Search

The recent availability of large-scale data combining multiple data modalities has opened various research and commercial opportunities in Artificial Intelligence (AI). Machine Learning (ML) has achieved important results in this area mostly by adopting a sub-symbolic distributed representation. It is generally accepted now that such purely sub-symbolic approaches can be data inefficient and struggle at extrapolation and reasoning. By contrast, symbolic AI is based on rich, high-level representations ideally based on human-readable symbols. Despite being more explainable and having success at reasoning, symbolic AI usually struggles when faced with incomplete knowledge or inaccurate, large data sets and combinatorial knowledge. Neurosymbolic AI attempts to benefit from the strengths of both approaches combining reasoning with complex representation of knowledge and efficient learning from multiple data modalities. Hence, neurosymbolic AI seeks to ground rich knowledge into efficient sub-symbolic representations and to explain sub-symbolic representations and deep learning by offering high-level symbolic descriptions for such learning systems. Logic Tensor Networks (LTN) are a neurosymbolic AI system for querying, learning and reasoning with rich data and abstract knowledge. LTN introduces Real Logic, a fully differentiable first-order language with concrete semantics such that every symbolic expression has an interpretation that is grounded onto real numbers in the domain. In particular, LTN converts Real Logic formulas into computational graphs that enable gradient-based optimization. This chapter presents the LTN framework and illustrates its use on knowledge completion tasks to ground the relational predicates (symbols) into a concrete interpretation (vectors and tensors). It then investigates the use of LTN on semi-supervised learning, learning of embeddings and reasoning. LTN has been applied recently to many important AI tasks, including semantic image interpretation, ontology learning and reasoning, and reinforcement learning, which use LTN for supervised classification, data clustering, semi-supervised learning, embedding learning, reasoning and query answering. The chapter presents some of the main recent applications of LTN before analyzing results in the context of related work and discussing the next steps for neurosymbolic AI and LTN-based AI models.

Observations of aerosol and trace gases in the remote troposphere are vital to quantify background concentrations and identify long-term trends in atmospheric composition on large spatial scales. Measurements made at high altitude are often used to study free-tropospheric air; however such high-altitude sites can be influenced by boundary layer air masses. Thus, accurate information on air mass origin and transport pathways to high-altitude sites is required. Here we present a new method, based on the source-receptor relationship (SRR) obtained from backwards WRF-FLEXPART simulations and a k-means clustering approach, to identify source regions of air masses arriving at measurement sites. Our method is tailored to areas of complex terrain and to stations influenced by both local and long-range sources. We have applied this method to the Chacaltaya (CHC) GAW station (5240 m a.s.l.; 16.35 degrees S, 68.13 degrees W) for the 6-month duration of the "Southern Hemisphere high-altitude experiment on particle nucleation and growth" (SALILNA) to identify where sampled air masses originate and to quantify the influence of the surface and the free troposphere. A key aspect of our method is that it is probabilistic, and for each observation time, more than one air mass (cluster) can influence the station, and the percentage influence of each air mass can be quantified. This is in contrast to binary methods, which label each observation time as influenced by either boundary layer or free-troposphere air masses. Air sampled at CHC is a mix of different provenance. We find that on average 9 % of the air, at any given observation time, has been in contact with the surface within 4 d prior to arriving at CHC. Furthermore, 24 % of the air has been located within the first 1.5 km above ground level (surface included). Consequently, 76 % of the air sampled at CHC originates from the free troposphere. However, pure free-tropospheric influences are rare, and often samples are concurrently influenced by both boundary layer and free-tropospheric air masses. A clear diurnal cycle is present, with very few air masses that have been in contact with the surface being detected at night. The 6-month analysis also shows that the most dominant air mass (cluster) originates in the Amazon and is responsible for 29 % of the sampled air. Furthermore, short-range clusters (origins within 100 km of CHC) have high temporal frequency modulated by local meteorology driven by the diurnal cycle, whereas the mid- and long-range clusters' (> 200 km) variability occurs on timescales governed by synoptic-scale dynamics. To verify the reliability of our method, in situ sulfate observations from CHC are combined with the SRR clusters to correctly identify the (pre-known) source of the sulfate: the Sabancaya volcano located 400 km north-west from the station.

