technical condition for the window size. For the multiclass scenario, we recursively evaluate the density evolution equations to determine the boundaries of the stability region. Numerical results demonstrate that the stability region of CCPRs can be enlarged compared to that of CPRs by leveraging the spatially-coupled method. Moreover, the stability region of CCPRs is close to our outer bounds when the window size is large., Comment: Part of this work was presented in 2023 IEEE International Symposium on Information Theory (ISIT) [1] and 2024 IEEE International Symposium on Information Theory (ISIT) [2]

PURPOSE: Current tools in predicting survival outcomes for patients with colon cancer predominantly rely on clinical and pathologic characteristics, but increasing evidence suggests that diet and lifestyle habits are associated with patient outcomes and should be considered to enhance model accuracy. METHODS: Using an adjuvant chemotherapy trial for stage III colon cancer (CALGB 89803), we developed prediction models of disease-free survival (DFS) and overall survival by additionally incorporating self-reported nine diet and lifestyle factors. Both models were assessed by multivariable Cox proportional hazards regression and externally validated using another trial for stage III colon cancer (CALGB/SWOG 80702), and visual nomograms of prediction models were constructed accordingly. We also proposed three hypothetical scenarios for patients with (1) good-risk, (2) average-risk, and (3) poor-risk clinical and pathologic features, and estimated their predictive survival by considering clinical and pathologic features with or without adding self-reported diet and lifestyle factors. RESULTS: Among 1,024 patients (median age 60.0 years, 43.8% female), we observed 394 DFS events and 311 deaths after median follow-up of 7.3 years. Adding self-reported diet and lifestyle factors to clinical and pathologic characteristics meaningfully improved performance of prediction models (c-index from 0.64 [95% CI, 0.62 to 0.67] to 0.69 [95% CI, 0.67 to 0.72] for DFS, and from 0.67 [95% CI, 0.64 to 0.70] to 0.71 [95% CI, 0.69 to 0.75] for overall survival). External validation also indicated good performance of discrimination and calibration. Adding most self-reported favorable diet and lifestyle exposures to multivariate modeling improved 5-year DFS of all patients and by 6.3% for good-risk, 21.4% for average-risk, and 42.6% for poor-risk clinical and pathologic features. CONCLUSION: Diet and lifestyle factors further inform current recurrence and survival prediction models for patients

To understand the clinical risks associated with the transfer of embryos classified as mosaic by preimplantation genetic testing for aneuploidy. Analysis of multi-center data collected between 2017 and 2023. Patients of infertility treatment. Comparison of pregnancies resulting from embryos classified as euploid or mosaic using the 20-80% interval in chromosomal intermediate copy numbers to define a mosaic result. Rates of spontaneous abortion, birth weight, length of gestation, incidence of birth defects, and chromosomal status during gestation. Implanted euploid embryos had significantly lower risk of spontaneous abortion compared to mosaic embryos (8.9% [n=8672; CI95 8.3,9.5] vs. 22.2% [n=914; CI95 19.6,25.0]). Embryos with mosaicism affecting whole chromosomes (not segmental) had the highest risk of spontaneous abortion (27.6% [n=395; CI95 23.2,32.3]). Babies born from euploid, mosaic, and whole chromosome mosaic embryos had average birth weights and lengths of gestation that were not statistically different (3118g and 267 days [n=488; CI95 3067,3169 and 266,268], 3052g and 265 days [n=488; CI95 2993,3112 and 264,267], 3159g and 268 days [n=194; CI95 3070,3249 and 266,270], respectively). Out of 488 babies from mosaic embryo transfers, one had overt gross abnormalities as defined by the Centers of Disease Controls and Prevention. Most prenatal tests performed on pregnancies from mosaic embryo transfers had normal results, and only three pregnancies produced prenatal test results reflecting the mosaicism detected at the embryonic stage (3 out of 250, 1.2%; CI95 0.25,3.5). While embryos classified as mosaic experience higher rates of miscarriage than euploid embryos (with a particularly high frequency shortly after implantation), babies born of mosaic embryo transfers are similar to babies of euploid embryo transfers. Prenatal testing indicates that mosaicism resolves during most pregnancies, although this process is not perfectly efficient. In a small percentage

Artificial intelligence (AI) has brought tremendous impacts on biomedical sciences from academic researches to clinical applications, such as in biomarkers' detection and diagnosis, optimization of treatment, and identification of new therapeutic targets in drug discovery. However, the contemporary AI technologies, particularly deep machine learning (ML), severely suffer from non-interpretability, which might uncontrollably lead to incorrect predictions. Interpretability is particularly crucial to ML for clinical diagnosis as the consumers must gain necessary sense of security and trust from firm grounds or convincing interpretations. In this work, we propose a tensor-network (TN)-ML method to reliably predict lung cancer patients and their stages via screening Raman spectra data of Volatile organic compounds (VOCs) in exhaled breath, which are generally suitable as biomarkers and are considered to be an ideal way for non-invasive lung cancer screening. The prediction of TN-ML is based on the mutual distances of the breath samples mapped to the quantum Hilbert space. Thanks to the quantum probabilistic interpretation, the certainty of the predictions can be quantitatively characterized. The accuracy of the samples with high certainty is almost 100$\%$. The incorrectly-classified samples exhibit obviously lower certainty, and thus can be decipherably identified as anomalies, which will be handled by human experts to guarantee high reliability. Our work sheds light on shifting the ``AI for biomedical sciences'' from the conventional non-interpretable ML schemes to the interpretable human-ML interactive approaches, for the purpose of high accuracy and reliability., Comment: 10 pages, 7 figures

This paper presents a novel approach to network pruning, targeting block pruning in deep neural networks for edge computing environments. Our method diverges from traditional techniques that utilize proxy metrics, instead employing a direct block removal strategy to assess the impact on classification accuracy. This hands-on approach allows for an accurate evaluation of each block's importance. We conducted extensive experiments on CIFAR-10, CIFAR-100, and ImageNet datasets using ResNet architectures. Our results demonstrate the efficacy of our method, particularly on large-scale datasets like ImageNet with ResNet50, where it excelled in reducing model size while retaining high accuracy, even when pruning a significant portion of the network. The findings underscore our method's capability in maintaining an optimal balance between model size and performance, especially in resource-constrained edge computing scenarios.

