Showing papers by "IBM published in 2022"

LegalGNN: Legal Information Enhanced Graph Neural Network for Recommendation

Abstract: Recommendation in legal scenario (Legal-Rec) is a specialized recommendation task that aims to provide potential helpful legal documents for users. While there are mainly three differences compared...

Crown Nanopores in Graphene for CO₂ Capture and Filtration

Abstract: With growing concerns about global warming, it has become urgent and critical to capture carbon from various emission sources (such as power plants) and even directly from air. Recent advances in materials research permit the design of various efficient approaches for capturing CO2 with high selectivity over other gases. Here, we show that crown nanopores (resembling crown ethers) embedded in graphene can efficaciously allow CO2 to pass and block other flue gas components (such as N2 and O2). We carried out extensive density functional theory-based calculations as well as classical and ab initio molecular dynamics simulations to reveal the energetics and dynamics of gas transport through crown nanopores. Our results highlight that the designed crown nanopores in graphene possess not only an excellent selectivity for CO2 separation/capture but also fast transport (flow) rates, which are ideal for the treatment of flue gas in power plants.

A Comprehensive Guideline for Bengali Sentiment Annotation

Abstract: Sentiment Analysis (SA) is a Natural Language Processing (NLP) and an Information Extraction (IE) task that primarily aims to obtain the writer’s feelings expressed in positive or negative by analy...

Large-scale chemical language representations capture molecular structure and properties

Abstract: Models based on machine learning can enable accurate and fast molecular property predictions, which is of interest in drug discovery and material design. Various supervised machine learning models have demonstrated promising performance, but the vast chemical space and the limited availability of property labels make supervised learning challenging. Recently, unsupervised transformer-based language models pretrained on a large unlabelled corpus have produced state-of-the-art results in many downstream natural language processing tasks. Inspired by this development, we present molecular embeddings obtained by training an efficient transformer encoder model, MoLFormer, which uses rotary positional embeddings. This model employs a linear attention mechanism, coupled with highly distributed training, on SMILES sequences of 1.1 billion unlabelled molecules from the PubChem and ZINC datasets. We show that the learned molecular representation outperforms existing baselines, including supervised and self-supervised graph neural networks and language models, on several downstream tasks from ten benchmark datasets. They perform competitively on two others. Further analyses, specifically through the lens of attention, demonstrate that MoLFormer trained on chemical SMILES indeed learns the spatial relationships between atoms within a molecule. These results provide encouraging evidence that large-scale molecular language models can capture sufficient chemical and structural information to predict various distinct molecular properties, including quantum-chemical properties. Large language models have recently emerged with extraordinary capabilities, and these methods can be applied to model other kinds of sequence, such as string representations of molecules. Ross and colleagues have created a transformer-based model, trained on a large dataset of molecules, which provides good results on property prediction tasks.

Association of Blood Biochemical Indexes and Antibiotic Exposure With Severe Immune-related Adverse Events in Patients With Advanced Cancers Receiving PD-1 Inhibitors

Abstract: Some patients with cancer treated with programmed death 1 (PD-1) inhibitors experience immune-related severe adverse events (ir-SAEs), however, predictors are limited. The objective was to identify clinicopathologic features that may be associated with a higher ir-SAE risk. This was a nested case-control study. After screening a total of 832 PD-1 inhibitor-treated patients, we identified 42 ir-SAE cases. According to the Common Terminology Criteria for Adverse Events (CTCAE) version 5.0, ir-SAEs were defined as grade ≥3 toxic effects associated with immunotherapy. A total of 126 controls were matched. The crude and adjusted risks of ir-SAEs were estimated by odds ratio (ORs) and 95% CIs using multivariate logistic regression models. Baseline neutrophil-to-lymphocyte ratio (NLR) [per SD increment-adjusted (aOR): 1.16], lactate dehydrogenase (LDH) ≥245 U/L (aOR: 2.39), and antibiotic exposure (aOR: 4.39) were associated with a higher risk of ir-SAEs. When NLR was categorized in 3 groups, significantly higher risks of ir-SAEs (aOR: 4.95) were found in participants in group 3 (>6) than in those in group 1 (<3). Furthermore, NLR (per SD increment-adjusted hazard ratio:1.08) were also significantly associated with shorter overall survival (OS). Baseline LDH ≥245 U/L and antibiotic exposure were no significant association with OS. In conclusion, ir-SAEs were associated between baseline NLR, LDH ≥245 U/L and antibiotic exposure. Lower NLR was correlated with longer OS for cancer.

