Real-time low latency estimation of brain rhythms with deep neural networks.

Objective.Neurofeedback and brain-computer interfacing technology open the exciting opportunity for establishing interactive closed-loop real-time communication with the human brain. This requires interpreting brain's rhythmic activity and generating timely feedback to the brain. Lower delay between neuronal events and the appropriate feedback increases the efficacy of such interaction. Novel more efficient approaches capable of tracking brain rhythm's phase and envelope are needed for scenarios that entail instantaneous interaction with the brain circuits.Approach.Isolating narrow-band signals incurs fundamental delays. To some extent they can be compensated using forecasting models. Given the high quality of modern time series forecasting neural networks we explored their utility for low-latency extraction of brain rhythm parameters. We tested five neural networks with conceptually distinct architectures in forecasting synthetic EEG rhythms. The strongest architecture was then trained to simultaneously filter and forecast EEG data. We compared it against the state-of-the-art techniques using synthetic and real data from 25 subjects.Main results.The temporal convolutional network (TCN) remained the strongest forecasting model that achieved in the majority of testing scenarios>90% rhythm's envelope correlation with<10 ms effective delay and<20∘circular standard deviation of phase estimates. It also remained stable enough to noise level perturbations. Trained to filter and predict the TCN outperformed the cFIR, the Kalman filter based state-space estimation technique and remained on par with the larger Conv-TasNet architecture.Significance.Here we have for the first time demonstrated the utility of the neural network approach for low-latency narrow-band filtering of brain activity signals. Our proposed approach coupled with efficient implementation enhances the effectiveness of brain-state dependent paradigms across various applications. Moreover, our framework for forecasting EEG signals holds promise for investigating the predictability of brain activity, providing valuable insights into the fundamental questions surrounding the functional organization and hierarchical information processing properties of the brain.

Electronic Structure and Minimal Models for Flat and Corrugated CuO Monolayers: An Ab Initio Study.

CuO atomic thin monolayer (mlCuO) was synthesized recently. Interest in the mlCuO is based on its close relation to CuO2 layers in typical high temperature cuprate superconductors. Here, we present the calculation of the band structure, the density of states and the Fermi surface of the flat mlCuO as well as the corrugated mlCuO within the density functional theory (DFT) in the generalized gradient approximation (GGA). In the flat mlCuO, the Cu-3dx2-y2 band crosses the Fermi level, while the Cu-3dxz,yz hybridized band is located just below it. The corrugation leads to a significant shift of the Cu-3dxz,yz hybridized band down in energy and a degeneracy lifting for the Cu-3dx2-y2 bands. Corrugated mlCuO is more energetically favorable than the flat one. In addition, we compared the electronic structure of the considered CuO monolayers with bulk CuO systems. We also investigated the influence of a crystal lattice strain (which might occur on some interfaces) on the electronic structure of both mlCuO and determined the critical strains of topological Lifshitz transitions. Finally, we proposed a number of different minimal models for the flat and the corrugated mlCuO using projections onto different Wannier functions basis sets and obtained the corresponding Hamiltonian matrix elements in a real space.

Analysis of the Human Protein Atlas Weakly Supervised Single-Cell Classification competition.

While spatial proteomics by fluorescence imaging has quickly become an essential discovery tool for researchers, fast and scalable methods to classify and embed single-cell protein distributions in such images are lacking. Here, we present the design and analysis of the results from the competition Human Protein Atlas - Single-Cell Classification hosted on the Kaggle platform. This represents a crowd-sourced competition to develop machine learning models trained on limited annotations to label single-cell protein patterns in fluorescent images. The particular challenges of this competition include class imbalance, weak labels and multi-label classification, prompting competitors to apply a wide range of approaches in their solutions. The winning models serve as the first subcellular omics tools that can annotate single-cell locations, extract single-cell features and capture cellular dynamics.

Self-supervised recurrent depth estimation with attention mechanisms.

Depth estimation has been an essential task for many computer vision applications, especially in autonomous driving, where safety is paramount. Depth can be estimated not only with traditional supervised learning but also via a self-supervised approach that relies on camera motion and does not require ground truth depth maps. Recently, major improvements have been introduced to make self-supervised depth prediction more precise. However, most existing approaches still focus on single-frame depth estimation, even in the self-supervised setting. Since most methods can operate with frame sequences, we believe that the quality of current models can be significantly improved with the help of information about previous frames. In this work, we study different ways of integrating recurrent blocks and attention mechanisms into a common self-supervised depth estimation pipeline. We propose a set of modifications that utilize temporal information from previous frames and provide new neural network architectures for monocular depth estimation in a self-supervised manner. Our experiments on the KITTI dataset show that proposed modifications can be an effective tool for exploiting temporal information in a depth prediction pipeline.