Vision-language models such as CLIP learn a generic text-image embedding from large-scale training data. A vision-language model can be adapted to a new classification task through few-shot prompt tuning. We find that such a prompt tuning process is highly robust to label noises. This intrigues us to study the key reasons contributing to the robustness of the prompt tuning paradigm. We conducted extensive experiments to explore this property and find the key factors are: 1) the fixed classname tokens provide a strong regularization to the optimization of the model, reducing gradients induced by the noisy samples; 2) the powerful pre-trained image-text embedding that is learned from diverse and generic web data provides strong prior knowledge for image classification. Further, we demonstrate that noisy zero-shot predictions from CLIP can be used to tune its own prompt, significantly enhancing prediction accuracy in the unsupervised setting. The code is available at https://github.com/CEWu/PTNL., Comment: Accepted by ICCV2023

Tax audits are a crucial process adopted in all tax departments to ensure tax compliance and fairness. Traditionally, tax audit leads have been selected based on empirical rules and randomization methods, which are not adaptive, may miss major cases and can introduce bias. Here, we present an audit lead tool based on artificial neural networks that have been trained and evaluated on an integrated dataset of 93,413 unique tax records from 8,647 restaurant businesses over 10 years in the Northern California, provided by the California Department of Tax and Fee Administration (CDTFA). The tool achieved a 40.1% precision and 58.7% recall (F1-score of 0.42) on classifying positive audit leads, and the corresponding regressor provided estimated audit gains (MAE of $155,490). Finally, we evaluated the statistical significance of various empirical rules for use in lead selection, with two out of five being supported by the data. This work demonstrates how data can be leveraged for creating evidence-based models of audit selection and validating empirical hypotheses, resulting in higher audit yields and more fair audit selection processes.

Most existing works on analyzing the performance of a random ensemble of low-density parity-check (LDPC) codes assume that the degree distributions of the two ends of a randomly selected edge are independent. In the paper, we take one step further and consider ensembles of LDPC codes with degree-degree correlations. For this, we propose two methods to construct an ensemble of degree-degree correlated LDPC codes. We then derive a system of density evolution equations for such degree-degree correlated LDPC codes over a binary erasure channel (BEC). By conducting extensive numerical experiments, we show how the degree-degree correlation affects the performance of LDPC codes. Our numerical results show that LDPC codes with negative degree-degree correlation could improve the maximum tolerable erasure probability. Moreover, increasing the negative degree-degree correlation could lead to better unequal error protection (UEP) design., Comment: accepted by the 2023 IEEE International Symposium on Information Theory (ISIT)

In-memory computing is becoming a popular architecture for deep-learning hardware accelerators recently due to its highly parallel computing, low power, and low area cost. However, in-RRAM computing (IRC) suffered from large device variation and numerous nonideal effects in hardware. Although previous approaches including these effects in model training successfully improved variation tolerance, they only considered part of the nonideal effects and relatively simple classification tasks. This paper proposes a joint hardware and software optimization strategy to design a hardware-robust IRC macro for object detection. We lower the cell current by using a low word-line voltage to enable a complete convolution calculation in one operation that minimizes the impact of nonlinear addition. We also implement ternary weight mapping and remove batch normalization for better tolerance against device variation, sense amplifier variation, and IR drop problem. An extra bias is included to overcome the limitation of the current sensing range. The proposed approach has been successfully applied to a complex object detection task with only 3.85\% mAP drop, whereas a naive design suffers catastrophic failure under these nonideal effects., Comment: 10 pages, 18 figures

Self-supervised video representation learning has been shown to effectively improve downstream tasks such as video retrieval and action recognition. In this paper, we present the Cascade Positive Retrieval (CPR) that successively mines positive examples w.r.t. the query for contrastive learning in a cascade of stages. Specifically, CPR exploits multiple views of a query example in different modalities, where an alternative view may help find another positive example dissimilar in the query view. We explore the effects of possible CPR configurations in ablations including the number of mining stages, the top similar example selection ratio in each stage, and progressive training with an incremental number of the final Top-k selection. The overall mining quality is measured to reflect the recall across training set classes. CPR reaches a median class mining recall of 83.3%, outperforming previous work by 5.5%. Implementation-wise, CPR is complementary to pretext tasks and can be easily applied to previous work. In the evaluation of pretraining on UCF101, CPR consistently improves existing work and even achieves state-of-the-art R@1 of 56.7% and 24.4% in video retrieval as well as 83.8% and 54.8% in action recognition on UCF101 and HMDB51. The code is available at https://github.com/necla-ml/CPR., Comment: To appear in CVPR 2022 L3D-IVU Workshop

