Training Epochs Research Articles

Life regression based patch slimming for vision transformers

Vision transformers have achieved remarkable success in computer vision tasks by using multi-head self-attention modules to capture long-range dependencies within images. However, the high inference computation cost poses a new challenge. Several methods have been proposed to address this problem, mainly by slimming patches. In the inference stage, these methods classify patches into two classes, one to keep and the other to discard in multiple layers. This approach results in additional computation at every layer where patches are discarded, which hinders inference acceleration. In this study, we tackle the patch slimming problem from a different perspective by proposing a life regression module that determines the lifespan of each image patch in one go. During inference, the patch is discarded once the current layer index exceeds its life. Our proposed method avoids additional computation and parameters in multiple layers to enhance inference speed while maintaining competitive performance. Additionally, our approach11https://github.com/cjwcommuny/life-regression. requires fewer training epochs than other patch slimming methods.

BiLSTM-CNN with fixed weight approach for tracking speech articulatory features

Speech production is a complex sequential process which involve the coordination of various articulatory features. Among them tongue being a highly versatile active articulator responsible for shaping airflow to produce targeted speech sounds that are intellectual, clear, and distinct. This paper presents a novel approach for predicting tongue and lip articulatory features involved in a given speech acoustics using a stacked Bidirectional Long Short-Term Memory (BiLSTM) architecture, combined with a one-dimensional Convolutional Neural Network (CNN) for post-processing with fixed weights initialization. The proposed network is trained with two datasets consisting of simultaneously recorded speech and Electromagnetic Articulography (EMA) datasets, each introducing variations in terms of geographical origin, linguistic characteristics, phonetic diversity, and recording equipment. The performance of the model is assessed in Speaker Dependent (SD), Speaker Independent (SI), corpus dependent (CD) and cross corpus (CC) modes. Experimental results indicate that the proposed model with fixed weights approach outperformed the adaptive weights initialization with in relatively minimal number of training epochs. These findings contribute to the development of robust and efficient models for articulatory feature prediction, paving the way for advancements in speech production research and applications.

Evaluating the stealth of reinforcement learning-based cyber attacks against unknown scenarios using knowledge transfer techniques

Organizations are vulnerable to cyber attacks as they rely on computer networks and the internet for communication and data storage. While Reinforcement Learning (RL) is a widely used strategy to simulate and learn from these attacks, RL-guided offensives against unknown scenarios often lead to early exposure due to low stealth resulting from mistakes during the training phase. To address this issue, this work evaluates if the use of Knowledge Transfer Techniques (KTT), such as Transfer Learning and Imitation Learning, reduces the probability of early exposure by smoothing mistakes during training. This study developed a laboratory platform and a method to compare RL-based cyber attacks using KTT for unknown scenarios. The experiments simulated 2 unknown scenarios using 4 traditional RL algorithms and 4 KTT. In the results, although some algorithms using KTT obtained superior results, they were not so significant for stealth during the initial epochs of training. Nevertheless, experiments also revealed that throughout the entire learning cycle, Trust Region Policy Optimization (TRPO) is a promising algorithm for conducting cyber offensives based on Reinforcement Learning.

BERT-based language model for accurate drug adverse event extraction from social media: implementation, evaluation, and contributions to pharmacovigilance practices.

Social media platforms serve as a valuable resource for users to share health-related information, aiding in the monitoring of adverse events linked to medications and treatments in drug safety surveillance. However, extracting drug-related adverse events accurately and efficiently from social media poses challenges in both natural language processing research and the pharmacovigilance domain. Recognizing the lack of detailed implementation and evaluation of Bidirectional Encoder Representations from Transformers (BERT)-based models for drug adverse event extraction on social media, we developed a BERT-based language model tailored to identifying drug adverse events in this context. Our model utilized publicly available labeled adverse event data from the ADE-Corpus-V2. Constructing the BERT-based model involved optimizing key hyperparameters, such as the number of training epochs, batch size, and learning rate. Through ten hold-out evaluations on ADE-Corpus-V2 data and external social media datasets, our model consistently demonstrated high accuracy in drug adverse event detection. The hold-out evaluations resulted in average F1 scores of 0.8575, 0.9049, and 0.9813 for detecting words of adverse events, words in adverse events, and words not in adverse events, respectively. External validation using human-labeled adverse event tweets data from SMM4H further substantiated the effectiveness of our model, yielding F1 scores 0.8127, 0.8068, and 0.9790 for detecting words of adverse events, words in adverse events, and words not in adverse events, respectively. This study not only showcases the effectiveness of BERT-based language models in accurately identifying drug-related adverse events in the dynamic landscape of social media data, but also addresses the need for the implementation of a comprehensive study design and evaluation. By doing so, we contribute to the advancement of pharmacovigilance practices and methodologies in the context of emerging information sources like social media.