Temporal network embedding framework with causal anonymous walks representations.

Many tasks in graph machine learning, such as link prediction and node classification, are typically solved using representation learning. Each node or edge in the network is encoded via an embedding. Though there exists a lot of network embeddings for static graphs, the task becomes much more complicated when the dynamic (i.e., temporal) network is analyzed. In this paper, we propose a novel approach for dynamic network representation learning based on Temporal Graph Network by using a highly custom message generating function by extracting Causal Anonymous Walks. We provide a benchmark pipeline for the evaluation of temporal network embeddings. This work provides the first comprehensive comparison framework for temporal network representation learning for graph machine learning problems involving node classification and link prediction in every available setting. The proposed model outperforms state-of-the-art baseline models. The work also justifies their difference based on evaluation in various transductive/inductive edge/node classification tasks. In addition, we show the applicability and superior performance of our model in the real-world downstream graph machine learning task provided by one of the top European banks, involving credit scoring based on transaction data.

Fusion of text and graph information for machine learning problems on networks.

Today, increased attention is drawn towards network representation learning, a technique that maps nodes of a network into vectors of a low-dimensional embedding space. A network embedding constructed this way aims to preserve nodes similarity and other specific network properties. Embedding vectors can later be used for downstream machine learning problems, such as node classification, link prediction and network visualization. Naturally, some networks have text information associated with them. For instance, in a citation network, each node is a scientific paper associated with its abstract or title; in a social network, all users may be viewed as nodes of a network and posts of each user as textual attributes. In this work, we explore how combining existing methods of text and network embeddings can increase accuracy for downstream tasks and propose modifications to popular architectures to better capture textual information in network embedding and fusion frameworks.

Survey on graph embeddings and their applications to machine learning problems on graphs.

Dealing with relational data always required significant computational resources, domain expertise and task-dependent feature engineering to incorporate structural information into a predictive model. Nowadays, a family of automated graph feature engineering techniques has been proposed in different streams of literature. So-called graph embeddings provide a powerful tool to construct vectorized feature spaces for graphs and their components, such as nodes, edges and subgraphs under preserving inner graph properties. Using the constructed feature spaces, many machine learning problems on graphs can be solved via standard frameworks suitable for vectorized feature representation. Our survey aims to describe the core concepts of graph embeddings and provide several taxonomies for their description. First, we start with the methodological approach and extract three types of graph embedding models based on matrix factorization, random-walks and deep learning approaches. Next, we describe how different types of networks impact the ability of models to incorporate structural and attributed data into a unified embedding. Going further, we perform a thorough evaluation of graph embedding applications to machine learning problems on graphs, among which are node classification, link prediction, clustering, visualization, compression, and a family of the whole graph embedding algorithms suitable for graph classification, similarity and alignment problems. Finally, we overview the existing applications of graph embeddings to computer science domains, formulate open problems and provide experiment results, explaining how different networks properties result in graph embeddings quality in the four classic machine learning problems on graphs, such as node classification, link prediction, clustering and graph visualization. As a result, our survey covers a new rapidly growing field of network feature engineering, presents an in-depth analysis of models based on network types, and overviews a wide range of applications to machine learning problems on graphs.

A Predictive Model Towards Site-Selective Metalations of Functionalized Heterocycles, Arenes, Olefins, and Alkanes using TMPZnCl⋠LiCl.

The development of a predictive model towards site-selective deprotometalation reactions using TMPZnCl⋅LiCl is reported (TMP=2,2,6,6-tetramethylpiperidinyl). The pKa values of functionalized N-, S-, and O-heterocycles, arenes, alkenes, or alkanes were calculated and compared to the experimental deprotonation sites. Large overlap (>80 %) between the calculated and empirical deprotonation sites was observed, showing that thermodynamic factors strongly govern the metalation regioselectivity. In the case of olefins, calculated frozen state energies of the deprotonated substrates allowed a more accurate prediction. Additionally, various new N-heterocycles were analyzed and the metalation regioselectivities rationalized using the predictive model.

Online supervised attention-based recurrent depth estimation from monocular video.

Autonomous driving highly depends on depth information for safe driving. Recently, major improvements have been taken towards improving both supervised and self-supervised methods for depth reconstruction. However, most of the current approaches focus on single frame depth estimation, where quality limit is hard to beat due to limitations of supervised learning of deep neural networks in general. One of the way to improve quality of existing methods is to utilize temporal information from frame sequences. In this paper, we study intelligent ways of integrating recurrent block in common supervised depth estimation pipeline. We propose a novel method, which takes advantage of the convolutional gated recurrent unit (convGRU) and convolutional long short-term memory (convLSTM). We compare use of convGRU and convLSTM blocks and determine the best model for real-time depth estimation task. We carefully study training strategy and provide new deep neural networks architectures for the task of depth estimation from monocular video using information from past frames based on attention mechanism. We demonstrate the efficiency of exploiting temporal information by comparing our best recurrent method with existing image-based and video-based solutions for monocular depth reconstruction.