As the effects of anthropogenic climate change become more severe, several approaches for deliberate climate intervention to reduce or stabilize Earth's surface temperature have been proposed. Solar radiation modification (SRM) is one potential approach to partially counteract anthropogenic warming by reflecting a small proportion of the incoming solar radiation to increase Earth's albedo. While climate science research has focused on the predicted climate effects of SRM, almost no studies have investigated the impacts that SRM would have on ecological systems. The impacts and risks posed by SRM would vary by implementation scenario, anthropogenic climate effects, geographic region, and by ecosystem, community, population, and organism. Complex interactions among Earth's climate system and living systems would further affect SRM impacts and risks. We focus here on stratospheric aerosol intervention (SAI), a well-studied and relatively feasible SRM scheme that is likely to have a large impact on Earth's surface temperature. We outline current gaps in knowledge about both helpful and harmful predicted effects of SAI on ecological systems. Desired ecological outcomes might also inform development of future SAI implementation scenarios. In addition to filling these knowledge gaps, increased collaboration between ecologists and climate scientists would identify a common set of SAI research goals and improve the communication about potential SAI impacts and risks with the public. Without this collaboration, forecasts of SAI impacts will overlook potential effects on biodiversity and ecosystem services for humanity.

As the effects of anthropogenic climate change become more severe, several approaches for deliberate climate intervention to reduce or stabilize Earth's surface temperature have been proposed. Solar radiation modification (SRM) is one potential approach to partially counteract anthropogenic warming by reflecting a small proportion of the incoming solar radiation to increase Earth's albedo. While climate science research has focused on the predicted climate effects of SRM, almost no studies have investigated the impacts that SRM would have on ecological systems. The impacts and risks posed by SRM would vary by implementation scenario, anthropogenic climate effects, geographic region, and by ecosystem, community, population, and organism. Complex interactions among Earth's climate system and living systems would further affect SRM impacts and risks. We focus here on stratospheric aerosol intervention (SAI), a well-studied and relatively feasible SRM scheme that is likely to have a large impact on Earth's surface temperature. We outline current gaps in knowledge about both helpful and harmful predicted effects of SAI on ecological systems. Desired ecological outcomes might also inform development of future SAI implementation scenarios. In addition to filling these knowledge gaps, increased collaboration between ecologists and climate scientists would identify a common set of SAI research goals and improve the communication about potential SAI impacts and risks with the public. Without this collaboration, forecasts of SAI impacts will overlook potential effects on biodiversity and ecosystem services for humanity.

Recent work on enhancing BERT-based language representation models with knowledge graphs (KGs) and knowledge bases (KBs) has yielded promising results on multiple NLP tasks. State-of-the-art approaches typically integrate the original input sentences with KG triples and feed the combined representation into a BERT model. However, as the sequence length of a BERT model is limited, such a framework supports little knowledge other than the original input sentences and is thus forced to discard some knowledge. This problem is especially severe for downstream tasks for which the input is a long paragraph or even a document, such as QA or reading comprehension tasks. We address this problem with Roof-Transformer, a model with two underlying BERTs and a fusion layer on top. One underlying BERT encodes the knowledge resources and the other one encodes the original input sentences, and the fusion layer integrates the two resultant encodings. Experimental results on a QA task and the GLUE benchmark attest the effectiveness of the proposed model.

In this paper, we propose an iterative receiver based on gridless variational Bayesian line spectra estimation (VALSE) named JCCD-VALSE that \emph{j}ointly estimates the \emph{c}arrier frequency offset (CFO), the \emph{c}hannel with high resolution and carries out \emph{d}ata decoding. Based on a modularized point of view and motivated by the high resolution and low complexity gridless VALSE algorithm, three modules named the VALSE module, the minimum mean squared error (MMSE) module and the decoder module are built. Soft information is exchanged between the modules to progressively improve the channel estimation and data decoding accuracy. Since the delays of multipaths of the channel are treated as continuous parameters, instead of on a grid, the leakage effect is avoided. Besides, the proposed approach is a more complete Bayesian approach as all the nuisance parameters such as the noise variance, the parameters of the prior distribution of the channel, the number of paths are automatically estimated. Numerical simulations and sea test data are utilized to demonstrate that the proposed approach performs significantly better than the existing grid-based generalized approximate message passing (GAMP) based \emph{j}oint \emph{c}hannel and \emph{d}ata decoding approach (JCD-GAMP). Furthermore, it is also verified that joint processing including CFO estimation provides performance gain.

Lu-Huai Feng,1 Kun-Peng Bu,1 Shuang Ren,1 Zhenhua Yang,2 Bi-Xun Li,1 Cheng-En Deng3 1Department of Comprehensive Internal Medicine, The Affiliated Tumor Hospital of Guangxi Medical University, Nanning, People’s Republic of China; 2Department of Nephrology, The First Affiliated Hospital of Guangxi Medical University, Nanning, People’s Republic of China; 3Department of Urology, The Affiliated Tumor Hospital of Guangxi Medical University, Nanning, People’s Republic of ChinaCorrespondence: Cheng-En DengDepartment of Urology, The Affiliated Tumor Hospital of Guangxi Medical University, Nanning 530021, Guangxi Zhuang Autonomous Region, People’s Republic of ChinaTel +86 18775391817Fax +86 771-5719573Email dengchengen66@126.comBi-Xun LiDepartment of Comprehensive Internal Medicine, The Affiliated Tumor Hospital of Guangxi Medical University, Nanning 530021, Guangxi Zhuang Autonomous Region, People’s Republic of ChinaTel +86 18977100069Fax +86 771-5719573Email libixun@126.comPurpose: Digestive carcinomas remain a major health burden worldwide and are closely related to type 2 diabetes. The aim of this study was to develop and validate a digestive carcinoma risk prediction model to identify high-risk individuals among those with type 2 diabetes.Patients and Methods: The prediction model was developed in a primary cohort that consisted of 655 patients with type 2 diabetes. Data were collected from November 2013 to December 2018. Clinical parameters and demographic characteristics were analyzed by logistic regression to develop a model to predict the risk of digestive carcinomas; then, a nomogram was constructed. The performance of the nomogram was assessed with respect to calibration, discrimination, and clinical usefulness. The results were internally validated by a bootstrapping procedure. The independent validation cohort consisted of 275 patients from January 2019 to December 2019.Results: Predictors in the prediction nomog