Programmable and Open-Access Millimeter-Wave Radios in the PAWR COSMOS Testbed

Abstract: While millimeter-wave (mmWave) wireless has recently gained tremendous attention with the transition to 5G, developing a broadly accessible experimental infrastructure will allow the research community to make significant progress in this area. Hence, in this paper, we present the design and implementation of various programmable and open-access 28/60 GHz software-defined radios (SDRs), deployed in the PAWR COSMOS advanced wireless testbed. These programmable mmWave radios are based on the IBM 28 GHz 64-element dual-polarized phased array antenna module (PAAM) subsystem board and the Sivers IMA 60 GHz WiGig transceiver. These front ends are integrated with USRP SDRs or Xilinx RF-SoC boards, which provide baseband signal processing capabilities. Moreover, we present measurements of the TX/RX beamforming performance and example experiments (e.g., real-time channel sounding and RFNoC-based 802.11ad preamble detection), using the mmWave radios. Finally, we discuss ongoing enhancement and development efforts focusing on these radios.

Artificial Intelligence and Machine Learning Technologies in Cancer Care: Addressing Disparities, Bias, and Data Diversity

Abstract: Artificial intelligence (AI) and machine learning (ML) technologies have not only tremendous potential to augment clinical decision-making and enhance quality care and precision medicine efforts, but also the potential to worsen existing health disparities without a thoughtful, transparent, and inclusive approach that includes addressing bias in their design and implementation along the cancer discovery and care continuum. We discuss applications of AI/ML tools in cancer and provide recommendations for addressing and mitigating potential bias with AI and ML technologies while promoting cancer health equity.

Batched Learning in Generalized Linear Contextual Bandits With General Decision Sets

Abstract: In real-world adaptive personalized decision making, due to physical and/or resource constraints, a decision maker often does not have the luxury of immediately incorporating the feedback from the previous individual into forming new policies for future individuals. This is an important aspect that has largely been abstracted away from the traditional online learning/decision making literature. In this letter, we study the problem of batched learning in generalized linear contextual bandits where the decision maker, unlike in traditional online learning, can only access feedback at the end of a limited number of batches, and when selecting actions within a batch, can only use information from prior batches. We provide a lower bound that characterizes the fundamental limit of performance in this setting and then give a UCB-based batched learning algorithm whose regret bound, obtained using a self-normalized martingale style analysis, nearly matches this lower bound. Our results provide a novel inquiry into generalized linear contextual bandits with arbitrary action sets, which include several bandits setting as special cases and thus shed light on batch-constrained decision making in general.

Measuring and Modeling Group Dynamics in Open-Source Software Development: A Tensor Decomposition Approach

Abstract: Many open-source software projects depend on a few core developers, who take over both the bulk of coordination and programming tasks. They are supported by peripheral developers, who contribute ei...

A Real-world Claims Data Analysis of Meningococcal Serogroup B Vaccine Series Completion and Potential Missed Opportunities in the United States

Abstract: In the United States, meningococcal serogroup B (MenB) vaccination is recommended for 16-23-year-olds based on shared clinical decision-making. We estimated series completion among individuals initiating MenB vaccination for the 2 available vaccines: MenB 4-component (MenB-4C, doses at 0 and ≥1 month) and MenB factor H binding protein (MenB-FHbp, doses at 0 and 6 months).This retrospective health insurance claims data analysis included 16-23-year-olds who initiated MenB vaccination (index date) during January 2017 to November 2018 (MarketScan Commercial Claims and Encounters Database) or January 2017 to September 2018 (MarketScan Multi-State Medicaid Database) and had continuous enrollment for ≥6 months before and ≥15 months after index. The main outcome was MenB vaccine series completion within 15 months. Among noncompleters, preventive care/well-child and vaccine administrative office visits were identified as potential missed opportunities for series completion. Robust Poisson regression models identified independent predictors of series completion.In the Commercial (n = 156,080) and Medicaid (n = 57,082) populations, series completion was 56.7% and 44.7%, respectively, and was higher among those who initiated MenB-4C versus MenB-FHbp (61.1% versus 49.8% and 47.8% versus 33.9%, respectively; both P < 0.001). Among noncompleters, 40.2% and 34.7% of the Commercial and Medicaid populations, respectively, had ≥1 missed opportunity for series completion. Receipt of MenB-4C and younger age were independently associated with a higher probability of series completion.Series completion rates were suboptimal but were higher among those who initiated MenB-4C. To maximize the benefits of MenB vaccination, interventions to improve completion and reduce missed opportunities should be implemented.