Grape pomace (GP), is the main winemaking by-product and could represent a valuable functional food ingredient being a source of bioactive compounds, like polyphenols. Polyphenols prevent many non-communicable diseases and could contrast the oxidation reaction in foods. However, the high content in polyunsaturated fatty acid, the described pro-oxidant potential action of some polyphenols and the complex interactions with other components of matrices during food processing must be considered. Indeed, all these factors could promote oxidative reactions and require focused and specific assay. The aims of this study were to evaluate the effects of GP powder (GPP) addition (at 0%, 5% and 10% concentrations) in breadsticks formulations both on the antioxidant activity at room temperature during storage and on the shelf-life by the OXITEST predictive approach. GPP fortification increased the total polyphenols content and the antioxidant activities of breadsticks. FRAP reduced during the first two days of storage at room temperature, TPC increased during the 75 days, while ABTS showed a slight progressive decrease. However, GP negatively influenced OXITEST estimated shelf-life of breadsticks, incrementing the oxidation rate. In conclusion, even if GP fortification of breadsticks could improve the nutritional value of the products, the increased commercial perishability represents a drawback that must be considered.

Carbonaceous aerosol is one of the main components of atmospheric particulate matter, which is of great significance due to its role in climate change, earth's radiation balance, visibility, and human health. In this work, carbonaceous aerosols were measured in Shijiazhuang and Beijing using the OC/EC analyzer from December 1, 2019 to March 15, 2020, which covered the Coronavirus Disease 2019 (COVID-19) pandemic. The observed results show that the gas-phase pollutants, such as NO, NO2, and aerosol-phase pollutants (Primary Organic Compounds, POC) from anthropogenic emissions, were significantly reduced during the lockdown period due to limited human activities in North China Plain (NCP). However, the atmospheric oxidation capacity (Ox/CO) shows a significantly increase during the lockdown period. Meanwhile, additional sources of nighttime Secondary Organic Carbon (SOC), Secondary Organic Aerosol (SOA), and babs, BrC(370 nm) are observed and ascribed to the nocturnal chemistry related to NO3 radical. The Potential Source Contribution Function (PSCF) analysis indicates that the southeast areas of the NCP region contributed more to the SOC during the lockdown period than the normal period. Our results highlight the importance of regional nocturnal chemistry in SOA formation., Graphical abstract Potential source of SOC during COVID-19 pandemic.Unlabelled Image

Aim To describe the epidemiological aspects of nocturnal enuresis (NE). In this study we identify the prevalence and the familial conditions in a large, representative sample of children with monosymptomatic NE (MNE) and nonmonosyptomatic NE (NMNE). Material and methods In this descriptive-analytic study the Italian Society of Pediatrics (SIP) promoted a prevalence study of NE using a questionnaire involved 320 primary care Pediatricians from Northern, Central and Southern Italy, from January 2019 to July 2019, with a total of 130,000 children analyzed by questionnaire related to epidemiology and type of NE, familiarity, quality of sleep, eating and drinking habits, pharmacological and psychological/behavioural interventions and family involvement. Results 270/320 (84.4%) Paediatrician replied to our questionnaire. We enrolled a total of 9307/130,000 (7.2%) children with NE, aged between 5 and 14 years: 2141 diagnosed with MNE and 7176 qualified as NMNE. Poor quality of sleep were reported in 7064 patients; 90% of children did not consider a dietary and drinking recommendation. Pediatrician reported a total of 54.1% of parents who declared to have a negative reaction to their children because of the bedwetting. A percentage of 71.4% of parents declared to use or to have used alternative therapies and not to prefer, at first, a pharmacological intervention. Conclusion The choice of treatment should include psychological/behavioural interventions in all cases to improve the therapeutic outcome. All primary care Pediatricians should be aware of the all aspects of NE to choose the best way to treat every patient.

Objective: The objective of the present study was to selectively evaluate the long-term impact of posterior fossa pilocytic astrocytomas, which are known to be among the most benign forms of paediatric brain tumours on neurocognitive and behavioural functions. Methods: Children that were operated on for a posterior fossa pilocytic astrocytoma in the Pediatric Neurosurgery Department of the Catholic University Medical School were selected according to the following criteria: (a) age &gt, 5 years (in order to have a complete set of neurocognitive evaluations data), (b) ability to perform a complete set of tests before and after surgery, and (c) children that had a regular follow-up up to 10 years from the surgical treatment. Results: Forty-three percent of the children selected for the present study showed a borderline IQ before surgery, which is a result corresponding to those previously reported in the literature for children affected by posterior fossa pilocytic astrocytomas, praxis and visual perception were the selective functions that were more frequently affected. Language performance tests scores were below average in 40% of the cases but tended to improve in terms of expressive and receptive skills even at the 1-year follow-up, the improvements became significant at the 5-year and 10-year follow-ups. Conclusions: Recognising and measuring the short- and long-term effects of cerebellar tumours in children and their treatment are the first step towards improving their clinical course and quality of life. Early interventions should be offered to all of them, with specific attention bestowed on visual-spatial stimulation, speech and occupational therapies in order to act on praxic and visuo-perceptive skills, as well as on emotion and behaviour tracts of the neurocognitive profile, which more commonly tend to persist in the long term.