Continual lifelong learning is essential to many applications. In this paper, we propose a simple but effective approach to continual deep learning. Our approach leverages the principles of deep model compression, critical weights selection, and progressive networks expansion. By enforcing their integration in an iterative manner, we introduce an incremental learning method that is scalable to the number of sequential tasks in a continual learning process. Our approach is easy to implement and owns several favorable characteristics. First, it can avoid forgetting (i.e., learn new tasks while remembering all previous tasks). Second, it allows model expansion but can maintain the model compactness when handling sequential tasks. Besides, through our compaction and selection/expansion mechanism, we show that the knowledge accumulated through learning previous tasks is helpful to build a better model for the new tasks compared to training the models independently with tasks. Experimental results show that our approach can incrementally learn a deep model tackling multiple tasks without forgetting, while the model compactness is maintained with the performance more satisfiable than individual task training., Comment: To appear in NeurIPS 2019

It is hard to detect on-road objects under various lighting conditions. To improve the quality of the classifier, three techniques are used. We define subclasses to separate daytime and nighttime samples. Then we skip similar samples in the training set to prevent overfitting. With the help of the outside training samples, the detection accuracy is also improved. To detect objects in an edge device, Nvidia Jetson TX2 platform, we exert the lightweight model ResNet-18 FPN as the backbone feature extractor. The FPN (Feature Pyramid Network) generates good features for detecting objects over various scales. With Cascade R-CNN technique, the bounding boxes are iteratively refined for better results., Comment: Accepted at IEEE 21st International Workshop on Multimedia Signal Processing (MMSP 2019)

Nowadays, the path routing over road networks has become increasingly important, yet challenging, in many real-world applications such as location-based services (LBS), logistics and supply chain management, transportation systems, map utilities, and so on. While many prior works aimed to find a path between a source and a destination with the smallest traveling distance/time, they do not take into account the quality constraints (e.g., obstacles) of the returned paths, such as uneven roads, roads under construction, and weather conditions on roads. Inspired by this, in this paper, we consider two types of ad-hoc obstacles, keyword-based and weather-based obstacles, on road networks, which can be used for modeling roads that the returned paths should not pass through. In the presence of such ad-hoc obstacles on roads, we formulate a path routing query over road networks with ad-hoc obstacles (PRAO), which retrieves paths from source to destination on road networks that do not pass ad-hoc keyword and weather obstacles and have the smallest traveling time. In order to efficiently answer PRAO queries, we design effective pruning methods and indexing mechanism to facilitate efficient PRAO query answering. Extensive experiments have demonstrated the efficiency and effectiveness of our approaches over real/synthetic data sets., Comment: Accepted for Information Systems (IS) Journal, 2019

Hung-Hsien Chiang,1 Deng-Chyang Wu,2 Pin-I Hsu,3 Chao-Hung Kuo,2 Wei-Chen Tai,1 Shih-Cheng Yang,1 Keng-Liang Wu,1 Chih-Chien Yao,1 Cheng-En Tsai,1 Chih-Ming Liang,1 Yao-Kuang Wang,2 Jiunn-Wei Wang,2 Chih-Fang Huang,4 Seng-Kee Chuah,1  on behalf of the Taiwan Acid-Related Disease Study Group 1Division of Hepato-Gastroenterology, Department of Internal Medicine, Kaohsiung Chang Gung Memorial Hospital and Chang Gung University College of Medicine, Kaohsiung, Taiwan; 2Division of Gastroenterology, Department of Internal Medicine, Kaohsiung Medical University Hospital and Kaohsiung Medical University, Kaohsiung, Taiwan; 3Division of Gastroenterology, Department of Internal Medicine, Kaohsiung Veterans General Hospital, National Yang-Ming University, Kaohsiung, Taiwan; 4Division of Family Physicians, Kaohsiung Chang Gung Memorial Hospital and Chang Gung University College of Medicine, Kaohsiung, TaiwanPurpose: Research comparing the clinical efficacy of dexlansoprazole and esomeprazole has been limited. This study aims to compare the clinical efficacy of single doses of dexlansoprazole (modified-release 60 mg) and esomeprazole (40 mg) after 24-week follow-up in patients with mild erosive esophagitis.Methods: We enrolled 86 adult GERD subjects, randomized in a 1:1 ratio to two sequence groups defining the order in which they received single doses of dexlansoprazole (n=43) and esomeprazole (n=43) for 8 weeks as initial treatment. Patients displaying complete symptom resolution (CSR) by the end of initial treatment (8 weeks) were switched to on-demand therapy until the end of 24 weeks. Follow-up endoscopy was performed either at the end of 24 weeks or when severe reflux symptoms occurred. Five patients were lost to follow-up, leaving 81 patients (dexlansoprazole, n=41; esomeprazole, n=40) in the per-protocol analysis.Results: The GERDQ scores at 4-, 8-, 12-, 16-, 20-, and 24-week posttreatment were less than the

Simplified representations of processes influencing forest biomass in Earth system models (ESMs) contribute to large uncertainty in projections. We evaluate forest biomass from eight ESMs outputs archived in the Coupled Model Intercomparison Project Phase 5 (CMIP5) using the biomass data synthesized from radar remote sensing and ground-based observations across northern extratropical latitudes. ESMs exhibit large biases in the forest distribution, forest fraction, and mass of carbon pools that contribute to uncertainty in forest total biomass (biases range from -20 Pg C to 135 Pg C). Forest total biomass is primarily positively correlated with precipitation variations, with surface temperature becoming equally important at higher latitudes, in both simulations and observations. Relatively small differences in forest biomass between the pre-industrial period and the contemporary period indicate uncertainties in forest biomass were introduced in the pre-industrial model equilibration (spin-up), suggesting parametric or structural model differences are a larger source of uncertainty than differences in transient responses. Our findings emphasize the importance of improved (1) models of carbon allocation to biomass compartments, (2) distribution of vegetation types in models, and (3) reproduction of pre-industrial vegetation conditions, in order to reduce the uncertainty in forest biomass simulated by ESMs.

