Original Counterparts Research Articles

Square-root Floquet topological phases and time crystals

Periodically driven (Floquet) phases are attractive due to their ability to host unique physical phenomena with no static counterparts. We propose a general approach in nontrivially devising a square-root version of existing Floquet phases, applicable both in noninteracting and in interacting setting. The resulting systems are found to yield richer physics that is otherwise absent in the original counterparts and is robust against parameter imperfection. These include the emergence of Floquet topological superconductors with arbitrarily many zero, $\ensuremath{\pi}$, and $\ensuremath{\pi}/2$ edge modes, as well as $4T$-period Floquet time crystals in disordered and disorder-free systems ($T$ being the driving period). Remarkably, our approach can be repeated indefinitely to obtain a ${2}^{n}\mathrm{th}$-root version of any periodically driven system, thus, allowing for the discovery and systematic construction of exotic Floquet phases.

Amplitude Reordering Accelerates the Adaptive Variational Quantum Eigensolver Algorithms

The variational quantum eigensolver (VQE) algorithm can simulate the chemical systems such as molecules in the noisy-intermediate-scale quantum devices and shows promising applications in quantum chemistry simulations. The accuracy and computational cost of the VQE simulations are determined by the underlying ansatz. Therefore, the most important issue is to generate a compact and accurate ansatz, which requires a shallower parametric quantum circuit and can achieve an acceptable accuracy. The newly developed adaptive algorithms (AAs) such as the adaptive derivative-assembled pseudo-Trotter VQE (ADAPT-VQE) can solve this issue via generating compact and accurate ansatzes. However, these AAs show very low computational efficiency because they require a large number of additional measurements. Here we propose an amplitude reordering (AR) strategy to accelerate the promising but expensive AAs by adding operators in a "batched" fashion in a way that their order is still quasi-optimal. We first introduce the AR method into ADAPT-VQE and build the AR-ADAPT-VQE algorithm. We then endow the energy-sorting VQE (ES-VQE) algorithm with the adaptive feature and introduce the AR into AES-VQE to form the AR-AES-VQE algorithm. To demonstrate the performance of these algorithms, we calculate the dissociation curves of three small molecules, LiH, linear BeH2, and linear H6, by using (AR-)ADAPT-VQE and (AR-)AES-VQE algorithms. It is found that all of the AR-equipped AAs (AR-AAs) can significantly reduce the number of iterations and subsequently accelerate the calculations with a speedup of up to more than ten times without the obvious loss of accuracy. The final ansatz generated by the AR-AAs not only avoids extra circuit depth but also maintains the computational accuracy; sometimes the AR-AAs even outperforms their original counterparts.

Cross-Interactive Residual Smoothing for Global and Block Lanczos-Type Solvers for Linear Systems with Multiple Right-Hand Sides

Global and block Krylov subspace methods are efficient iterative solvers for large sparse linear systems with multiple right-hand sides. However, global or block Lanczos-type solvers often exhibit large oscillations in the residual norms and may have a large residual gap relating to the loss of attainable accuracy of the approximations. Conventional residual smoothing schemes suppress these oscillations but cannot improve the attainable accuracy, whereas a recent residual smoothing scheme enables the improvement of the attainable accuracy for single right-hand side Lanczos-type solvers. The underlying concept of this scheme is that the primary and smoothed sequences of the approximations and residuals influence one another, thereby avoiding the severe propagation of rounding errors. In the present study, we extend this cross-interactive residual smoothing to the case of solving linear systems with multiple right-hand sides. The resulting smoothed methods can reduce the residual gap with a low additional cost compared to their original counterparts. We demonstrate the effectiveness of the proposed approach through rounding error analysis and numerical experiments.