Gaseous hydrochloric (HCl) and hydrobromic acid (HBr) are vital halogen species that play essential roles in tropospheric physicochemical processes. Yet, the majority of the current studies on these halogen species were conducted in marine or coastal areas. Detection and source identification of HCl and HBr in inland urban areas remain scarce, thus limiting the full understanding of halogen chemistry and potential atmospheric impacts in the environments with limited influence from the marine sources. Here, both gaseous HCl and HBr were concurrently measured in urban Beijing, China, during winter and early spring of 2019. We observed significant HCl and HBr concentrations ranging from a minimum value at 1 × 108 molecules cm−3 (4 ppt) and 4 × 107 molecules cm−3 (1 ppt) up to 6 × 109 molecules cm−3 (222 ppt) and 1 × 109 molecules cm−3 (37 ppt), respectively. The HCl and HBr concentrations are enhanced along with the increase of atmospheric temperature, UVB and levels of gaseous HNO3. Based on the air mass analysis and high correlations of HCl and HBr with the burning indicators (HCN and HCNO), gaseous HCl and HBr are found to be related to anthropogenic burning aerosols. The gas–particle partitioning may also play a dominant role in the elevated daytime HCl and HBr. During the daytime, the reactions of HCl and HBr with OH radicals lead to significant production of atomic Cl and Br, up to 2 × 104 molecules cm−3 s−1 and 8 × 104 molecules cm−3 s−1, respectively. The production rate of atomic Br (via HBr + OH) is 2–3 times higher than that of atomic Cl (via HCl + OH), highlighting the potential importance of bromine chemistry in the urban area. On polluted days, the production rates of atomic Cl and Br are faster than those on clean days. Furthermore, our observations of elevated HCl and HBr may suggest an important recycling pathway of halogen species in inland megacities and may provide a plausible explanation for the widespread halogen chemistry, which could affect the atmospheric oxidation in China.

Thermal desorption coupled with different detectors is an important analysis method for ambient carbonaceous aerosols. However, it is unclear how the compounds coexisting in both the gas and particle phases affect carbonaceous aerosol concentrations and measurements during thermal desorption. We observed matrix effects leading to a redistribution of different OC fractions (OC1 to OC4) during the thermal desorption process. These factors led to the formation of OC with low volatility (OC4), mainly from high-volatility OC (OC1 and OC2). Laboratory studies further indicated that ammonium promotes such matrix effects by transforming OC in the particle phase. Therefore, in addition to providing insights into the chemical evolution of OC during haze events, we argue that thermal-desorption-based OC measurements should be used with caution, which is an important step towards a more accurate measurement of OC in the ambient atmosphere.

