Small Input Research Articles

Orthogonal inducible control of Cas13 circuits enables programmable RNA regulation in mammalian cells.

RNA plays an indispensable role in mammalian cell functions. Cas13, a class of RNA-guided ribonuclease, is a flexible tool for modifying and regulating coding and non-coding RNAs, with enormous potential for creating new cell functions. However, the lack of control over Cas13 activity has limited its cell engineering capability. Here, we present the CRISTAL (Control of RNA with Inducible SpliT CAs13 Orthologs and Exogenous Ligands) platform. CRISTAL is powered by a collection (10 total) of orthogonal split inducible Cas13 effectors that can be turned ON or OFF via small molecules in multiple cell types, providing precise temporal control. Also, we engineer Cas13 logic circuits that can respond to endogenous signaling and exogenous small molecule inputs. Furthermore, the orthogonality, low leakiness, and high dynamic range of our inducible Cas13d and Cas13b enable the design and construction of a robust incoherent feedforward loop, leading to near-perfect and tunable adaptation response. Finally, using our inducible Cas13 effectors, we achieve simultaneous multiplexed control of multiple genes in vitro and in mice. Together, our CRISTAL design represents a powerful platform for precisely regulating RNA dynamics to advance cell engineering and elucidate RNA biology.

RC-Effects on the Oxide of SOI MOSFET under Off-State TDDB Degradation: RF Characterization and Modeling.

Based on S-parameter measurements, the effect of dynamic trapping and de-trapping of charge in the gate oxide, the increase of dielectric loss due to polarization, and the impact of leakage current on the small-signal input impedance at RF is analyzed and represented. This is achieved by systematically extracting the corresponding model parameters from single device measurements at different frequency ranges, and then the methodology is applied to analyze the evolution of these parameters when the device is submitted to non-conducting electrical stress. This approach not only allows to inspect the impact of effects not occurring under DC conditions, such as the current due to the time varying dielectric polarization, but also to clearly distinguish effects in accordance with the functional form of their contribution to the device's impedance. In fact, it is shown that minor changes in the model of the gate capacitance by including additional resistive and capacitive components allows for an excellent model-experiment correlation up to 30 GHz. Moreover, the accuracy of the correlation is shown to be maintained when applying the proposal to the device under different gate-to-source bias conditions and at several stages during off-state degradation.

Executives' perceptions of the steel industry's materials management system effectiveness

Materials management is a key component of supply chain management in public sector enterprises, especially in the steel industry. It entails managing the flow of materials while concurrently evaluating factors including demand, price, availability, quality, and delivery dates. This focuses on the necessity of thorough planning for effective material management and control. Small material input cost reductions can lower manufacturing costs and increase profits. The study aims to analyse the executives' perception towards the dimensions of materials management efficacy and also examines respondents’ demographic status. The results reveal that there is a significant association between demographic variables designation, education and experience with the dimensions quality, convenience, and cost while there does not exist a significant association with the dimensions time, quantity and human resource. The study develops a better understanding of the factors determining the efficacy of materials management systems in steel sector enterprises.

Response to photic stimulation as a measure of cortical excitability in epilepsy patients.

Studying states and state transitions in the brain is challenging due to nonlinear, complex dynamics. In this research, we analyze the brain's response to non-invasive perturbations. Perturbation techniques offer a powerful method for studying complex dynamics, though their translation to human brain data is under-explored. This method involves applying small inputs, in this case via photic stimulation, to a system and measuring its response. Sensitivity to perturbations can forewarn a state transition. Therefore, biomarkers of the brain's perturbation response or "cortical excitability" could be used to indicate seizure transitions. However, perturbing the brain often involves invasive intracranial surgeries or expensive equipment such as transcranial magnetic stimulation (TMS) which is only accessible to a minority of patient groups, or animal model studies. Photic stimulation is a widely used diagnostic technique in epilepsy that can be used as a non-invasive perturbation paradigm to probe brain dynamics during routine electroencephalography (EEG) studies in humans. This involves changing the frequency of strobing light, sometimes triggering a photo-paroxysmal response (PPR), which is an electrographic event that can be studied as a state transition to a seizure state. We investigate alterations in the response to these perturbations in patients with genetic generalized epilepsy (GGE), with (n = 10) and without (n = 10) PPR, and patients with psychogenic non-epileptic seizures (PNES; n = 10), compared to resting controls (n = 10). Metrics of EEG time-series data were evaluated as biomarkers of the perturbation response including variance, autocorrelation, and phase-based synchrony measures. We observed considerable differences in all group biomarker distributions during stimulation compared to controls. In particular, variance and autocorrelation demonstrated greater changes in epochs close to PPR transitions compared to earlier stimulation epochs. Comparison of PPR and spontaneous seizure morphology found them indistinguishable, suggesting PPR is a valid proxy for seizure dynamics. Also, as expected, posterior channels demonstrated the greatest change in synchrony measures, possibly reflecting underlying PPR pathophysiologic mechanisms. We clearly demonstrate observable changes at a group level in cortical excitability in epilepsy patients as a response to perturbation in EEG data. Our work re-frames photic stimulation as a non-invasive perturbation paradigm capable of inducing measurable changes to brain dynamics.