Following Marr's insight, we propose a generative image representation called primal sketch, which integrates two modeling components. The first component explains the structural part of an image, such as object boundaries, by a hidden layer of image primitives. The second component models the remaining textural part without distinguishable elements by Markov random fields that interpolate the structural part of the image. We adopt an artist's notion by calling the two components "sketchable" and "non-sketchable" parts respectively. A dictionary of image primitives are used for modeling structures in natural images, and each primitive is specified by variables for its photometric, geometric, and topological attributes. The primitives in the image representation are not independent but organized in an sketch graph. This sketch graph is modeled by a spatial Markov model that enforces Gestalt organizations. The inference of the sketch graph consists of two phases. Phase I sequentially adds the most prominent image primitives in a procedure similar to matching pursuit. Phase II edits the sketch graph by a number of graph operators to achieve good Gestalt organizations. Experiments show that the primal sketch model produces satisfactory results for a large number of generic images. The primal sketch model is not only a parsimonious image representation for lossy image coding, but also provides a meaningful mid-level generic representation for other vision tasks.

Following Marr's insight, we propose a generative image representation called primal sketch, which integrates two modeling components. The first component explains the structural part of an image, such as object boundaries, by a hidden layer of image primitives. The second component models the remaining textural part without distinguishable elements by Markov random fields that interpolate the structural part of the image. We adopt an artist's notion by calling the two components "sketchable" and "non-sketchable" parts respectively. A dictionary of image primitives are used for modeling structures in natural images, and each primitive is specified by variables for its photometric, geometric, and topological attributes. The primitives in the image representation are not independent but organized in an sketch graph. This sketch graph is modeled by a spatial Markov model that enforces Gestalt organizations. The inference of the sketch graph consists of two phases. Phase I sequentially adds the most prominent image primitives in a procedure similar to matching pursuit. Phase II edits the sketch graph by a number of graph operators to achieve good Gestalt organizations. Experiments show that the primal sketch model produces satisfactory results for a large number of generic images. The primal sketch model is not only a parsimonious image representation for lossy image coding, but also provides a meaningful mid-level generic representation for other vision tasks.

Following Marr’s insight, we propose a generative image representation called primal sketch, which integrates two modeling components. The first component explains the structural part of an image, such as ob ject boundaries, by a hidden layer of image primitives. The second component models the remaining textural part without distinguishable elements by Markov random fields that interpolate the structural part of the image. We adopt an artist’s notion by calling the two components “sketchable” and “non-sketchable” parts respectively. A dictionary of image primitives are used for modeling structures in natural images, and each primitive is specified by variables for its photometric, geometric, and topological attributes. The primitives in the image representation are not independent but organized in an sketch graph. This sketch graph is modeled by a spatial Markov model that enforces Gestalt organizations. The inference of the sketch graph consists of two phases. Phase I sequentially adds the most prominent image primitives in a procedure similar to matching pursuit. Phase II edits the sketch graph by a number of graph operators to achieve good Gestalt organizations. Experiments show that the primal sketch model produces satisfactory results for a large number of generic images. The primal sketch model is not only a parsimonious image representation for lossy image coding, but also provides a meaningful mid-level generic representation for other vision tasks.

Following Marr’s insight, we propose a generative image representation called primal sketch, which integrates two modeling components. The first component explains the structural part of an image, such as ob ject boundaries, by a hidden layer of image primitives. The second component models the remaining textural part without distinguishable elements by Markov random fields that interpolate the structural part of the image. We adopt an artist’s notion by calling the two components “sketchable” and “non-sketchable” parts respectively. A dictionary of image primitives are used for modeling structures in natural images, and each primitive is specified by variables for its photometric, geometric, and topological attributes. The primitives in the image representation are not independent but organized in an sketch graph. This sketch graph is modeled by a spatial Markov model that enforces Gestalt organizations. The inference of the sketch graph consists of two phases. Phase I sequentially adds the most prominent image primitives in a procedure similar to matching pursuit. Phase II edits the sketch graph by a number of graph operators to achieve good Gestalt organizations. Experiments show that the primal sketch model produces satisfactory results for a large number of generic images. The primal sketch model is not only a parsimonious image representation for lossy image coding, but also provides a meaningful mid-level generic representation for other vision tasks.

This paper presents a mathematical framework for visual learning that integrates two popular statistical learning paradigms in the literature: I) Descriptive learning, such as Markov random fields and minimax entropy learning, and II) Generative learning, such as PCA, ICA, TCA, and HMM. We apply the integrated learning framework to texture modeling, and we assume that an observed texture image is generated by multiple layers of hidden stochastic processes with various texture elements called "textons". Each texton is expressed as a window function like a mini-template or a wavelet, and each hidden stochastic process is a spatial pattern with a number of textons subject to affine transfomrations. The hidden layers are characterized by minimax entropy models, and they generate images by occlusion or linear addition. Thus given a raw input image, the learning framework achieves four goals: i) Computing the appearance of textons. ii) Inferring the hidden stochastic processes iii) Learning Gibbs models for each hidden stochastic process and iv) Verifying the learnt textons and models through random sampling. The integrated framework subsumes the minimax entropy learning paradigm and creates a richer class of probability models for visual patterns. Furthermore we show that the integration of descriptive and generative methods is a natural path of visual learning. We demonstrate the proposed framework and algorithms on many real images.