The molecular composition and volatility of gaseous organic compounds were investigated during April–July 2019 at the Station for Measuring Ecosystem – Atmosphere Relations (SMEAR) II situated in a boreal forest in Hyytiälä, southern Finland. In order to obtain a more complete picture and full understanding of the molecular composition and volatility of ambient gaseous organic compounds (from volatile organic compounds, VOCs, to highly oxygenated organic molecules, HOMs), two different instruments were used. A Vocus proton-transfer-reaction time-of-flight mass spectrometer (Vocus PTR-ToF; hereafter Vocus) was deployed to measure VOCs and less oxygenated VOCs (i.e., OVOCs). In addition, a multi-scheme chemical ionization inlet coupled to an atmospheric pressure interface time-of-flight mass spectrometer (MION API-ToF) was used to detect less oxygenated VOCs (using Br− as the reagent ion; hereafter MION-Br) and more oxygenated VOCs (including HOMs; using NO3- as the reagent ion; hereafter MION-NO3). The comparison among different measurement techniques revealed that the highest elemental oxygen-to-carbon ratios (O : C) of organic compounds were observed by the MION-NO3 (0.9 ± 0.1, average ± 1 standard deviation), followed by the MION-Br (0.8 ± 0.1); lowest O : C ratios were observed by Vocus (0.2 ± 0.1). Diurnal patterns of the measured organic compounds were found to vary among different measurement techniques, even for compounds with the same molecular formula, suggesting contributions of different isomers detected by the different techniques and/or fragmentation from different parent compounds inside the instruments. Based on the complementary molecular information obtained from Vocus, MION-Br, and MION-NO3, a more complete picture of the bulk volatility of all measured organic compounds in this boreal forest was obtained. As expected, the VOC class was the most abundant (about 53.2 %), followed by intermediate-volatility organic compounds (IVOCs, about 45.9 %). Although condensable organic compounds (low-volatility organic compounds, LVOCs; extremely low volatility organic compounds, ELVOCs; and ultralow-volatility organic compounds, ULVOCs) only comprised about 0.2 % of the total gaseous organic compounds, they play an important role in new particle formation as shown in previous studies in this boreal forest. Our study shows the full characterization of the gaseous organic compounds in the boreal forest and the advantages of combining Vocus and MION API-ToF for measuring ambient organic compounds with different oxidation extents (from VOCs to HOMs). The results therefore provide a more comprehensive understanding of the molecular composition and volatility of atmospheric organic compounds as well as new insights into interpreting ambient measurements or testing/improving parameterizations in transport and climate models.

Purpose The assessment of Autism Spectrum Disorders (ASD) in childhood has two essential aspects: the identification of the risk (under 30 months of age) and the definition of a diagnosis that takes into account its core areas as well as further non-specific aspects. The purpose of this paper is to present an approach that considers the combination of clinical evaluation with the use of tools that analyse the various levels of the child’s functioning as fundamental. Design/methodology/approach The comprehensive assessment at the Institute of Ortofonologia in Rome provides the ADOS-2 and the Leiter-R for the evaluation of the symptomatology, the severity level, the non-verbal cognitive functioning and the fluid reasoning; the TCE and the UOI are used to identify, respectively, the child’s emotional skills and the ability to understand the intentions of others, as precursors of the theory of mind. Within this assessment, the Brief-P, the Short Sensory Profile and the RBS are also included for the evaluation of executive functions, sensory pattern and of restricted and repetitive behaviours, as observed by parents. Findings How to define a reliable development profile, which allows to plan a specific intervention calibrated on the potential of the child and on his development trajectory, is described. Two clinical cases are also presented. Originality/value The entire process is aimed both at a detailed assessment of the child’s functioning and at identifying a specific therapeutic project and predictive factors for achieving an optimal outcome.

Infection may complicate the outcome of cranial repair with significant additional morbidity, related to hospitalization, surgery and long antibiotic therapy, that may become even dramatic in case of multi-resistant germs and in particular in the paediatric population. Additionally, the economic costs for the health system are obvious. Moreover, surgical decisions concerning the timing of cranioplasty and choice of the material may be strongly affected by the risk of infection. Despite, management and prevention of cranioplasty infections are not systematically treated through the literature so far. We reviewed pertinent literature dealing with cranioplasty infection starting from the diagnosis to treatment options, namely conservative versus surgical ones. Our institutional bundle, specific to the paediatric population, is also presented. This approach aims to significantly reduce the risk of infection in first-line cranioplasty and redo cranioplasty after previous infection. A thorough knowledge and understanding of risk factors may lead to surgical strategies and bundles, aiming to reduce infectious complications of cranioplasty. Finally, innovation in materials used for cranial repair should also aim to enhance the antimicrobial properties of these inert materials.