Medication-related osteonecrosis of the jaw: An update

Abstract: Antiresorptive medications, such as bisphosphonates and denosumab, are an important class of medication used to treat a wide range of diseases from osteoporosis to multiple myeloma. Unfortunately, they are also associated with a rare but devastating side effect - medication-related osteonecrosis of the jaw (MRONJ). First reported in 2003, much research has been done into the area; however, the exact pathophysiology continues to elude clinicians and researchers. What has been ascertained is that intravenous treatment, duration of treatment, and tooth extraction are major risk factors. Staging and treatment guidelines have been proposed; however, there has been no universal acceptance, and clinicians rely on various position papers. Over the next 30 years, the aging population is set to double, and with it, the prescription of antiresorptive medication and incidence of MRONJ will undoubtedly increase. In 2013, Gupta et al. published a paper on bisphosphonate-related osteonecrosis of the jaw; however, there have many changes since then. This paper aims to provide a succinct update on those changes.

BAT Framework Modeling of RMG HKMG GAA-SNS FETs

Abstract: In this chapter, the BAT framework presented in Chaps. 4– 6 is used to analyze and model the measured NBTI time kinetics in Silicon channel GAA-SNS FETs with RMG HKMG gate insulator stack. The ultra-fast measured stress and recovery data at different stress bias and temperature are modeled in devices having different sheet dimensions (length and width). The changes in voltage acceleration and temperature activation for changes in the sheet dimensions are modeled. The calibrated BAT framework is used to determine the impact of dimension scaling on the extrapolated EOL degradation under use condition.

Magnetic Biosignatures of Magnetosomal Greigite From Micromagnetic Calculation

Abstract: Greigite magnetosomes produced by magnetotactic bacteria (MTB) are widely distributed in natural environments, but large uncertainties remain regarding their magnetic biosignatures. Here, we have constructed micromagnetic models with realistic biogenic greigite particles to quantify the magnetic properties and magnetotaxis efficiency of greigite-producing MTB cells. Our calculations suggest coercivity (Bc) of ∼15–21 mT for intact greigite-producing rod-shaped MTB and many-celled magnetotactic prokaryotes, with Bc decreasing to ∼11 mT for greigite magnetofossils with clumped particles. These magnetic signatures make biogenic greigite distinguishable from typical biogenic magnetite and inorganic greigite, providing reliable magnetic criteria to detect biogenic greigite in a wide range of environmental and geological settings. Our numerical calculations suggest that rod-shaped greigite-producing MTB have a similar magnetotaxis efficiency to magnetite MTB, likely by biomineralizing more greigite crystals to compensate for the lower saturation magnetization of greigite and less ordered chains in greigite MTB cells, demonstrating biological-controlled optimization of their magnetic nanostructure.

Bending and Collapse: Magnetic Recording Fidelity of Magnetofossils From Micromagnetic Simulation

Abstract: Magnetosome chains produced by magnetotactic bacteria are important paleoenvironmental and paleomagnetic recorders. It has been shown that magnetic properties of magnetosome chains are closely related to their morphology and chain structures; however, the in situ structures of magnetosome chains in sediments (magnetofossils) are not known. Magnetosome chains are subject to various deformations after cell dissolution and are therefore unlikely to be fully intact, obscuring their original magnetic signals. Here, we use finite element micromagnetic simulations to quantify changes in magnetic signals in response to chain deformation, in particular, as a function of variable degrees of bending and collapse. Our results indicate that bending/collapse leads to a significant coercivity reduction and domain state transition of the chain. Therefore, hysteresis parameters can be used to assess the degree of chain bending/collapse in magnetofossil-rich sediments. Calculations of the contributions of chain bending/collapse to the post-depositional remanent magnetization (pDRM) of magnetofossils indicate that pDRM remains both faithful to the pre-bending/collapse natural remanent magnetization, and that the remanence of some structurally deformed magnetofossil assemblages remains thermally stable over billion-year timescales, suggesting that even strongly deformed magnetosome chains in ancient geological materials retain faithful paleomagnetic records and thus have potentials for tracing ancient geomagnetic field variations and microbial activities on early Earth.