Extending the GLM Framework of the Lee-Carter Model with Random Forest Recursive Feature Elimination Based Determinants of Mortality

The Lee-Carter (LC) model led to the development of many prominent mortality models. This study aims to modify the generalised linear model (GLM) (Poisson, negative binomial, and binomial) framework of the LC model by incorporating factors that affect mortality into the model. The top three factors which affect the mortality for each of the 14 countries studied were selected using the random forest recursive feature elimination (RF-RFE) method which eliminates the least important factors based on the correlation of the predictors with the log-mortality rate. These selected factors were integrated in the form of additional bilinear variates to the GLM models and compared to their original counterparts. The RF-RFE method is effective in selecting the best determinants of mortality by avoiding multicollinearity among predictor variables. The inclusion of the time-factor modulation based on the factors selected improved the model adequacy significantly. Vast improvement was evident in the Poisson and binomial settings. Furthermore, the modified GLM version fits short-base-period data well. This study shows that the inclusion of exogenous determinants of mortality improves the performance of the model significantly.

Adversarial Training for Improving Model Robustness? Look at Both Prediction and Interpretation

Neural language models show vulnerability to adversarial examples which are semantically similar to their original counterparts with a few words replaced by their synonyms. A common way to improve model robustness is adversarial training which follows two steps—collecting adversarial examples by attacking a target model, and fine-tuning the model on the augmented dataset with these adversarial examples. The objective of traditional adversarial training is to make a model produce the same correct predictions on an original/adversarial example pair. However, the consistency between model decision-makings on two similar texts is ignored. We argue that a robust model should behave consistently on original/adversarial example pairs, that is making the same predictions (what) based on the same reasons (how) which can be reflected by consistent interpretations. In this work, we propose a novel feature-level adversarial training method named FLAT. FLAT aims at improving model robustness in terms of both predictions and interpretations. FLAT incorporates variational word masks in neural networks to learn global word importance and play as a bottleneck teaching the model to make predictions based on important words. FLAT explicitly shoots at the vulnerability problem caused by the mismatch between model understandings on the replaced words and their synonyms in original/adversarial example pairs by regularizing the corresponding global word importance scores. Experiments show the effectiveness of FLAT in improving the robustness with respect to both predictions and interpretations of four neural network models (LSTM, CNN, BERT, and DeBERTa) to two adversarial attacks on four text classification tasks. The models trained via FLAT also show better robustness than baseline models on unforeseen adversarial examples across different attacks.

What Can We Learn Even from the Weakest? Learning Sketches for Programmatic Strategies

In this paper we show that behavioral cloning can be used to learn effective sketches of programmatic strategies. We show that even the sketches learned by cloning the behavior of weak players can help the synthesis of programmatic strategies. This is because even weak players can provide helpful information, e.g., that a player must choose an action in their turn of the game. If behavioral cloning is not employed, the synthesizer needs to learn even the most basic information by playing the game, which can be computationally expensive. We demonstrate empirically the advantages of our sketch-learning approach with simulated annealing and UCT synthesizers. We evaluate our synthesizers in the games of Can't Stop and MicroRTS. The sketch-based synthesizers are able to learn stronger programmatic strategies than their original counterparts. Our synthesizers generate strategies of Can't Stop that defeat a traditional programmatic strategy for the game. They also synthesize strategies that defeat the best performing method from the latest MicroRTS competition.

Evaluation of Deep Learning and Conventional Approaches for Image Recaptured Detection in Multimedia Forensics

Image recaptured from a high-resolution LED screen or a good quality printer is difficult to distinguish from its original counterpart. The forensic community paid less attention to this type of forgery than to other image alterations such as splicing, copy-move, removal, or image retouching. It is significant to develop secure and automatic techniques to distinguish real and recaptured images without prior knowledge. Image manipulation traces can be hidden using recaptured images. For this reason, being able to detect recapture images becomes a hot research topic for a forensic analyst. The attacker can recapture the manipulated images to fool image forensic system. As far as we know, there is no prior research that has examined the pros and cons of up-to-date image recaptured techniques. The main objective of this survey was to succinctly review the recent outcomes in the field of image recaptured detection and investigated the limitations in existing approaches and datasets. The outcome of this study provides several promising directions for further significant research on image recaptured detection. Finally, some of the challenges in the existing datasets and numerous promising directions on recaptured image detection are proposed to demonstrate how these difficulties might be carried into promising directions for future research. We also discussed the existing image recaptured datasets, their limitations, and dataset collection challenges.