Background Piezosurgery (PS) has gained increasing dispersion in neurosurgery. In pediatric neurosurgery, the experience is limited to craniosynostosis surgery. The present study assesses PS in the pediatric population, also considering outcomes and complications in cranial and spinal procedures. Methods All consecutive craniotomies and laminotomies, performed with PS (group A) or conventional osteotomes (group B) in the 2014–2017 period were reviewed. The following variables were analyzed: dural tears, estimated blood loss and need of transfusion, cosmetic outcome (Sloan score), and operative times. A review of the pertinent literature is included. Results 172 children were enrolled, 90 in group A and 82 in group B. The mean follow-up time was 2.1 years. A statistically significant difference in favor of group A was found about EBL (105 vs. 113 ml) and late outcome (Sloan class A 98.5% vs. 91.5%). PS also reduced the risk of dural tears (1 vs. 7 cases in groups A and B, respectively) and blood transfusion (52% vs. 55.5%) but without statistical significance. The operative times were significantly shorter in group B (13 vs. 23 minutes), although the newer PS plus (PSP) was demonstrated to significantly shorten these times compared with the traditional PS (3.5 vs. 6.5 minutes for orbitotomy and 7.5 vs. 9.5 minutes for hemicraniotomy). Conclusions PS is a safe and effective tool that can be specifically recommended for bone splitting and graft, laminotomy, and craniotomy in cosmetically eloquent areas. The limit of operation times can be overcome by a learning curve in neurosurgery and PSP.

Trackers have become popular devices these days. They are extensively used to record sports activities (e.g., hiking, skiing), mainly in terms of GPS trajectories, which can be shared on social networking platforms with other users looking for leisure tips. Notably, as the number of available trajectories drastically increased over time, in many cases, it has become challenging, if not impossible, the extensive evaluation of all possible alternatives and the manual selection of the most suitable one. Paths are characterized by multiple features (e.g., dirt, asphalt), and a good representation is needed to satisfy user needs. Moreover, paths can be composed to generate new routes. This calls for a recommender system capable to handle both the multi-feature path representation and the implicit definition of alternatives by composition. This paper suggests a novel approach that features a richer trajectory representation based on a semantic annotation to describe significant path features. Annotations are then used for automatic recommendation of new paths that maximize the presence of characteristics matching the user preferences. Finally, a class of algorithm variants is evaluated using an off-line validation process and compared with a baseline solution to test the underlying assumptions.

We investigate grounded language learning through real-world data, by modelling a teacher-learner dynamics through the natural interactions occurring between users and search engines; in particular, we explore the emergence of semantic generalization from unsupervised dense representations outside of synthetic environments. A grounding domain, a denotation function and a composition function are learned from user data only. We show how the resulting semantics for noun phrases exhibits compositional properties while being fully learnable without any explicit labelling. We benchmark our grounded semantics on compositionality and zero-shot inference tasks, and we show that it provides better results and better generalizations than SOTA non-grounded models, such as word2vec and BERT., Published as a conference paper at NAACL2021

We present Query2Prod2Vec, a model that grounds lexical representations for product search in product embeddings: in our model, meaning is a mapping between words and a latent space of products in a digital shop. We leverage shopping sessions to learn the underlying space and use merchandising annotations to build lexical analogies for evaluation: our experiments show that our model is more accurate than known techniques from the NLP and IR literature. Finally, we stress the importance of data efficiency for product search outside of retail giants, and highlight how Query2Prod2Vec fits with practical constraints faced by most practitioners., Published as a conference paper at NAACL2021 - Industry Track

In this study, we use temporally aligned word embeddings and a large diachronic corpus of English to quantify language change in a data‐driven, scalable way, which is grounded in language use. We show a unique and reliable relation between measures of language change and age of acquisition (AoA) while controlling for frequency, contextual diversity, concreteness, length, dominant part of speech, orthographic neighborhood density, and diachronic frequency variation. We analyze measures of language change tackling both the change in lexical representations and the change in the relation between lexical representations and the words with the most similar usage patterns, showing that they capture different aspects of language change. Our results show a unique relation between language change and AoA, which is stronger when considering neighborhood‐level measures of language change: Words with more coherent diachronic usage patterns tend to be acquired earlier. The results support theories positing a link between ontogenetic and ethnogenetic processes in language.