A bilateral filtering-based image enhancement for Alzheimer disease classification using CNN.

This study aims to develop an optimally performing convolutional neural network to classify Alzheimer's disease into mild cognitive impairment, normal controls, or Alzheimer's disease classes using a magnetic resonance imaging dataset. To achieve this, we focused the study on addressing the challenge of image noise, which impacts the performance of deep learning models. The study introduced a scheme for enhancing images to improve the quality of the datasets. Specifically, an image enhancement algorithm based on histogram equalization and bilateral filtering techniques was deployed to reduce noise and enhance the quality of the images. Subsequently, a convolutional neural network model comprising four convolutional layers and two hidden layers was devised for classifying Alzheimer's disease into three (3) distinct categories, namely mild cognitive impairment, Alzheimer's disease, and normal controls. The model was trained and evaluated using a 10-fold cross-validation sampling approach with a learning rate of 0.001 and 200 training epochs at each instance. The proposed model yielded notable results, such as an accuracy of 93.45% and an area under the curve value of 0.99 when trained on the three classes. The model further showed superior results on binary classification compared with existing methods. The model recorded 94.39%, 94.92%, and 95.62% accuracies for Alzheimer's disease versus normal controls, Alzheimer's disease versus mild cognitive impairment, and mild cognitive impairment versus normal controls classes, respectively.

EH-former: Regional easy-hard-aware transformer for breast lesion segmentation in ultrasound images

Breast lesion segmentation of ultrasound images plays a crucial role in early screening and diagnosis of breast lesions. However, accurately segmenting lesions in breast ultrasound (BUS) images is challenging due to prevalent issues such as low contrast, intense speckle noise, and blurred lesion boundaries. Although existing deep learning-based segmentation models have made significant progress, few have strategically addressed these complex and noisy regional features in BUS images. The easy-to-hard manner in Curriculum Learning (CL) appears promising, but it often remains at the sample level and does not adequately address regional complexities. To address this, we design a region-wise CL to dynamically adjust the focus on hard regional features in BUS images. Specifically, we propose a Regional Easy-Hard-Aware Transformer (EH-Former), structured in two stages for lesion segmentation in BUS images. The first stage incorporates uncertainty estimation for dividing regional difficulty. In the second stage, we propose a novel Adaptive Easy-Hard region Separator (AdaSep), a module employing uncertainty-aware regularization to separate features of varying difficulties, allowing the two streams within EH-Former to focus on learning regional features of different complexities. Additionally, we develop a Dynamic Easy-Hard Feature Fusion (D-Fusion) module, dynamically adjusting the fusion weight of easy and hard regional features based on the current training epoch to achieve progressive regional feature learning. Extensive experimental results on five public datasets show that the proposed EH-Former consistently outperforms state-of-the-art methods in most metrics and exhibits better domain generalization capabilities. Furthermore, our region-wise CL significantly enhances the performance of EH-Former in detecting complex tissue structures and noisy areas that are challenging to segment accurately. The source code is available at https://github.com/lele0109/EH-Former.