Abstract LB563: Integrated multiomic profiling identifies the epigenetic regulator PRC2 as a therapeutic target to counteract leukemia immune escape and relapse

Abstract Background: Evasion from immune control represents one of the main drivers of acute myeloid leukemia (AML) relapse after allogeneic hematopoietic cell transplantation (allo-HCT). In particular, up to 40% of AML relapses display complete loss of surface expression of HLA class II molecules without any genetic lesion explaining this phenotype (Christopher et al, N Engl J Med, 2018; Toffalori et al, Nat Med, 2019). This led us to investigate the links between epigenetic changes, immune evasion and post-transplantation relapse. Methods: Starting from primary AML samples pairwise collected from five patients at diagnosis and relapse with non-genomic loss of HLA class II expression, we generated Patients-Derived Xenografts (PDXs) into NOD-SCID γ-chain null mice. Leukemic cells expanded in the mice and their original human counterparts were characterized for changes in gene expression (by RNA-Seq), DNA methylation profile (by RRBS), and chromatin accessibility (by ATAC-Seq). The results obtained by all these approaches and in the different patients were integrated by Multi-Omics Factor Analysis (MOFA) and Gene Set Enrichment Analysis (GSEA). Finally, we tested the immunological effects of epigenetic drugs on AML cells and on their recognition by T cells in ex-vivo short-term cultures and in PDXs. Results: We verified that PDXs faithfully recapitulate immune-related differences between AML diagnosis and post-transplantation relapse, including loss of expression of HLA class II molecules. Differences between diagnosis and post-transplantation relapse samples were mostly explained by changes in chromatin accessibility, and largely unrelated to the DNA methylation profile. In particular, in all five patients analyzed, we documented genomewide chromatin compaction at time of relapse, that was particularly evident for HLA class II genes and their master regulator CIITA, and not detected in relapses after sole chemotherapy. Integration of all the high-throughput technologies by MOFA, and of results from different patients by GSEA, pointed to the Polycomb Repressive Complex 2 (PRC2) as the main candidate mediator of HLA class II silencing.Pharmacological inhibition of PRC2 subunits rescued HLA class II expression in AML relapses ex vivo and in vivo, with consequent recovery of leukemia recognition by CD4+ T cells. Conclusions: Our results uncover a novel link between epigenetics and leukemia immune escape, which may rapidly translate into innovative strategies to cure or prevent AML post-transplantation relapse. Citation Format: Valentina Gambacorta, Stefano Beretta, Martina Ciccimarra, Laura Zito, Kety Giannetti, Angela Andrisani, Daniela Gnani, Lucia Zanotti, Giacomo Oliveira, Matteo G. Carrabba, Davide Cittaro, Ivan Merelli, Fabio Ciceri, Raffaella Di Micco, Luca Vago. Integrated multiomic profiling identifies the epigenetic regulator PRC2 as a therapeutic target to counteract leukemia immune escape and relapse [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr LB563.