Background The optimal management of neonatal post-hemorrhagic hydrocephalus (PHH) is still debated, though several treatment options have been proposed. In the last years, ventriculosubgaleal shunt (VSgS) and neuroendosdcopic lavage (NEL) have been proposed to overcome the drawbacks of more traditional options, such as external ventricular drainage and ventricular access device. Methods We retrospectively reviewed neonates affected by PHH treated at our institution since September 2012 to September 2020. Until 2017 patients received VSgS as initial treatment. After the introduction of NEL, this treatment option was offered to patients with large intraventricular clots. After NEL, VSgS was always placed. Primary VSgS was reserved to patients without significant intraventricular clots and critically ill patients that could not be transferred to the operating room and undergo a longer surgery. Results We collected 63 babies (38 males and 25 females) with mean gestational age of 27.8 ± 3.8SD weeks (range 23–38.5 weeks) and mean birthweight of 1199.7 ± 690.6 SD grams (range 500–3320 g). In 6 patients, hemorrhage occurred in the third trimester of gestation, while in the remaining cases hemorrhage complicated prematurity. This group included 37 inborn and 26 outborn babies. Intraventricular hemorrhage was classified as low grade (I–II according to modified Papile grading scale) in 7 cases, while in the remaining cases the grade of hemorrhage was III to IV. Mean age at first neurosurgical procedure was 32.2 ± 3.6SD weeks (range 25.4–40 weeks). Death due to prematurity occurred in 5 patients. First-line treatment was VSgS in 49 patients and NEL in the remaining 14 cases. Mean longevity of VSgS was 30.3 days (range 10–97 days) in patients finally requiring an additional treatment of hydrocephalus. Thirty-two patients required one to three redo VSgS. Interval from initial treatment to permanent shunt ranged from 14 to 312 days (mean 70.9 days). CSF infection was observed in 5 patients (7.9%). Shunt dependency was observed in 51 out of 58 surviving patients, while 7 cases remained shunt-free at the last follow-up. Multiloculated hydrocephalus was observed in 14 cases. Among these, only one patient initially received NEL and was complicated by isolated trapped temporal horn. Conclusions VSgS and NEL are two effective treatment options in the management of PHH. Both procedures should be part of the neurosurgical armamentarium to deal with PHH, since they offer specific advantages in selected patients. A treatment algorithm combining these two options may reduce the infectious risk and the risk of multiloculated hydrocephalus.

Parents of children with autism spectrum disorder (ASD) have to address various challenges mainly due to their children’s atypia related to communication, emotion regulation and behaviors, arising also within the relationship with their caregivers. Several studies have pointed out that children with ASD can exhibit many difficulties regarding initiating and maintaining meaningful relationships with others. To date, little research has explored the interactions between children with ASD and their fathers, focusing more on mothers. In the context of parent–child interactions in the presence of autism, recent studies have highlighted the importance of parental attunement but there is little research considering solely the affective-bodily dimension. Therefore, the aim of the present study was to explore the parental attunement in fathers of children with ASD observed during play interactions and to investigate the relationship between paternal attunement and the perception of their psychological characteristics related to care relationships. The results highlight that fathers who describe themselves as better in affective care and sensitivity toward others more likely have an absence of paternal attunement during play interactions. The data presented are discussed in the light of intervention hypotheses, aimed at improving the relationship between fathers and children with autism.

As with most Machine Learning systems, recommender systems are typically evaluated through performance metrics computed over held-out data points. However, real-world behavior is undoubtedly nuanced: ad hoc error analysis and deployment-specific tests must be employed to ensure the desired quality in actual deployments. In this paper, we propose RecList, a behavioral-based testing methodology. RecList organizes recommender systems by use case and introduces a general plug-and-play procedure to scale up behavioral testing. We demonstrate its capabilities by analyzing known algorithms and black-box commercial systems, and we release RecList as an open source, extensible package for the community., Comment: Paper accepted to the WebConf 2022

Many data sets (e.g., reviews, forums, news, etc.) exist parallelly in multiple languages. They all cover the same content, but the linguistic differences make it impossible to use traditional, bag-of-word-based topic models. Models have to be either single-language or suffer from a huge, but extremely sparse vocabulary. Both issues can be addressed by transfer learning. In this paper, we introduce a zero-shot cross-lingual topic model. Our model learns topics on one language (here, English), and predicts them for unseen documents in different languages (here, Italian, French, German, and Portuguese). We evaluate the quality of the topic predictions for the same document in different languages. Our results show that the transferred topics are coherent and stable across languages, which suggests exciting future research directions., Comment: Updated version. Published as a conference paper at EACL2021