Dual network embedding for representing research interests in the link prediction problem on co-authorship networks.

We present a study on co-authorship network representation based on network embedding together with additional information on topic modeling of research papers and new edge embedding operator. We use the link prediction (LP) model for constructing a recommender system for searching collaborators with similar research interests. Extracting topics for each paper, we construct keywords co-occurrence network and use its embedding for further generalizing author attributes. Standard graph feature engineering and network embedding methods were combined for constructing co-author recommender system formulated as LP problem and prediction of future graph structure. We evaluate our survey on the dataset containing temporal information on National Research University Higher School of Economics over 25 years of research articles indexed in Russian Science Citation Index and Scopus. Our model of network representation shows better performance for stated binary classification tasks on several co-authorship networks.

Synthesis of Bicyclo[1.1.1]pentane Bioisosteres of Internal Alkynes and para-Disubstituted Benzenes from [1.1.1]Propellane.

We report a general preparation of arylated bicyclo[1.1.1]pentanes through the opening of [1.1.1]propellane with various arylmagnesium halides. After transmetalation with ZnCl2 and Negishi cross-coupling with aryl and heteroaryl halides, bis-arylated bicyclo[1.1.1]pentanes are obtained. These bis-arylated bicyclo[1.1.1]pentanes may be considered as bioisosteres of internal alkynes. Bioisosteres of tazarotene and the metabotropic glutamate receptor 5 (mGluR5) antagonist 2-methyl-6-(phenylethynyl)pyridine were prepared and their physicochemical properties were evaluated.

Zn-, Mg-, and Li-TMP Bases for the Successive Regioselective Metalations of the 1,5-Naphthyridine Scaffold (TMP=2,2,6,6-Tetramethylpiperidyl).

A set of successive regioselective metalations and functionalizations of the 1,5-naphthyridine scaffold are described. A combination of Zn-, Mg-, and Li-TMP (TMP=2,2,6,6-tetramethylpiperidyl) bases and the presence or absence of a Lewis acid (BF3 ⋅OEt2 ) allows the introduction of up to three substituents to the 1,5-naphthyridine core. Also, a novel "halogen dance" reaction was discovered upon metalation of an 8-iodo-2,4-trifunctionalized 1,5-naphthyridine allowing a fourth regioselective functionalization. Additionally, reactions leading to key 1,5-naphthyridines for the preparation of OLED materials and a potential antibacterial agent were performed.

Synthetic Applications and Mechanistic Studies of the Hydroxide-Mediated Cleavage of Carbon-Carbon Bonds in Ketones.

The hydroxide-mediated cleavage of ketones into alkanes and carboxylic acids has been reinvestigated and the substrate scope extended to benzyl carbonyl compounds. The transformation is performed with a 0.05 M ketone solution in refluxing xylene in the presence of 10 equiv of potassium hydroxide. The reaction constitutes a straightforward protocol for the synthesis of certain phenyl-substituted carboxylic acids from 2-phenylcycloalkanones. The mechanism was investigated by kinetic experiments which indicated a first order reaction in hydroxide and a full negative charge in the rate-determining step. The studies were complemented by a theoretical investigation where two possible pathways were characterized by DFT/M06-2X. The calculations showed that the scission takes place by nucleophilic attack of hydroxide on the ketone followed by fragmentation of the resulting oxyanion into the carboxylic acid and a benzyl anion.

Dehydrogenative Synthesis of Carboxylic Acids from Primary Alcohols and Hydroxide Catalyzed by a Ruthenium N-Heterocyclic Carbene Complex.

Primary alcohols have been reacted with hydroxide and the ruthenium complex [RuCl2(IiPr)(p-cymene)] to afford carboxylic acids and dihydrogen. The dehydrogenative reaction is performed in toluene, which allows for a simple isolation of the products by precipitation and extraction. The transformation can be applied to a range of benzylic and saturated aliphatic alcohols containing halide and (thio)ether substituents, while olefins and ester groups are not compatible with the reaction conditions. Benzylic alcohols undergo faster conversion than other substrates, and a competing Cannizzaro reaction is most likely involved in this case. The kinetic isotope effect was determined to be 0.67 using 1-butanol as the substrate. A plausible catalytic cycle was characterized by DFT/B3LYP-D3 and involved coordination of the alcohol to the metal, ß-hydride elimination, hydroxide attack on the coordinated aldehyde, and a second ß-hydride elimination to furnish the carboxylate.