Multi scale-aware attention for pyramid convolution network on finger vein recognition

In recent years, biometrics has been the most popular style of personal identification. The finger vein is an intrinsic and stable trait, and with the ability to detect liveness, it receives academic and industry attention. However, convolution neural networks (CNNs) based finger vein recognition generally can only cover a small input region by using small kernels. Hence, the performance is poor, facing low-quality finger vein images. It is a challenge to effectively use the critical feature of multi-scale for finger veins. In this article, we extract multi-scale features via pyramid convolution. We propose scale attention, namely, the scale-aware attention (SA) module, which enables dynamic adjustment of the weight of each scale to information aggregation. Utilize the complementation of different scale detail features to enhance the discriminativeness of extracted features, thus improving the finger vein recognition performance. In order to verify the present method’s efficiency, we carried out experiments on two public data sets and one internal data, and the wide range of experimental results proves the proposed method’s effectiveness.

Evaluation of Malware Classification Models for Heterogeneous Data.

Machine learning (ML) has found widespread application in various domains. Additionally, ML-based techniques have been employed to address security issues in technology, with numerous studies showcasing their potential and effectiveness in tackling security problems. Over the years, ML methods for identifying malicious software have been developed across various security domains. However, recent research has highlighted the susceptibility of ML models to small input perturbations, known as adversarial examples, which can significantly alter model predictions. While prior studies on adversarial examples primarily focused on ML models for image processing, they have progressively extended to other applications, including security. Interestingly, adversarial attacks have proven to be particularly effective in the realm of malware classification. This study aims to explore the transparency of malware classification and develop an explanation method for malware classifiers. The challenge at hand is more complex than those associated with explainable AI for homogeneous data due to the intricate data structure of malware compared to traditional image datasets. The research revealed that existing explanations fall short in interpreting heterogeneous data. Our employed methods demonstrated that current malware detectors, despite high classification accuracy, may provide a misleading sense of security and measuring classification accuracy is insufficient for validating detectors.

HEP@home: A Volunteer Computing Project to Run Fast Simulation with Delphes for CEPC

Delphes is a C++ framework to perform a fast multipurpose detector response simulation. The Circular Electron Positron Collider (CEPC) experiment runs fast simulation with a modified Delphes based on its own scientific objectives. The CEPC fast simulation with Delphes is a High Throughput Computing (HTC) application with small input and output files. Besides, to compile and run Delphes, only ROOT software is necessary.Therefore, all these features make it appropriate to run CEPC fast simulation as a Volunteer Computing application. As a result, a BOINC project named HEP@home is developed to run fast simulation with Delphes for CEPC. This paper describes the internal structure of the project, pre and post data operations, and its development status.

CuXCMP: CUDA-Accelerated Private Comparison Based on Homomorphic Encryption

Private comparison schemes constructed on homomorphic encryption offer the noninteractive and parallelizable features, and have advantages in communication bandwidth and performance. In this work, we propose cuXCMP, an extension of the privacy comparison scheme XCMP (AsiaCCS 2018). We address the relatively small input domain and the incompletely expressible output of XCMP, by modifying the encoding method and devising a constant term extraction (CTX) approach. Then, we describe a method for constructing privacy-preserving decision tree (PPDT) using this scheme. Considering the high computational overhead of CTX, we exploit the massive parallelism of the GPU to accelerate this function. Based on the results of the kernel profiling, we utilize several optimization techniques to improve the performance, including using multiple CUDA streams, reducing the grid dimension, kernel fusion, etc. By accelerating this function, we boost the execution time of the scheme and demonstrate 130× and 1.9× speedups for CTX and cuXCMP, respectively, as well as a 35% reduction in the evaluation time of PPDT.

