Search

In the era of big data, the issue of data quality has become increasingly prominent. One of the main challenges is the problem of duplicate data, which can arise from repeated entry or the merging of multiple data sources. These "dirty data" problems can significantly limit the effective application of big data. To address the issue of data deduplication, we propose a pre-trained deduplication model based on active learning, which is the first work that utilizes active learning to address the problem of deduplication at the semantic level. The model is built on a pre-trained Transformer and fine-tuned to solve the deduplication problem as a sequence to classification task, which firstly integrate the transformer with active learning into an end-to-end architecture to select the most valuable data for deduplication model training, and also firstly employ the R-Drop method to perform data augmentation on each round of labeled data, which can reduce the cost of manual labeling and improve the model's performance. Experimental results demonstrate that our proposed model outperforms previous state-of-the-art (SOTA) for deduplicated data identification, achieving up to a 28% improvement in Recall score on benchmark datasets.

Small Object Detection (SOD) is an important machine vision topic because (i) a variety of real-world applications require object detection for distant objects and (ii) SOD is a challenging task due to the noisy, blurred, and less-informative image appearances of small objects. This paper proposes a new SOD dataset consisting of 39,070 images including 137,121 bird instances, which is called the Small Object Detection for Spotting Birds (SOD4SB) dataset. The detail of the challenge with the SOD4SB dataset is introduced in this paper. In total, 223 participants joined this challenge. This paper briefly introduces the award-winning methods. The dataset, the baseline code, and the website for evaluation on the public testset are publicly available., Comment: This paper is included in the proceedings of the 18th International Conference on Machine Vision Applications (MVA2023). It will be officially published at a later date. Project page : https://www.mva-org.jp/mva2023/challenge

With the advent of the big data era, the data quality problem is becoming more critical. Among many factors, data with missing values is one primary issue, and thus developing effective imputation models is a key topic in the research community. Recently, a major research direction is to employ neural network models such as self-organizing mappings or automatic encoders for filling missing values. However, these classical methods can hardly discover interrelated features and common features simultaneously among data attributes. Especially, it is a very typical problem for classical autoencoders that they often learn invalid constant mappings, which dramatically hurts the filling performance. To solve the above-mentioned problems, we propose a missing-value-filling model based on a feature-fusion-enhanced autoencoder. We first incorporate into an autoencoder a hidden layer that consists of de-tracking neurons and radial basis function neurons, which can enhance the ability of learning interrelated features and common features. Besides, we develop a missing value filling strategy based on dynamic clustering that is incorporated into an iterative optimization process. This design can enhance the multi-dimensional feature fusion ability and thus improves the dynamic collaborative missing-value-filling performance. The effectiveness of the proposed model is validated by extensive experiments compared to a variety of baseline methods on thirteen data sets.

Ultrafast diffraction imaging is a powerful tool to retrieve the geometric structure of gas-phase molecules with combined picometre spatial and attosecond temporal resolution. However, structural retrieval becomes progressively difficult with increasing structural complexity, given that a global extremum must be found in a multi-dimensional solution space. Worse, pre-calculating many thousands of molecular configurations for all orientations becomes simply intractable. As a remedy, here, we propose a machine learning algorithm with a convolutional neural network which can be trained with a limited set of molecular configurations. We demonstrate structural retrieval of a complex and large molecule, Fenchone (C$_{10}$H$_{16}$O), from laser-induced electron diffraction (LIED) data without fitting algorithms or ab initio calculations. Retrieval of such a large molecular structure is not possible with other variants of LIED or ultrafast electron diffraction. Combining electron diffraction with machine learning presents new opportunities to image complex and larger molecules in static and time-resolved studies.

We report the design and implementation of multiple Tesla type micro valves in the target delivery system of a reaction microscope (ReMi) to study gas phase structural dynamics of complex polyatomic molecules, when no delivery system currently exists that can deliver dense enough molecular jets of neutral complex molecules without ionizing or exciting the target. We show, the Tesla valves provide an efficient unidirectional flow of the cis-stilbene molecules into the ReMi. We demonstrate using a bubbler with Tesla valves an order-of-magnitude increase in the detected stilbene molecular ion signal following the strong-field tunnel ionization (SFTI), compared to conventional bubbler without any Tesla valves. Our results for the first time, opens the door for studying large, complex neutral molecules in the gas-phase with low vapour pressures in future ultrafast studies.

Visualizing molecular transformations in real-time requires a structural retrieval method with {\AA}ngstr\"om spatial and femtosecond temporal atomic resolution. Imaging of hydrogen-containing molecules additionally requires an imaging method that is sensitive to the atomic positions of hydrogen nuclei, with most methods possessing relatively low sensitivity to hydrogen scattering. Laser-induced electron diffraction (LIED) is a table top technique that can image ultrafast structural changes of gas-phase polyatomic molecules with sub-{\AA}ngstr\"om and femtosecond spatiotemporal resolution together with relatively high sensitivity to hydrogen scattering. Here, we image the umbrella motion of an isolated ammonia molecule (NH$_3$) following its strong field ionization. Upon ionization of a neutral ammonia molecule, the ammonia cation (NH$_3^+$) undergoes an ultrafast geometrical transformation from a pyramidal ($\Phi_{HNH} = 107 ^\circ$) to planar ($\Phi_{HNH}=120^\circ$) structure in approximately 8 femtoseconds. Using LIED, we retrieve a near-planar ($\Phi_{HNH}=117 \pm 5^\circ$) field-dressed NH$_3^+$ molecular structure $7.8-9.8$ femtoseconds after ionization. Our measured field-dressed NH$_3^+$ structure is in excellent agreement with our calculated equilibrium field dressed structure using quantum chemical ab initio calculations.

Observing changes in molecular structure requires atomic-scale {\AA}ngstrom and femtosecond spatio-temporal resolution. We use the Fourier transform (FT) variant of laser-induced electron diffraction (LIED), FT-LIED, to directly retrieve the molecular structure of ${\rm H_2O^+}$ with picometre and femtosecond resolution without a priori knowledge of the molecular structure nor the use of retrieval algorithms or ab initio calculations. We identify a symmetrically stretched ${\rm H_2O^+}$ field-dressed structure that is most likely in the ground electronic state. We subsequently study the nuclear response of an isolated water molecule to an external laser field at four different field strengths. We show that upon increasing the laser field strength from 2.5 to 3.8 V/{\AA}, the O-H bond is further stretched and the molecule slightly bends. The observed ultrafast structural changes lead to an increase in the dipole moment of water and, in turn, a stronger dipole interaction between the nuclear framework of the molecule and the intense laser field. Our results provide important insights into the coupling of the nuclear framework to a laser field as the molecular geometry of ${\rm H_2O^+}$ is altered in the presence of an external field.