This paper presents a mathematical framework for visual learning that integrates two popular statistical learning paradigms in the literature: I) Descriptive learning, such as Markov random fields and minimax entropy learning, and II) Generative learning, such as PCA, ICA, TCA, and HMM. We apply the integrated learning framework to texture modeling, and we assume that an observed texture image is generated by multiple layers of hidden stochastic processes with various texture elements called "textons". Each texton is expressed as a window function like a mini-template or a wavelet, and each hidden stochastic process is a spatial pattern with a number of textons subject to affine transfomrations. The hidden layers are characterized by minimax entropy models, and they generate images by occlusion or linear addition. Thus given a raw input image, the learning framework achieves four goals: i) Computing the appearance of textons. ii) Inferring the hidden stochastic processes iii) Learning Gibbs models for each hidden stochastic process and iv) Verifying the learnt textons and models through random sampling. The integrated framework subsumes the minimax entropy learning paradigm and creates a richer class of probability models for visual patterns. Furthermore we show that the integration of descriptive and generative methods is a natural path of visual learning. We demonstrate the proposed framework and algorithms on many real images.

Computer vision can be considered a highly specialized data collection and data analysis problem. We need to understand the special properties of image data in order to construct statistical models for representing the wide variety of image patterns. One special property of vision that distinguishes itself from other sensory data such as speech data is that distance or scale plays a profound role in image data. More specifically, visual objects and patterns can appear at a wide range of distances or scales, and the same visual pattern appearing at different distances or scales produces different image data with different statistical properties, thus entails different regimes of statistical models. In particular, we show that the entropy rate of the image data changes over the viewing distance (as well as the camera resolution). Moreover, the inferential uncertainty changes with viewing distance too. We call these changes information scaling. From this perspective, we examine both empirically and theoretically two prominent and yet largely isolated research themes in image modeling literature, namely, wavelet sparse coding and Markov random fields. Our results indicate that the two models are appropriate on two different entropy regimes: sparse coding targets the low entropy regime, whereas the random fields are suitable for the high entropy regime. Because of information scaling, both models are necessary for representing and interpreting image intensity patterns in the whole entropy range, and information scaling triggers transitions between these two regimes of models. This motivates us to propose a full-zoom primal sketch model that integrates both sparse coding and Markov random fields.

Computer vision can be considered a highly specialized data collection and data analysis problem. We need to understand the special properties of image data in order to construct statistical models for representing the wide variety of image patterns. One special property of vision that distinguishes itself from other sensory data such as speech data is that distance or scale plays a profound role in image data. More specifically, visual objects and patterns can appear at a wide range of distances or scales, and the same visual pattern appearing at different distances or scales produces different image data with different statistical properties, thus entails different regimes of statistical models. In particular, we show that the entropy rate of the image data changes over the viewing distance (as well as the camera resolution). Moreover, the inferential uncertainty changes with viewing distance too. We call these changes information scaling. From this perspective, we examine both empirically and theoretically two prominent and yet largely isolated research themes in image modeling literature, namely, wavelet sparse coding and Markov random fields. Our results indicate that the two models are appropriate on two different entropy regimes: sparse coding targets the low entropy regime, whereas the random fields are suitable for the high entropy regime. Because of information scaling, both models are necessary for representing and interpreting image intensity patterns in the whole entropy range, and information scaling triggers transitions between these two regimes of models. This motivates us to propose a full-zoom primal sketch model that integrates both sparse coding and Markov random fields.

Vision can be posed as a statistical learning and inference problem. As an over-simplified account, let W be a description of the outside scene in terms of ìwhat is where,î let I be the retina image, and let p(W, I) be the joint distribution of W and I. 1 Then visual learning is to learn p(W, I) from training data, and visual perception is to infer W from I based on p(W|I).There are two major schools on visual learning and perception. One school is operation-oriented and learns the inferential process defined by p(W|I) directly, often in the form of an explicit transformation W ? F(I). This scheme is mostly used in supervised learning, where W is object category, and is given in training data. The other school is representation-oriented and learns the generative process p(W) and p(I|W) explicitly, then perception is to invert the generative process by maximizing or sampling p(W|I) ? p(W)p(I|W). In this scheme, p(W) may also be accounted for by a regularization term such as smoothness or sparsity. This scheme is often used in unsupervised learning where W is not available in training data.

Computer vision can be considered a highly specialized data collection and data analysis problem. We need to understand the special properties of image data in order to construct statistical models for representing the wide variety of image patterns. One special property of vision that distinguishes itself from other sensory data such as speech data is that distance or scale plays a profound role in image data. More specifically, visual objects and patterns can appear at a wide range of distances or scales, and the same visual pattern appearing at different distances or scales produces different image data with different statistical properties, thus entails different regimes of statistical models. In particular, we show that the entropy rate of the image data changes over the viewing distance (as well as the camera resolution). Moreover, the inferential uncertainty changes with viewing distance too. We call these changes information scaling. From this perspective, we examine both empirically and theoretically two prominent and yet largely isolated research themes in image modeling literature, namely, wavelet sparse coding and Markov random fields. Our results indicate that the two models are appropriate on two different entropy regimes: sparse coding targets the low entropy regime, whereas the random fields are suitable for the high entropy regime. Because of information scaling, both models are necessary for representing and interpreting image intensity patterns in the whole entropy range, and information scaling triggers transitions between these two regimes of models. This motivates us to propose a full-zoom primal sketch model that integrates both sparse coding and Markov random fields. In this model, local image intensity patterns are classified into “sketchable regime” and “non-sketchable regime” by a sketchability criterion. In the sketchable regime, the image data are represented deterministically by highly parametrized sketch primitives. In the non-sketchab