Structural Response and Residual Capacity of S700 High-Strength Steel CHS Columns after Exposure to Elevated Temperatures

The structural behavior and residual capacity of S700 high-strength steel circular hollow section (CHS) columns after exposure to elevated temperatures were studied through testing and numerical modeling. The testing program was conducted on 10 S700 high-strength steel CHS columns and comprised heating and cooling of the specimens as well as postfire material testing, measurements of initial geometric imperfections, and pin-ended column tests. Numerical simulations were subsequently performed, where finite element models were built and validated with reference to the experimental results and afterward used to conduct parametric studies to obtain further numerical data. Given that there are no specific provisions for the design of steel structures after exposure to elevated temperatures, the relevant room-temperature design buckling curves were evaluated, using postfire material properties, for their applicability to S700 high-strength steel CHS columns after exposure to elevated temperatures, based on the experimental and numerical data. The evaluation results indicated that the buckling curves prescribed in the American specification and Australian standard lead to accurate and consistent residual capacity predictions for S700 high-strength steel CHS columns after exposure to elevated temperatures, while the Eurocode buckling curve yields conservative predictions of residual capacity. A revised Eurocode buckling curve was then proposed, and resulted in a higher level of accuracy than its original counterpart.

Overlapping biosimilar and originator follitropin alfa preparations: How much closer can they get?

Unfounded skepticism relating to biosimilars, arising from the assertion that they are not molecularly identical to their original counterpart, fails to acknowledge that no biological medicine, including Gonal-f® (from Merck Serono) is identical to itself. Molecular differences between the biosimilar and the reference medicines are irrelevant and clinically undetectable as long as they are contained within the accepted variability for the original medicine. Accordingly, the minor differences in ‘ongoing pregnancy rate’ and ‘live birth’ rate reported in a recent meta-analysis of biosimilars of Gonal-f® from Chua et al. are probably driven by product-unrelated factors, notwithstanding the fact that of the four products under analysis, only Ovaleap® (from Theramex UK Ltd) and Bemfola® (from Gedeon Richter Plc) can unambiguously be considered to be biosimilars. The EU Biosimilars model has proven successful, but some healthcare professionals, building on highly arguable premises, voice a distrust in biosimilars. Only if such scientifically unfounded distrust is reverted, the full promise of rFSH alfa biosimilars for reproductive medicine patients is likely to be fulfilled.

Italian Translation, Adaptation, and Validation of the Novel Satisfaction Measure Assessment after Primary Total Joint Arthroplasty: The Goodman Score Questionnaire.

Patient satisfaction after total joint arthroplasties (TJA) represents a key element for the evaluation of surgery success in relation to subjects’ needs and expectations. The assessment tools are applied inconsistently throughout the literature, and thus, it is difficult to compare results among different studies. Goodman et al. proposed a standardized questionnaire with strong psychometric properties for the assessment of satisfaction. The present study aims to translate, adapt, and validate the Goodman questionnaire for the Italian population. After translation and back translation, the questionnaire was administrated to 50 patients. Internal consistency, test–retest reliability, floor and ceiling effects, and construct validity were evaluated (correlation with KOOS/HOOS, SF-12 PCS/MCS, EQ-5D). Responsiveness was evaluated with respect to SF-12 PCS improvements. The Italian version of the Goodman score questionnaire demonstrated psychometric properties similar to those of the original version. The translated questionnaire showed good internal consistency (Cronbach’s alpha = 0.836) and test–retest reliability (ICC: 0.507). Moderate/strong correlations were observed between the Italian version of the Goodman score and other scores. The score significantly discriminated patients who improved from those who did not improve in SF-12 PCS after treatment. This study provides an adapted and validated Italian version of the Goodman score questionnaire, with psychometric properties similar to those of its original counterpart.

A landscape-based analysis of fixed temperature and simulated annealing

Since the introduction of Simulated Annealing (SA), researchers have considered variants that keep the same temperature value throughout the whole search and tried to determine whether this strategy can be more effective than the original cooling scheme. Several studied have tried to answer this question without a conclusive answer and without providing indications that could be useful for a practical implementation. In this work, we address this question following an experimental approach, relating the characteristics of the algorithms with the characteristics of the landscapes they encounter. We use problem-independent landscape features to study the algorithmic behaviour across different problems. We consider three different objective functions and various instance classes and determine the conditions under which the fixed-temperature variant of SA can outperform its original counterpart and when SA is instead a better choice.