Research on High-Performance Fourier Transform Algorithms Based on the NPU

Backpack computers require powerful, intelligent computing capabilities for field wearables while taking energy consumption into careful consideration. A recommended solution for this demand is the CPU + NPU-based SoC. In many wearable intelligence applications, the Fourier Transform is an essential, computationally intensive preprocessing task. However, due to the unique structure of the NPU, the conventional Fourier Transform algorithms cannot be applied directly to it. This paper proposes two NPU-accelerated Fourier Transform algorithms that leverage the unique hardware structure of the NPU and provides three implementations of those algorithms, namely MM-2DFT, MV-2FFTm, and MV-2FFTv. Then, we benchmarked the speed and energy efficiency of our algorithms for the gray image edge filtering task on the Huawei Atlas200I-DK-A2 development kits against the Cooley-Tukey algorithm running on CPU and GPU platforms. The experiment results reveal MM-2DFT outperforms OpenCL-based FFT on NVIDIA Tegra X2 GPU for small input sizes, with a 4- to 8-time speedup. As the input image resolution exceeds 2048, MV-2FFTv approaches GPU computation speed. Additionally, two scenarios were tested and analyzed for energy efficiency, revealing that cube units of the NPU are more energy efficient. The vector and CPU units are better suited for sparse matrix multiplication and small-scale inputs, respectively.

On Boundary Controllability for the Higher-Order Nonlinear Schrödinger Equation

A control problem with final overdetermination is considered for the higher-order nonlinear Schrödinger equation on a bounded interval. The boundary condition on the space derivative is chosen as the control. Results on the global existence of solutions under small input data are established.

Nonlinear Vibration Analysis of Bistable Morphing Composite Laminated Shell Based on High-Dimensional Dynamical Model

The bistable morphing composite laminated shell presents promising potential for aerospace morphing structures design, as it is capable of shape-changing under small energy inputs and maintaining two stable states without external loading. The bistable nonlinearity is distinct from the typical geometrical nonlinearity, as it arises from thermal residual stresses that cannot be equalized. During the derivation of a three-degree-of-freedom nonlinear dynamic model, Sanders’ nonlinear strain is incorporated into the strain–displacement relationship, while thermal residual stresses are taken into account in the constitutive relationship. The nonlinear dynamic model effectively captures the snap-through phenomena with orthogonal asymmetry in bistable morphing composite laminated shells. In order to reveal the resonance response and chaotic characteristics due to the interaction of vibrations between degrees of freedom, high-dimensional nonlinear dynamical model is derived by multiple scales method based on the 1:1:1 internal resonance relationship. The resonance response curves reflect the parametric influence law of the dynamical system stability. The nonlinear dynamic behaviors are described by the bifurcation diagrams and the maximum Lyapunov exponent. The significant motions under particular excitation conditions are visualized by phase portraits, time histories and Poincaré maps.

SOx Functionalized NiOOH Nanosheets Embedded in Ni(OH)2 Microarray for High-Efficiency Seawater Oxidation.

A nano-micro heterostructure has been established to address the challenges of selectivity, stress, pitting corrosion, and long-term durability of anodes in unpurified seawater. The heterostructure comprisedNiOOH nanosheets embedded within a high surface area Ni(OH)2 microarray, and the surface structure is further functionalized with sulfate (SOx). This cation-selective protective layer impedes chloride (Cl-) diffusion and abstracts H from reaction intermediates, leading to enhanced selectivity and corrosion resistance of the anode. The multilevel porosity within the randomly oriented nanosheets and the underlying support provide short diffusion channels for ions and mass migration, ensuring efficient ion transport and long-term structural and mechanical durability of the active sites, even at high current density. Remarkably, the catalyst requires a small input voltage of 400mV to deliver a current density of 1 Acm-2 and maintains it for over 168 h without noticeable degradation or hypochlorite formation. Spectroscopic analysis and density functional theory (DFT) calculations reveal that the Ni electronic structure in the +3 valence state, its strong structural interaction with the underlying microarray, and the functionality of SOx significantly reduce the required potential for O-O coupling.