Synoptic ¹²⁹ I and CFC–SF ₆ Transit Time Distribution (TTD) Sections Across the Central Arctic Ocean From the 2015 GEOTRACES Cruises

Abstract: Measurements of the tracers, 129I, CFC-11, and SF6 on water samples collected in the Arctic Ocean in 2015 have been used to calculate mean ages, Γ and mixing, Δ parameters using transit time distributions (TTDs) to constrain water circulation and mixing time scales. Values of Γ and Δ determined separately using the two tracer pairs, SF6-CFC-11, and 129I-CFC-11 are in good agreement for gas solubilities estimated for saturation levels of 0.90, but agreement decreases for other gas saturation levels. Both Γ and Δ increase rapidly with increasing depth below the base of the intermediate water layer (ca. 1,000 m), but maintaining a value of Δ/Γ ≅ 1 supporting the use of this proportionality in applications of TTDs to deep ocean transport of substances such as anthropogenic carbon. Isolines of Γ = 20 years deepening to depths below 1,000 m over the flank of the Mendeleyev Ridge near the North Pole outline the bathymetrically steered, return flow of recently ventilated Atlantic Water toward Fram Strait. Basin interior waters are significantly older with the Γ = 25 years mean age isoline shallowing upward to depths above 500 m in the Makarov, Canada, and Eurasian Basins. Values of Δ remain relatively constant in the 6–10 years range in upper intermediate water across all three basins indicating that flow is principally advective and that the mixing specified by Δ likely occurs upstream of the central basins in regions proximal to the outflow from the Santa Anna Trough.

The effects of spectral dimensionality reduction on hyperspectral pixel classification: A case study

Abstract: This paper presents a systematic study of the effects of hyperspectral pixel dimensionality reduction on the pixel classification task. We use five dimensionality reduction methods -- PCA, KPCA, ICA, AE, and DAE -- to compress 301-dimensional hyperspectral pixels. Compressed pixels are subsequently used to perform pixel classifications. Pixel classification accuracies together with compression method, compression rates, and reconstruction errors provide a new lens to study the suitability of a compression method for the task of pixel classification. We use three high-resolution hyperspectral image datasets, representing three common landscape types (i.e. urban, transitional suburban, and forests) collected by the Remote Sensing and Spatial Ecosystem Modeling laboratory of the University of Toronto. We found that PCA, KPCA, and ICA post greater signal reconstruction capability; however, when compression rates are more than 90\% these methods show lower classification scores. AE and DAE methods post better classification accuracy at 95\% compression rate, however their performance drops as compression rate approaches 97\%. Our results suggest that both the compression method and the compression rate are important considerations when designing a hyperspectral pixel classification pipeline.

Accuracy of a Method to Monitor Root Position Using a 3D Digital Crown/Root Model during Orthodontic Treatments

Abstract: This study aimed to assess the accuracy of a method of predicting post-movement root position during orthodontic treatment using a 3D digital crown/root model (3DCRM) created with pre-movement records of both cone-beam computed tomography (CBCT) and dental arch digital scans. Pre- and post-movement CBCT scans and dental arch digital scans of five patients who had completed orthodontic treatments were used in this study. The 3DCRM was superimposed onto the post-movement scanned dental arch to identify the post-movement root position (test method). Post-movement CBCT (referenced as the current method) served as the control to identify the actual post-movement root position. 3D-coordinate analysis revealed no significant differences between the test and current methods along the X and Y axes. However, the discrepancy on the Z axis (especially in cases of intrusion) was greater than that in all other directions for all three tooth types examined (p < 0.05). A strong positive correlation between the degree of discrepancy and the distance of tooth movement was observed on the Z axis (r = 0.71). The 3DCRM method showed promising potential to accurately predict root position during orthodontic treatments without the need for a second CBCT. However, root resorption, which affected the Z axis prediction, needs to be closely monitored using periapical radiographs to complement this method.

Optics enabled networks and architectures for data center cost and power efficiency

Abstract: Bandwidth demand for data center networks continues as performance increases and is further fueled by the exploding demand for artificial intelligence and new high-performance computing workloads. Managing power and costs will require a range of solutions including new networking and workload specialized architectures, composable systems, and optical circuit switching. In this paper, we focus primarily on two topics: examining the benefits of flatter networks (enabled mainly by means of co-packaged-optics-enabled switches) and the utilization improvement potential for composable (disaggregated) systems, while discussing specialized hardware and networks, and optical circuit switching more briefly.