An Evaluation of Phylogenetic Workflows in Viral Molecular Epidemiology.

The use of viral sequence data to inform public health intervention has become increasingly common in the realm of epidemiology. Such methods typically utilize multiple sequence alignments and phylogenies estimated from the sequence data. Like all estimation techniques, they are error prone, yet the impacts of such imperfections on downstream epidemiological inferences are poorly understood. To address this, we executed multiple commonly used viral phylogenetic analysis workflows on simulated viral sequence data, modeling Human Immunodeficiency Virus (HIV), Hepatitis C Virus (HCV), and Ebolavirus, and we computed multiple methods of accuracy, motivated by transmission-clustering techniques. For multiple sequence alignment, MAFFT consistently outperformed MUSCLE and Clustal Omega, in both accuracy and runtime. For phylogenetic inference, FastTree 2, IQ-TREE, RAxML-NG, and PhyML had similar topological accuracies, but branch lengths and pairwise distances were consistently most accurate in phylogenies inferred by RAxML-NG. However, FastTree 2 was the fastest, by orders of magnitude, and when the other tools were used to optimize branch lengths along a fixed FastTree 2 topology, the resulting phylogenies had accuracies that were indistinguishable from their original counterparts, but with a fraction of the runtime.

Nanosecond pulsed laser-assisted modified copper surface structure: Enhanced surface microhardness and microbial corrosion resistance

Microbiologically influenced corrosion (MIC) is an unavoidable problem in several industries. Copper (Cu) and its alloys are widely used engineering materials. However, MIC of Cu remains a persistent challenge to their performance and functional lifetime under aggressive environments. This study investigated nanosecond pulsed laser processing (LP), which may enhance the corrosion resistance of Cu. The microstructural evolution and corrosion behavior of LP-Cu in the presence of sulfate-reducing bacteria (SRB) were evaluated. Typical deformation-induced microstructural features of high-density dislocations were analyzed on the top surface of LP-Cu coupon. Electrochemical measurements suggested that LP-Cu coupons exhibited better corrosion resistance in SRB-inoculated solution compared with their original counterpart. The enhanced corrosion resistance by LP primarily resulted from the combined influences of compressive residual stress and work hardening in the surface. However, overlap percentage played a key role in improving corrosion resistance. LP produced optimal corrosion resistance at 50% overlap. Therefore, this study introduces a unique and an option for anticorrosion control in manufacturing processes and potentially implements it onto other materials to improve its microbial corrosion resistance through LP.

A Proposal upon the Figure of Hermes as an Ancient God of Fire (According to the Homeric Hymn to Hermes)

The origin and meaning of Hermes, whose figure is full of different and even contradictory aspects, has never been clarified. Now, starting from a passage from the Homeric Hymn dedicated to him – in which Hermes is considered the one “who first invented fire-sticks and fire” – as well as on the fact that he is often associated with Hestia, the goddess of the hearth, we propose here that he was originally a god of fire. This immediately explains his attributions, the meaning of his name and even the shape of the caduceus. Moreover, the Hermes-Hestia couple corresponds to the conception of the two fires in the Vedic world: one is the quadrangular “male fire” of the sky, while the other is the round “female fire” of the earth. This original dimension of Hermes as a god of fire also links him to lightning, that produces forest fires and is considered a divine omen, which explains his function as the messenger of the gods. At this point, it is even possible to identify the original counterpart of Hermes in the Roman world: he was not Mercury, but the ancient god Terminus, whose original dimension linked to fire gradually faded over the centuries, as it happened also for Hermes. Keywords: Hermes, Hestia, Homeric Hymns, Terminus

Compressed Deep Learning to Classify Arrhythmia in an Embedded Wearable Device.