Vision can be posed as a statistical learning and inference problem. As an over-simplified account, let W be a description of the outside scene in terms of ìwhat is where,î let I be the retina image, and let p(W, I) be the joint distribution of W and I. 1 Then visual learning is to learn p(W, I) from training data, and visual perception is to infer W from I based on p(W|I).There are two major schools on visual learning and perception. One school is operation-oriented and learns the inferential process defined by p(W|I) directly, often in the form of an explicit transformation W ? F(I). This scheme is mostly used in supervised learning, where W is object category, and is given in training data. The other school is representation-oriented and learns the generative process p(W) and p(I|W) explicitly, then perception is to invert the generative process by maximizing or sampling p(W|I) ? p(W)p(I|W). In this scheme, p(W) may also be accounted for by a regularization term such as smoothness or sparsity. This scheme is often used in unsupervised learning where W is not available in training data.

Computer vision can be considered a highly specialized data collection and data analysis problem. We need to understand the special properties of image data in order to construct statistical models for representing the wide variety of image patterns. One special property of vision that distinguishes itself from other sensory data such as speech data is that distance or scale plays a profound role in image data. More specifically, visual objects and patterns can appear at a wide range of distances or scales, and the same visual pattern appearing at different distances or scales produces different image data with different statistical properties, thus entails different regimes of statistical models. In particular, we show that the entropy rate of the image data changes over the viewing distance (as well as the camera resolution). Moreover, the inferential uncertainty changes with viewing distance too. We call these changes information scaling. From this perspective, we examine both empirically and theoretically two prominent and yet largely isolated research themes in image modeling literature, namely, wavelet sparse coding and Markov random fields. Our results indicate that the two models are appropriate on two different entropy regimes: sparse coding targets the low entropy regime, whereas the random fields are suitable for the high entropy regime. Because of information scaling, both models are necessary for representing and interpreting image intensity patterns in the whole entropy range, and information scaling triggers transitions between these two regimes of models. This motivates us to propose a full-zoom primal sketch model that integrates both sparse coding and Markov random fields. In this model, local image intensity patterns are classified into “sketchable regime” and “non-sketchable regime” by a sketchability criterion. In the sketchable regime, the image data are represented deterministically by highly parametrized sketch primitives. In the non-sketchab

Textons refer to fundamental micro-structures in generic natural images and thus constitute the basic elements in early (pre-attentive) visual perception. However, the word "textong" remains a vague concept in the literature of computer vision and visual perception, and a precist mathematical definition has yet to be found. In this article, we argue that the definition of texton should be governed by a sound mathematical model of images, and the set of textons must be learned from, or best tuned to, an image ensemble. We adopt a generative image model that an image is a superposition of bases from an over-complete dictionary, then a texton is defined as a mini-template that consists of a varying number of image bases with some geometric and photometric configurations. By analogy to physics, if image bases are like protons, neutrons and electrons, then textons are like atoms. Then a small number of textons can be learned from training images as repeating micro-structures. We report four experiments for comparison. The first experiment computes clusters in feature space of filter responses. The second transformed component analysis in both feature space and image patches. The third adopts a two-layer generative model where an image is generated by image bases and image bases are generated by textons. The fourth experiment shows textons from motion image sequences, which we call movetons.

Recent results on sparse coding and independent component analysis suggest that human vision first represents a visual image by a linear superposition of a relatively small number of localized elongate, oriented image bases. With this representation, the sketch of an image consists of the locations, orientations, and elongations of the images bases, and the sketch can be visually illustrated by depicting each image base by a linelet of the same length and orientation. Built on the insight of sparse and independent component analysis, we propose a two-level generative model for textures. At the bottom-level, the texture image is represented by a linear superposition of image bases. At the top-level, a Markov model is assumed for the placement of the image bases or the sketch, and the model is characterized by a set of simple geometrical feature statistics.

Textons refer to fundamental micro-structures in generic natural images and thus constitute the basic elements in early (pre-attentive) visual perception. However, the word "textong" remains a vague concept in the literature of computer vision and visual perception, and a precist mathematical definition has yet to be found. In this article, we argue that the definition of texton should be governed by a sound mathematical model of images, and the set of textons must be learned from, or best tuned to, an image ensemble. We adopt a generative image model that an image is a superposition of bases from an over-complete dictionary, then a texton is defined as a mini-template that consists of a varying number of image bases with some geometric and photometric configurations. By analogy to physics, if image bases are like protons, neutrons and electrons, then textons are like atoms. Then a small number of textons can be learned from training images as repeating micro-structures. We report four experiments for comparison. The first experiment computes clusters in feature space of filter responses. The second transformed component analysis in both feature space and image patches. The third adopts a two-layer generative model where an image is generated by image bases and image bases are generated by textons. The fourth experiment shows textons from motion image sequences, which we call movetons.

Recent results on sparse coding and independent component analysis suggest that human vision first represents a visual image by a linear superposition of a relatively small number of localized elongate, oriented image bases. With this representation, the sketch of an image consists of the locations, orientations, and elongations of the images bases, and the sketch can be visually illustrated by depicting each image base by a linelet of the same length and orientation. Built on the insight of sparse and independent component analysis, we propose a two-level generative model for textures. At the bottom-level, the texture image is represented by a linear superposition of image bases. At the top-level, a Markov model is assumed for the placement of the image bases or the sketch, and the model is characterized by a set of simple geometrical feature statistics.