Mechanical forces quench frontal polymerization: Experiments and theory

Frontal polymerization is a promising energy-saving method for rapid fabrication of polymer components with good mechanical properties. In these systems, a small energy input is sufficient to convert monomers, from a liquid or soft solid state, into a stiff polymer component. Once the reaction is initiated, it propagates as a self-sustaining front that is driven by the heat released from the reaction itself. While several studies have been proposed to capture the coupling between thermodynamics and extreme chemical kinetics in these systems, and can explain experimentally observed thermo-chemical instabilities, only few have considered the potential influence of mechanical forces that develop in these systems during fabrication. Nonetheless, some experiments do indicate that local volume changes induced by the competing effects of thermal expansion and chemical shrinkage, can lead to significant deformation or even failure in the resulting component. In this work, we present a unique experimental approach to elucidate the effect of mechanics on the propagation. Our experiments reveal that residual stresses that arise in frontal polymerization are not only a potential cause of undesired deformations in polymer products, but can also quench the reaction front. This thermo-chemo-mechanically coupled effect is captured by our theoretical model, which explains the mechanical limitations on frontal polymerization and can guide future fabrication. Overall, the findings of this work suggest that mechanical coupling needs to be taken into consideration to enable industrial applications of frontal polymerization at large scales.

Identifying Stress Granules As Determinants of Leukemia Stem Cell Maintenance and Stress Adaptation

The development of novel leukemic stem cell (LSC)-targeted therapeutics, which spare hematopoietic stem cells (HSC), is an urgent goal towards reducing the high, LSC-driven, relapse rates in acute myeloid leukemia (AML) patients. Towards elucidating key post-transcriptional networks underlying LSC maintenance we previously performed an in vivo pooled CRISPR dropout screen targeting 128 RNA binding proteins (RBPs) highly expressed in AML LSCs while having low expression in HSCs. This screen highlighted 12 RBPs crucial for murine AML LSC-mediated leukemic growth, of which there was a significant enrichment of stress granule (SG) RBPs. Further, gene set enrichment analysis shows the SG proteome is enriched in human AML LSCs vs non-LSCs while normal HSCs have reduced SG RBP expression relative to multipotent progenitor cells. SGs are dynamic biomolecular condensates, composed of ribonucleoprotein complexes, which may form as a protective mechanism against a variety of stress conditions. SGs are gaining increasing attention as contributors to cancer metastasis and chemoresistance, but their role in LSCs is virtually unexplored. The intriguing enrichment of SG RBPs in our screen hints at a larger role for SGs in LSC maintenance and/or stress resistance. To address this, we focused on the core SG nucleation RBP, G3BP1, a molecular switch that guides SG assembly. Fluorescence microscopy was first performed to describe SG dynamics in leukemia vs primitive normal cells where interestingly, a subset of patient AML cells contained G3BP1 SGs in the absence of added stress. In contrast, primitive healthy murine stem cells had reduced heat shock-induced SG formation in comparison to downstream normal progenitors. A series of shG3BP1 knockdown experiments performed in human leukemia cell lines and patient AML specimens demonstrated that shG3BP1-transduced cells showed increased apoptosis and differentiation and reduced colony forming capacity and competition in vitro compared to control cells. The same set of assays showed minimal impact of SG impairment to healthy umbilical cord blood cells. Importantly, in vivo xenotransplantation demonstrated that LSCs with G3BP1 knocked down were significantly compromised in their capacity to generate leukemic grafts whereas normal HSCs assessed in a similar manner display much less G3BP1 dependency. Towards determination of the SG mechanisms essential for bulk AML and LSC regulation, we performed matched RNA-sequencing and proteomics in G3BP1 knocked down AML cells as well as BioID in AML cell lines, primary patient AML cell subsets, and healthy hematopoietic cells to allow for the uncovering of an AML specific SG proteome. These analyses highlighted numerous potential SG mechanisms of cancer cell stress resistance and/or stem cell regulatory processes including sequestration of apoptotic factors and oncogenic signalling pathways. Of note, RNA-seq and proteomics in SG impaired shG3BP1 leukemia cells showed both a negative enrichment of the SG proteome and LSC signature as well as a positive enrichment of the innate immune response. Interestingly, our BioID uncovered that the AML specific SG proteome is also enriched for positive regulation of RIG-I signaling pathway (GO: Biological Process). Together this data suggests SGs play a protective role through the downregulation of overactive innate immune signalling in AML. Work using optimized small input approaches to map SG RNA and protein interactomes in LSCs vs HSCs to confirm similar regulatory relationships as well as identify LSC-context specific ones is ongoing. Overall, these results elucidate SGs as key AML LSC stress adaptive regulatory networks and highlights them as novel targets for consideration in AML therapeutics.