Introduction Electrical source imaging (ESI) and especially hdEEG represent a noninvasive, low cost and accurate method of localizing epileptic zone (EZ). Such capability can greatly increase seizure freedom rate in surgically treated drug resistant epilepsy cases. Furthermore, ESI might be important in intracranial record planning. Case report We report the case of a 15 years old boy suffering from drug resistant epilepsy with a previous history of DNET removal. The patient suffered from heterogeneous seizure semiology characterized by anesthesia and loss of tone in the left arm, twisting of the jaw to the left and dysarthria accompanied by daze; lightheadedness sometimes associated with headache and dizziness and at a relatively short time distance negative myoclonus involving the left hand. Clinical evidence poorly match scalp and video EEG monitoring thus requiring hdEEG recording followed by SEEG to define surgical target. Surgery was also guided by ECoG and obtained seizure freedom. Discussion ESI offers an excellent estimate of EZ, being hdEEG and intracranial recordings especially important in defining it. We analyzed our results together with the data from the literature showing how in children hdEEG might be even more crucial than in adults due to the heterogeneity in seizures phenomenology. The complexity of each case and the technical difficulties in dealing with children, stress even more the importance of a noninvasive tool for diagnosis. In fact, hdEEG not only guided in the presented case SEEG planning but may also in the future offer the possibility to replace it.

The identification of switched systems is a challenging problem, which entails both combinatorial (sample-mode assignment) and continuous (parameter estimation) features. A general framework for this problem has been recently developed, which alternates between parameter estimation and sample-mode assignment, solving both tasks to global optimality under mild conditions. This article extends this framework to the nonlinear case, which further aggravates the combinatorial complexity of the identification problem, since a model structure selection task has to be addressed for each mode of the system. To solve this issue, we reformulate the learning problem in terms of the optimization of a probability distribution over the space of all possible model structures. Then, a randomized approach is employed to tune this distribution. The performance of the proposed approach on some benchmark examples is analyzed in detail.

Spotting a lie is challenging but has an enormous potential impact on security as well as private and public safety. Several NLP methods have been proposed to classify texts as truthful or deceptive. In most cases, however, the target texts' preceding context is not considered. This is a severe limitation, as any communication takes place in context, not in a vacuum, and context can help to detect deception. We study a corpus of Italian dialogues containing deceptive statements and implement deep neural models that incorporate various linguistic contexts. We establish a new state-of-the-art identifying deception and find that not all context is equally useful to the task. Only the texts closest to the target, if from the same speaker (rather than questions by an interlocutor), boost performance. We also find that the semantic information in language models such as BERT contributes to the performance. However, BERT alone does not capture the implicit knowledge of deception cues: its contribution is conditional on the concurrent use of attention to learn cues from BERT's representations.

Word embeddings (e.g., word2vec) have been applied successfully to eCommerce products through~\textit{prod2vec}. Inspired by the recent performance improvements on several NLP tasks brought by contextualized embeddings, we propose to transfer BERT-like architectures to eCommerce: our model -- ~\textit{Prod2BERT} -- is trained to generate representations of products through masked session modeling. Through extensive experiments over multiple shops, different tasks, and a range of design choices, we systematically compare the accuracy of~\textit{Prod2BERT} and~\textit{prod2vec} embeddings: while~\textit{Prod2BERT} is found to be superior in several scenarios, we highlight the importance of resources and hyperparameters in the best performing models. Finally, we provide guidelines to practitioners for training embeddings under a variety of computational and data constraints., Comment: Updated version. Published as a workshop paper at ECNLP 4 at ACL-IJCNLP 2021

Quantum annealing has emerged in the last few years as a promising quantum computing approach to solving large-scale combinatorial optimization problems. In this paper, we formulate the problem of correctly placing pressure sensors on a Water Distribution Network (WDN) as a combinatorial optimization problem in the form of a Quadratic Unconstrained Binary Optimization (QUBO) or Ising model. Optimal sensor placement is indeed key to detect and isolate fault events. We outline the QUBO and Ising formulations for the sensor placement problem starting from the network topology and few other features. We present a detailed procedure to solve the problem by minimizing its Hamiltonian using PyQUBO, an open-source Python Library. We then apply our methods to the case of a real Water Distribution Network. Both simulated annealing and a hybrid quantum-classical approach on a D-Wave machine are employed., Comment: 10 pages, 3 figures. v2: minor corrections

Language models have revolutionized the field of NLP. However, language models capture and proliferate hurtful stereotypes, especially in text generation. Our results show that 4.3% of the time, language models complete a sentence with a hurtful word. These cases are not random, but follow language and gender-specific patterns. We propose a score to measure hurtful sentence completions in language models (HONEST). It uses a systematic template- and lexicon-based bias evaluation methodology for six languages. Our findings suggest that these models replicate and amplify deep-seated societal stereotypes about gender roles. Sentence completions refer to sexual promiscuity when the target is female in 9% of the time, and in 4% to homosexuality when the target is male. The results raise questions about the use of these models in production settings.