The importance of an embedded wearable device with automatic detection and alarming cannot be overstated, given that 15–30% of patients with atrial fibrillation are reported to be asymptomatic. These asymptomatic patients do not seek medical care, hence traditional diagnostic tools including Holter are not effective for the further prevention of associated stroke or heart failure. This is likely to be more so in the era of COVID-19, in which patients become more reluctant on hospitalization and checkups. However, little literature is available on this important topic. For this reason, this study developed efficient deep learning with model compression, which is designed to use ECG data and classify arrhythmia in an embedded wearable device. ECG-signal data came from Korea University Anam Hospital in Seoul, Korea, with 28,308 unique patients (15,412 normal and 12,896 arrhythmia). Resnets and Mobilenets with model compression (TensorFlow Lite) were applied and compared for the diagnosis of arrhythmia in an embedded wearable device. The weight size of the compressed model registered a remarkable decrease from 743 MB to 76 KB (1/10000), whereas its performance was almost the same as its original counterpart. Resnet and Mobilenet were similar in terms of accuracy, i.e., Resnet-50 Hz (97.3) vs. Mo-bilenet-50 Hz (97.2), Resnet-100 Hz (98.2) vs. Mobilenet-100 Hz (97.9). Here, 50 Hz/100 Hz denotes the down-sampling rate. However, Resnets took more flash memory and longer inference time than did Mobilenets. In conclusion, Mobilenet would be a more efficient model than Resnet to classify arrhythmia in an embedded wearable device.

Abstract P5-01-01: Intraductal patient-derived xenograft models of ductal carcinoma in situ (DCIS) to distinguish indolent DCIS from aggressive DCIS

Abstract Ductal Carcinoma in Situ (DCIS) is a starting lesion in the milk duct of the breast, which accounts for 25% of all ‘breast cancers’ detected. DCIS is usually treated by surgery combined with radiotherapy, which can have a large impact on the life of patients. However, there is little to no evidence that treatment of low and intermediate grade DCIS reduces mortality, while women diagnosed with DCIS do perceive their risk of dying the same as patients with invasive disease. To reduce the negative perception and the overtreatment of DCIS, but assure proper treatment for high risk DCIS, it is crucial to understand the biology underlying DCIS. In order to study the biology of DCIS we established patient-derived xenograft (PDX) models by intraductally injecting fresh patient DCIS material into the mammary ducts of female immunocompromised mice (Mouse Mammary Intraductal(MIND)). To date, 150 primary DCIS patient samples were engrafted and >100 have been incubated in vivo for a period of 12 months. Using this method we obtain a take rate of 89% with 36% of our models showing progression. Histology and molecular subtyping by PAM50 classification are well preserved in the outgrown PDX lesions compared to the original human counterpart, ensuring that our models are a good representation of the patient disease. For >80 primary samples we obtained RNAseq as well as >50 matched PDX profiles. In addition WES or WGS data is generated from the same primary DCIS samples together with 10 matched PDX WES profiles as well as >40 matched CNV PDX profiles. On top of this we used whole mount imaging, in order to create 3D sections of the injected mammary glands extracted from our PDX models, showing two distinct growth patterns correlated with invasion. And as this is all done in the context of the PRECISION consortium this allows us to confirm and validate our findings in larger sequencing and imaging efforts of human samples. Besides collecting large amounts of data which we can now link to DCIS progression we have successfully sequentially transplanted >25 of our PDX models in order to create models that can be shared with the scientific community and can be manipulated to validate the role of genes in DCIS progression. As an example we took HER2 overexpression, which in our models was found to be associated to progressive disease. On the one hand we treated three progressive HER2 overexpressing DCIS PDX models with a HER2-antibody Herceptin showing a massive decline in growth, while on the other hand we infected two indolent HER2 negative DCIS PDX models with a HER2-GFP lentivirus showing a transformation to HER2 positive progressive growth. In conclusion all this data together enables us to create a well-characterized biobank of DCIS models of which we can follow the progression, sequentially transplant >25 models, find mutational and expression profiles related to high risk DCIS and manipulate gene expression to validate the role of genes in DCIS progression. Citation Format: Jelle Wesseling, Stefan Hutten, Catrin Lutz, Colinda Scheele, Madelon Badoux, Timo Eijkman, Roebi de Bruijn, Mathilde Almekinders, Joyce Sanders, Esther Lips, Jacco van Rheenen, Jos Jonkers, Grand challenge PRECISION consortium. Intraductal patient-derived xenograft models of ductal carcinoma in situ (DCIS) to distinguish indolent DCIS from aggressive DCIS [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P5-01-01.