Fault estimation and tolerant bumpless transfer control for switched systems

A novel fault estimation and tolerant bumpless transfer control strategy is proposed for switched systems. In this control scheme, a switched estimator, a switched estimator-based tolerant bumpless transfer controller and a switching logic are jointly designed. The estimator relaxes the assumption on the usual observer matching condition in the fault estimation issue of switched linear systems. The controller can tolerate the fault with small input bumps at switching time. The switching law is an improvement of the traditional state-dependent switching law with dwell time still maintained, permitting the increase of the Lyapunov function for subsystems within the ensured dwell time interval. Through the presented control scheme, a criterion is developed to drive the solvability of the fault estimation and tolerant bumpless transfer control issue for the SSs. Also, with this criterion, this control issue of subsystems can be unsolvable. An example of regulating the current for a chua’s circuit system is provided for purpose of verifying the theoretical result.

Detecting emergency vehicles With 1D-CNN using fourier processed audio signals

Emergency Vehicle Detection (EVD) is a critical task in transportation management. Vehicle audio analysis provides affordable and insightful data, compared to established yet expensive vehicle detection methods. This research presents two contributions. Firstly, to decrease the processing power and memory demands, it offers a method to detect the optimal maximum sampling frequencies through an analysis of correlation between spectral entropy (SE) and dataset variance, grounded in the principles of the discrete Fourier transform (DFT) and Nyquist–Shannon theorem. Secondly, to address the inconsistency between recorded data in different datasets, this paper proposes segmenting audio into small input time windows (ITW) and applying individualized normalization. Training the model on dataset-1 achieves 98.37% accuracy on the validation set, with 86.83% accuracy on dataset-2. Comparative analysis with the baseline Long-Short-Term Memory model shows 22.79% performance improvement in favor of the proposed 1D-CNN model. Overall, this model outperforms state-of-the-art techniques, achieving accuracy rate of 98.37%.

Easyab Unravels Oncokinase Driven Regulation of Splicing Factors in AML

Acute myeloid leukemia (AML) is the most commonly diagnosed leukemia in adults. Gene mutations involved in various cellular functions are known to cause AML. A large proportion of AML is driven by aberrant activation of kinases and signal transducers such as FLT3, cKIT, PTPN11, JAK2, KRAS, NRAS, and to a lesser extent, JAK1, JAK3, and CSF3R. Treatment for AML has remained the same in the past five decades. Recent developments led to the availability of targeted therapies and enabled the practice of additional drugs like tyrosine kinase inhibitors. However, drug resistance and tolerance mechanism remain a challenge to finding a cure for the disease. Though we can identify and target kinase mutations individually, we have yet to learn how these mutations, like FLT3-ITD, change the cell's intrinsic signalling environment and cooperate with other oncogenes. Understanding these signalling behaviors is crucial for effective treatment outcomes and avoiding the development of drug resistance and tolerance mechanisms. Protein post-translational modifications (PTM) such as phosphorylation, ubiquitination, methylation, control the activity, fate, and spatial location of the proteins, thereby controlling processes such as signal transduction. Mass spectrometry-based proteomics offers unbiased profiling of PTMs. However, comprehensive analyses of PTMs require conventional antibody-based enrichment approaches and large starting materials, laborious, low throughput, cost prohibitive, and time-consuming, thus limiting the application to cell line-based studies. It also limits application in exploring unbiased-global cell signaling in patient samples. To overcome these limitations, we developed a loss-less sample preparation method named ‘EasyAb’. We show that our short, high throughput and low-cost approach enables capturing PTM peptides from very small inputs. As proof of principle, we enriched tyrosine-phosphorylated peptides from Molm13 and MV4-11 cell lines expressing FLT3-ITD kinase. Enrichment resulted in the ultra-deep identification of 4,416 tyrosine phosphosites, double the number of sites found in the conventional protocols, and took only one day of processing. About 40 % of the tyrosine phosphorylated proteins enriched are membrane proteins, including seven tyrosine phosphorylation sites on FLT3 kinase, of which FLT3 pY401 is a novel site discovered. To test the applicability of the EasyAb protocol in patient samples, we performed phosphotyrosine profiling of AML samples expressing FLT3 WT (n=9) and FLT3-ITD (n=5) mutation. EasyAb enrichment for phosphotyrosine peptides resulted in identifying 6,339 phosphotyrosine sites, of which 2300 sites were quantified. Comparing FLT3WT and ITD samples revealed more than 500 phosphosites significantly regulated. The significantly changing sites include several known signalling molecules such as STAT5, SYK, BTK, GAB2, GRB2, and PTPN11. The data also revealed differential tyrosine phosphorylation of proteins involved in various cellular functions such as autophagy, apoptosis, cell cycle, DNA damage repair proteins, phosphatases, and kinases. Interestingly, we also uncovered a class of proteins belonging to mRNA processing and splicing pathways distinctly tyrosine phosphorylated in FLT3-ITD samples. It is known that splicing factor mutations drive leukemia. Conversely, little is known about the regulatory mechanism of splicing factors by singling in cancer. To further investigate FLT3-ITD signalling regulation of splicing factors, we treated Molm13 and MV4-11 cells with Quizartinib and profiled tyrosine phosphorylation. Inhibition of FLT3-ITD lead to downregulation in tyrosine phosphorylation of several splicing factors. We have previously reported that splicing factor phosphorylation by JAK2 kinase plays a crucial role in Jak inhibitor persistence phenotype. Importantly, FLT3-ITD-mediated tyrosine phosphorylation-based regulation of splicing factors is unknown, and this process could be crucial for FLT3-ITD-mediated tumorigenesis. The functional significance of this novel regulatory mechanism is being studied. In conclusion, we have developed an effective tool for studying post-translational modifications that require antibody-based enrichment (tyrosine phosphorylation, ubiquitination, methylation, acetylation, and sumoylation) relatively quickly in primary patient samples.