Breads were prepared by substituting common wheat flour with 0 (GP0), 5 (GP5) and 10 (GP10) g/100 g (w/w) of grape pomace powder (GPP) and were analyzed for the phenolic profile bioaccessibility as well as the in vitro starch digestion during simulated digestion. The free and bound phenolic composition of native GPP and resulting breads were profiled using ultra-high-performance chromatography-quadrupole-time-of-flight (UHPLC-QTOF). The raw GPP was characterized by 190 polyphenols with the anthocyanins representing the most abundant class, accounting for 11.60 mg/g of cyanidin equivalents. Regarding the fortified bread, the greatest (p &lt, 0.05) content in phenolic compounds was recorded for the GP10 sample (considering both bound and free fractions) being 127.76 mg/100 g dry matter (DM), followed by the GP5 (106.96 mg/100 g DM), and GP0 (63.76 mg/100 g DM). The use of GPP determined an increase of anthocyanins (considered the markers of the GPP inclusion), recording 20.98 mg/100 g DM in GP5 and 35.82 mg/100 g DM in GP10. The bioaccessibility of anthocyanins increased in both GP5 and GP10 breads when moving from the gastric to the small intestine in vitro digestion phase with an average value of 24%. Both the starch hydrolysis and the predicted glycemic index decreased with the progressive inclusion of GPP in bread. Present findings showed that GPP in bread could promote an antioxidant environment in the digestive tract and influence the in vitro starch digestion.

Matching tables against Knowledge Graphs is a crucial task in many applications. A widely adopted solution to improve the precision of matching algorithms is to refine the set of candidate entities by their type in the Knowledge Graph. However, it is not rare that a type is missing for a given entity. In this paper, we propose a methodology to improve the refinement phase of matching algorithms based on type prediction and soft constraints. We apply our methodology to state-of-the-art algorithms, showing a performance boost on different datasets.

District heating networks (DHNs) are promising technologies to increase the efficiency and reduce emissions of heat distribution to residential and commercial buildings. The advent of the smart grid paradigm has introduced the usage of heating load forecasting tools in DHNs. They provide estimates of future heating load, improving the planning of heat production and power station maintenance. In this article, we propose a methodology based on the integrated use of regularized regression and clustering for generating predictive models of future heating load in DHNs. The methodology is tested on a real case study based on a dataset provided by AGSM, an Italian utility company that manages a DHN in the city of Verona, Italy. We generate a set of multiple-equation models having different degrees of complexity and show that models generated by the proposed approach outperform those trained by standard methods. Moreover, we provide an interpretation of patterns encoded by these models, and show that they identify real operational states of the network. The approach is completely data driven.

The thermo-elastic properties of a novel class of ceramic composites based on an ultra-refractory matrix and containing short carbon fibre to provide failure tolerance and thermal shock resistance were experimentally characterised and numerically simulated. The composites samples were fabricated with fibre volume fractions ranging from 5 to 60 vol% and were featured by homogeneous dispersion of fibre along the basal plane due to the processing route. A random sequential adsorption algorithm was implemented to generate Representative Volume Elements (RVEs), following which finite element (FE) analyses were conducted to determine the elastic modulus and the coefficient of thermal expansion of the composites . Comparison of the experimental and modelling results showed good agreement for the modulus at low fibre volume fractions, and for CTE at higher volume fractions. Discrepancies are attributed to significant evolution of the composite microstructure upon processing at high temperatures, as revealed by microscopic analysis. The observations emphasise the need to account for the effects of microstructural changes on the constituent properties in the modelling of the thermo-mechanical properties in sintered ceramic composites.

Forecasting is an important task for intelligent agents involved in dynamical processes. A specific application domain concerns district heating networks, in which the future heating load generated by centralized power plants and distributed to buildings must be optimized for better plant maintenance, energy consumption and environmental impact. In this paper we present XM_HeatForecast a Python tool designed to support district heating network operators. The tool provides an integrated architecture for i) generating and updating in real-time predictive models of heating load, ii) supporting the analysis of prediction performance and errors, iii) inspecting model parameters and analyzing the historical dataset from which models are trained. A case study is presented in which the software is used on a synthetic dataset of heat loads and weather forecast from which a regression model is generated and updated every 24 h, while predictions of load in the next 48 h are performed every hour.