Regioselective Transition-Metal-Free Allyl-Allyl Cross-Couplings.

Readily prepared allylic zinc halides undergo SN 2-type substitutions with allylic bromides in a 1:1 mixture of THF and DMPU providing 1,5-dienes regioselectively. The allylic zinc species reacts at the most branched end (γ-position) of the allylic system furnishing exclusively γ,α'-allyl-allyl cross-coupling products. Remarkably, the double bond stereochemistry of the allylic halide is maintained during the cross-coupling process. Also several functional groups (ester, nitrile) are tolerated. This cross-coupling of allylic zinc reagents can be extended to propargylic and benzylic halides. DFT calculations show the importance of lithium chloride in this substitution.

Palladium-catalyzed carbonylative couplings of vinylogous enolates: application to statin structures.

The first Pd-catalyzed carbonylative couplings of aryl and vinyl halides with vinylogous enolates are reported generating products derived from C-C bond formation exclusively at the γ-position. Good results were obtained with a dienolate derivative of acetoacetate (1,3-dioxin-4-one). These transformations occurred at room temperature and importantly with only stoichiometric carbon monoxide in a two-chamber reactor. The methodology was applied to the synthesis of two members of the statin family generating the cis-3,5-diol acid motif by a γ-selective carbonylation followed by a cis-stereoselective reduction of the 3,5-dicarbonyl acid intermediates.

Efficient fluoride-catalyzed conversion of CO2 to CO at room temperature.

A protocol for the efficient and selective reduction of carbon dioxide to carbon monoxide has been developed. Remarkably, this oxygen abstraction step can be performed with only the presence of catalytic cesium fluoride and a stoichiometric amount of a disilane in DMSO at room temperature. Rapid reduction of CO2 to CO could be achieved in only 2 h, which was observed by pressure measurements. To quantify the amount of CO produced, the reduction was coupled to an aminocarbonylation reaction using the two-chamber system, COware. The reduction was not limited to a specific disilane, since (Ph2MeSi)2 as well as (PhMe2Si)2 and (Me3Si)3SiH exhibited similar reactivity. Moreover, at a slightly elevated temperature, other fluoride salts were able to efficiently catalyze the CO2 to CO reduction. Employing a nonhygroscopic fluoride source, KHF2, omitted the need for an inert atmosphere. Substituting the disilane with silylborane, (pinacolato)BSiMe2Ph, maintained the high activity of the system, whereas the structurally related bis(pinacolato)diboron could not be activated with this fluoride methodology. Furthermore, this chemistry could be adapted to (13)C-isotope labeling of six pharmaceutically relevant compounds starting from Ba(13)CO3 in a newly developed three-chamber system.

Ruthenium-catalyzed self-coupling of primary and secondary alcohols with the liberation of dihydrogen.

The dehydrogenative self-condensation of primary and secondary alcohols has been studied in the presence of RuCl2(IiPr)(p-cymene). The conversion of primary alcohols into esters has been further optimized by using magnesium nitride as an additive, which allows the reaction to take place at a temperature and catalyst loading lower than those described previously. Secondary alcohols were dimerized into racemic ketones by a dehydrogenative Guerbet reaction with potassium hydroxide as the additive. The transformation gave good yields of the ketone dimers with a range of alkan-2-ols, whereas more substituted secondary alcohols were unreactive. The reaction proceeds by dehydrogenation to the ketone, followed by an aldol reaction and hydrogenation of the resulting enone.

Mechanistic investigation of the ruthenium-N-heterocyclic-carbene-catalyzed amidation of amines with alcohols.

The mechanism of the ruthenium-N-heterocyclic-carbene-catalyzed formation of amides from alcohols and amines was investigated by experimental techniques (Hammett studies, kinetic isotope effects) and by a computational study with dispersion-corrected density functional theory (DFT/M06). The Hammett study indicated that a small positive charge builds-up at the benzylic position in the transition state of the turnover-limiting step. The kinetic isotope effect was determined to be 2.29(±0.15), which suggests that the breakage of the C-H bond is not the rate-limiting step, but that it is one of several slow steps in the catalytic cycle. Rapid scrambling of hydrogen and deuterium at the α position of the alcohol was observed with deuterium-labeled substrates, which implies that the catalytically active species is a ruthenium dihydride. The experimental results were supported by the characterization of a plausible catalytic cycle by using DFT/M06. Both cis-dihydride and trans-dihydride intermediates were considered, but when the theoretical turnover frequencies (TOFs) were derived directly from the calculated DFT/M06 energies, we found that only the trans-dihydride pathway was in agreement with the experimentally determined TOFs.