C3PO: Cloud-based Confidentiality-preserving Continuous Query Processing

Abstract: With the advent of the Internet of things (IoT), billions of devices are expected to continuously collect and process sensitive data (e.g., location, personal health factors). Due to the limited co...

All about Nitrite: Exploring Nitrite Sources and Sinks in the Eastern Tropical North Pacific Oxygen Minimum Zone

Abstract: Abstract. Oxygen minimum zones (OMZs), due to their large volumes of perennially deoxygenated waters, are critical regions for understanding how the interplay between anaerobic and aerobic nitrogen (N) cycling microbial pathways affects the marine N budget. Here we present a suite of measurements of the most significant OMZ N cycling rates, which all involve nitrite (NO2–) as a product, reactant, or intermediate, in the Eastern Tropical North Pacific (ETNP) OMZ. These measurements and comparisons to data from previously published OMZ cruises present additional evidence that NO3– reduction is the predominant OMZ N flux, followed by NO2– oxidation back to NO3–. The combined rates of both of these N recycling processes were observed to be much greater (up to nearly 200x) than the combined rates of the N loss processes of anammox and denitrification, especially in waters near the anoxic / oxic interface. We also show that NO2– oxidation can occur in functionally anoxic incubations, measurements that further strengthen the case for truly anaerobic NO2– oxidation. We also evaluate the possibility that NO2– dismutation provides the oxidative power for anaerobic NO2– oxidation. Although almost all treatments returned little evidence for dismutation (as based on product inhibition, substrate stimulation, and stoichiometric hypotheses), results from one treatment under conditions closest to in situ NO2– values may support the occurrence of NO2– dismutation. The partitioning of N loss between anammox and denitrification differed widely from stoichiometric predictions of at most 29 % anammox; in fact, N loss rates at many depths consisted entirely of anammox. Through investigating the magnitudes of NO3– reduction and NO2– oxidation, testing for anaerobic NO2– oxidation, examining the possibility of NO2– dismutation, and further documenting the balance of N loss processes, these new data shed light on many open questions in OMZ N cycling research.

The Devil's Party

Abstract: This chapter provides a selective overview of the relationship between metal and literature. It seeks to demonstrate the depth and diversity of the topic, while at the same time developing a more specific argument: that the distinction between ‘literary’ and ‘genre’ writing is crucial to conceptualizing this kind of intermedial exchange. Whereas the feedback relationship between metal and genre fiction is a well-established and fruitful one, metal’s love for the literary remains largely unrequited, especially when compared to other styles of music that share an association with anti-establishment danger such as punk and hip hop. Nevertheless, metal has drawn sustenance from literature since in its inception and the chapter explores some strategically chosen examples of these complex, richly creative borrowings and adaptations. It draws on a variety of mainstream and underground bands, including Mastodon and Watain, as well as original interviews with the vocalists Kat Katz and J. R. Hayes.

Rock Magnetic Signatures of Hydrothermal Mineralization in the Trans‐Atlantic Geotraverse (TAG) Hydrothermal Field

Abstract: The Ocean Drilling Program Leg 158 drill holes from the Trans-Atlantic Geotraverse hydrothermal field are investigated to understand the rock magnetic signatures of hydrothermal mineralization. A composite columnar section has been constructed through hole correlation to understand the stratigraphic variation of magnetomineralogy within the stockwork. Isothermal remanent magnetization components unmixing, first-order reversal curve diagrams, low-temperature magnetic signatures, and electron microscopic analyses disclose magnetic minerals of disparate occurrences related to predominating hydrothermal mineralization reactions in three broad zones: For basaltic basements, serpentinization of olivine phenocrysts during preliminary hydrothermal alteration produces magnetite, in addition to primary titanomagnetite; Chloritized and silicified zone samples contain relict titanomagnetite and exsolved magnetite that survived hydrothermal dissolution; Anhydrite and sulfide zone samples are dominated by magnetite and hematite, likely from oxidation of polymetallic sulfides due to exposure in oxidative seafloor environments during drilling. Our findings suggest that seafloor oxidation potentially modifies the magnetic properties of polymetallic sulfides in hydrothermal deposits, which applies to magnetic tomography of sophisticated subseafloor vent structures and prospecting seafloor massive sulfides (SMS) deposits therein. Meanwhile, we alert future deep-sea mining that drilling may promote physicochemical alteration of SMS deposits, causing environmental risks. The established magnetic signatures ultimately contribute to understanding the in situ geological preservation of SMS deposits and optimizing exploitation procedures in the future.