Novel perturbation mechanism underlying the network fragility evolution

Studies have shown that fragility is an effective marker for seizures and seizure onset zone (SOZ). Through analysis and simulation of a probabilistic neural network under different inputs, the regularization mechanism of external input perturbations on the fragility is explored. It is theoretically found that the fragility of a perturbed node within seizure network is inversely associated with the received perturbation input, while the fragility of the other unperturbed nodes always oppositely changes with this perturbed node. By terming the node with high fragility as the fragile node (FN), it is interestingly shown that the FN would evolve to the node with the smallest input. Then, the network fragility is further investigated. Results show that the non-uniform perturbation inputs can more easily impact the network fragility. In addition, noise-induced variations of network connection can degrade the network fragility to some extent. Finally, the real data from a patient with epilepsy have verified the universality of the above obtained findings. These results may provide possible insights into stimulation strategies for seizure control in clinic.

Experimental study fluidized bed reactor using number hole 8 to see distribution gas fluid pressure

Innovations related to fluidization systems using fluidized bed reactor are still needed to be developed in order to improve the fluidization process services for particle in order to produce the right fluid pressure in certain fluidized bed spaces that are currently operating. In this research, the pressurized fluid in question is air sourced from a compressor which has been arranged in such a way. The aim of the research is to obtain information regarding the characteristics of bubble resulting from the air pressure process pressing fluidized system particle. The research method was carried out by experimenting with testing fluidized bed reactor as test model in the form 8 hole, providing air pressure from compressor then observing the characteristics visually. The results of this visually documented research have been carried out and obtained. By using the number of hole 8 with height silica sand inserted to height of 25 cm, which is measured from the beginning of the hole plate before pressure is applied. After being given high pressure the bed increased to 27.6 cm, the highest bubble diameter was obtained at the input air pressure of 8 bar, which was 3.9 cm, with bed silica sand produced 26.1 cm, where the babble condition began to appear after the 9th second. Other characteristics also obtained reactor temperature of 25.45 °C where this condition is the smallest when compared to other pressure input results, this is because the small input pressure causes the temperature to be low, while the large pressure input temperature becomes higher due to many factor, including friction between silica sand, silica sand collisions, and faster movement of particle material.

Robust machine learning modeling for predictive control using Lipschitz-Constrained Neural Networks

Neural networks (NNs) have emerged as a state-of-the-art method for modeling nonlinear systems in model predictive control (MPC). However, the robustness of NNs, in terms of sensitivity to small input perturbations, remains a critical challenge for practical applications. To address this, we develop Lipschitz-Constrained Neural Networks (LCNNs) for modeling nonlinear systems and derive rigorous theoretical results to demonstrate their effectiveness in approximating Lipschitz functions, reducing input sensitivity, and preventing over-fitting. Specifically, we first prove a universal approximation theorem to show that LCNNs using SpectralDense layers can approximate any 1-Lipschitz target function. Then, we prove a probabilistic generalization error bound for LCNNs using SpectralDense layers by using their empirical Rademacher complexity. Finally, the LCNNs are incorporated into the MPC scheme, and a chemical process example is utilized to show that LCNN-based MPC outperforms MPC using conventional feedforward NNs in the presence of training data noise.