In questa seconda parte del viaggio, episodi dal 24 al 48, assisteremo agli insegnamenti di Tripitaka e ai fantastici combattimenti di Scimmiotto e dei suoi compagni di viaggio contro demoni, draghi ed esseri maligni.

La quarta e ultima parte delle fantastiche vicende di Scimmiotto d'Oro e del suo maestro, durante il lungo viaggio che dalla Cina li conduce alla casa del Budda, in Occidente, per il recupero dei Sutra, le tavole della legge buddista. In questi ultimi 26 episodi assisteremo ai momenti più illuminanti del viaggio.Voci di Silvia Cecchini e di Serafino BalduzziMusiche originali di Ivan Genesio

Le fantastiche vicende di Scimmiotto d'Oro e del suo maestro, durante il lungo viaggio che dalla Cina li conduce alla casa del Budda, in Occidente, per il recupero dei Sutra, le tavole della legge buddista. Il protagonista, dal quale i manga giapponesi hanno tratto il loro indiscusso eroe del momento, è davvero eccezionale: dotato di capacità che lo rendono superiore a qualsiasi divinità nella quale si sia imbattuto, con l'eccezione del Budda e della Pus'a Kuan Yin, egli muta la sua tempra aggressiva in quella di un fervente sostenitore della regola celeste, sempre a modo suo, ovviamente. Umoristico, commovente, dissacratore, si può definire un romanzo sacro, un romanzo d'avventura, un romanzo storico, insomma, un vero e proprio viaggio per chi lo ascolta, così come lo è stato per chi l'ha tradotto e per chi l'ha letto; infine un vero romanzo di formazione, nella spiritualità orientale. Il mito di Scimmiotto è fra i più noti della Cina.La lettura è a due voci: una dà lettura al testo (Silvia Cecchini), l'altra alle note che lo accompagnano (Serafino Balduzzi, che ne è l’autore e il traduttore). Inoltre i tratti che sono in poesia nell'originale sono sottolineati dalla musica: l'ascolto diventa vivace, sorprendente, ricco di spunti .Musiche originali di Ivan Genesio.L’intera opera è divisa in quattro parti: in questa prima parte ascolteremo l’introduzione di Serafino Balduzzi, letta da lui, e i primi 23 episodi del viaggio.

In questa terza parte del viaggio, episodi dal 49 al 73, le lezioni da apprendere saranno sempre più dure, ma, anche Scimmiotto comincerà a nutrire, nel suo cuore, dei sentimenti sempre più profondi verso Tripitaka e i suoi compagni, e il suo percorso spirituale inizierà a delinearsi.Voci di Silvia Cecchini e di Serafino Balduzzi, musiche originali di Ivan Genesio.

Le fantastiche vicende di Scimmiotto d'Oro e del suo maestro, durante il lungo viaggio che dalla Cina li conduce alla casa del Budda, in Occidente, per il recupero dei Sutra, le tavole della legge buddista. Il protagonista, dal quale i manga giapponesi hanno tratto il loro indiscusso eroe del momento, è davvero eccezionale: dotato di capacità che lo rendono superiore a qualsiasi divinità nella quale si sia imbattuto, con l'eccezione del Budda e della Pus'a Kuan Yin, egli muta la sua tempra aggressiva in quella di un fervente sostenitore della regola celeste, sempre a modo suo, ovviamente. Umoristico, commovente, dissacratore, si può definire un romanzo sacro, un romanzo d'avventura, un romanzo storico, insomma, un vero e proprio viaggio per chi lo ascolta, così come lo è stato per chi l'ha tradotto e per chi l'ha letto; infine un vero romanzo di formazione, nella spiritualità orientale. Il mito di Scimmiotto è fra i più noti della Cina.La lettura è a due voci: una dà lettura al testo (Silvia Cecchini), l'altra alle note che lo accompagnano (Serafino Balduzzi, che ne è l’autore e il traduttore). Inoltre i tratti che sono in poesia nell'originale sono sottolineati dalla musica: l'ascolto diventa vivace, sorprendente, ricco di spunti .Musiche originali di Ivan Genesio.L’intera opera è divisa in quattro parti: in questa prima parte ascolteremo l’introduzione di Serafino Balduzzi, letta da lui, e i primi 23 episodi del viaggio.

La quarta e ultima parte delle fantastiche vicende di Scimmiotto d'Oro e del suo maestro, durante il lungo viaggio che dalla Cina li conduce alla casa del Budda, in Occidente, per il recupero dei Sutra, le tavole della legge buddista. In questi ultimi 26 episodi assisteremo ai momenti più illuminanti del viaggio.Voci di Silvia Cecchini e di Serafino BalduzziMusiche originali di Ivan Genesio

In questa seconda parte del viaggio, episodi dal 24 al 48, assisteremo agli insegnamenti di Tripitaka e ai fantastici combattimenti di Scimmiotto e dei suoi compagni di viaggio contro demoni, draghi ed esseri maligni.

In questa terza parte del viaggio, episodi dal 49 al 73, le lezioni da apprendere saranno sempre più dure, ma, anche Scimmiotto comincerà a nutrire, nel suo cuore, dei sentimenti sempre più profondi verso Tripitaka e i suoi compagni, e il suo percorso spirituale inizierà a delinearsi.Voci di Silvia Cecchini e di Serafino Balduzzi, musiche originali di Ivan Genesio.