Likelihood ascent search‐aided low complexity improved performance massive MIMO detection in perfect and imperfect channel state information

SummaryMassive multiple‐input multiple‐output (MIMO) systems improve spectral efficiency and link reliability. Linear minimum mean‐squared error (MMSE) detectors can achieve optimal performance in massive MIMO detection but require large dimension matrix inversion, which is computationally intensive. Therefore, low complexity iterative detection schemes are proposed in the literature as an alternative to the exact MMSE method. However, the performance of these schemes is greatly influenced by the choice of the initial solution. Therefore, to improve the detection performance in this paper, we proposed three hybrid detection schemes, which are Newton–Schultz–Richardson (NS‐RI), Newton–Schultz–Chebyshev (NS‐Cheby), and Newton–Schultz–Gauss–Seidel (NS‐GS). The proposed hybrid schemes show significant performance improvement and a higher convergence rate compared to their original counterpart. The performance of the proposed detectors is further improved by the likelihood ascent search (LAS) stage, which corrects the detected symbols obtained from iterative MMSE methods through a neighborhood search. However, the complexity of the LAS algorithm primarily depends on the initialization step. In this work, we introduce an efficient Gram matrix computation in the real domain. Additionally, we have applied a band approximation of the Gram matrix for the LAS initialization, which reduces the order of computational complexity of the Gram matrix from to O(ωNTNR) where ω < <2NT.

Enhanced initial biodegradation resistance of the biomedical Mg-Cu alloy by surface nanomodification

Mg-Cu alloys are promising antibacterial implant materials. However, their clinical applications have been impeded by their high initial biodegradation rate, which can be alleviated using nanotechnology by for example surface nanomodification to obtain a gradient nanostructured surface layer. The present work (i) produced a gradient nanostructured surface layer with a ∼500 µm thickness on a Mg-0.2 Cu alloy by a surface mechanical grinding treatment (SMGT), and (ii) studied the biodegradation behavior in Hank's solution. The initial biodegradation rate of the SMGTed samples was significantly lower than that of the unSMGTed original counterparts, which was attributed to the surface nanocrystallization, and the fragmentation and re-dissolution of Mg2Cu particles in the surface of the SMGTed Mg-0.2 Cu alloy. Furthermore, the SMGTed Mg-0.2 Cu alloy had good antibacterial efficacy. This work creatively used SMGT technology to produce a high-performance Mg alloy implant material.

Modifying the softening process for knowledge distillation

The prime focus of knowledge distillation (KD) seeks a light proxy termed student to mimic the outputs of its heavy neural networks termed teacher, and makes the student run real-time on the resource-limited devices. This paradigm requires aligning the soft logits of both teacher and student. However, few doubts whether the process of softening the logits truly give full play to the teacher-student paradigm. In this paper, we launch several analyses to delve into this issue from scratch. Subsequently, several simple yet effective functions are devised to replace the vanilla KD. The ultimate function can be an effective alternative to its original counterparts and work well with other skills like FitNets. To claim this point, we conduct several visual tasks on individual benchmarks, and experimental results verify the potential of our proposed function in terms of performance gains. For example, when the teacher and student networks are ShuffleNetV2-1.0 and ShuffleNetV2-0.5, our proposed method achieves 40.88%top-1 error rate on Tiny ImageNet.