On-device edge-learning for cardiac abnormality detection using a bio-inspired and spiking shallow network

This work introduces on-device edge learning for cardiac abnormality detection by merging spiking 2D Convolutional Long-Short-Term Memory (ConvLSTM2D) with a bio-inspired shallow neural network, referred to as Closed-form Continuous-time (CfC), to form the sCCfC model. The model achieves an F1 score and AUROC of 0.82 and 0.91 in cardiac abnormalities detection. These results are comparable to the non-spiking ConvLSTM2D–CfC (ConvCfC) model [Huang et al., J. Cardiovasc. Transl. Res. (published online, 2024)]. Notably, the sCCfC model demonstrates a significantly higher energy efficiency with an estimated power consumption of 4.68 μJ/Inf (per inference) on an emulated Loihi’s neuromorphic chip architecture, in contrast to ConvCfC model’s consumption of 450 μJ/Inf on a conventional processor. In addition, as a proof-of-concept, we deployed the sCCfC model on the conventional and relatively resource-constrained Radxa Zero, which is equipped with an Amlogic S905Y2 processor for on-device training, which resulted in performance improvements. After initial training of two epochs on a conventional Graphics Processing Unit, the F1 score and AUROC improved from 0.46 and 0.65 to 0.56 and 0.73, respectively, with five additional epochs of on-device training. Furthermore, when presented with a new dataset, the sCCfC model showcases strong out-of-sample generalization capabilities that can constitute a pseudo-perspective test, achieving an F1 score and AUROC of 0.71 and 0.86, respectively. The spiking sCCfC also outperforms the non-spiking ConvCfC model in robustness regarding effectively handling missing electrocardiogram (ECG) channels during inference. The model’s efficacy extends to single-lead ECG analysis, demonstrating reasonable accuracy in this context, while the focus of our work has been on the computational and memory complexities of the model.

Bandit-NAS: Bandit sampling and training method for Neural Architecture Search

Existing Neural Architecture Search algorithms achieve a low error rate in vision tasks, such as image classification, by training child networks with equal resources during the search. However, it is unnecessary to allocate equal resources or fully converge scores to assess which child architectures should be adopted, resulting in computational redundancy. In this study, we present Bandit-NAS, an approach that automatically computes data slicing and training time for each child network. Firstly, we formulate the search for the optimal training time for a given resource as an M-armed bandit problem. Secondly, we extend the original NAS methods by proposing an end-to-end bandit algorithm, combined with reinforcement learning-based NAS algorithms, to determine an update strategy. Bandit-NAS enables simultaneous training of M child networks within a specified resource constraint (one epoch training time), with the allocation of training data based on the current accuracy of the child networks, thereby minimizing their error rate. Experimental results on 3 different datasets, MNIST, CIFAR-10 and CIFAR-100 demonstrate the superiority of Bandit-NAS over baseline NAS algorithms, such as ENAS and DQNAS, achieving lower error rates and faster search time.

Flexible and Energy-Efficient Synaptic Transistor with Quasi-Linear Weight Update Protocol by Inkjet Printing of Orientated Polar-Electret/High-k Oxide Composite Dielectric.

Inkjet printing artificial synapse is cost-effective but challenging in emulating synaptic dynamics with a sufficient number of effective weight states under ultralow voltage spiking operation. A synaptic transistor gated by inkjet-printed composite dielectric of polar-electret polyvinylpyrrolidone (PVP) and high-k zirconia oxide (ZrOx) is proposed and thus synthesized to solve this issue. Quasi-linear weight update with a large variation margin is obtained through the coupling effect and the facilitation of dipole orientation, which can be attributed to the orderly arranged molecule chains induced by the carefully designed microfluidic flows. Crucial features of biological synapses including long-term plasticity, spike-timing-dependence-plasticity (STDP), "Learning-Experience" behavior, and ultralow energy consumption (<10 fJ/pulse) are successfully implemented on the device. Simulation results exhibit an excellent image recognition accuracy (97.1%) after 15 training epochs, which is the highest for printed synaptic transistors. Moreover, the device sustained excellent endurance against bending tests with radius down to 8 mm. This work presents a very viable solution for constructing the futuristic flexible and low-cost neural systems.

Principal Component Networks: Utilizing Low-Rank Activation Structure to Reduce Parameters Early in Training

Recent works show that overparameterized neural networks contain small subnetworks that exhibit comparable accuracy to the full model when trained in isolation. These results highlight the potential to reduce the computational costs of deep neural network training without sacrificing generalization performance. Initial approaches for finding these small networks relied on expensive multi-round train-and-prune procedures, limiting their practical potential, but more recent work identifies subnetworks using structured pruning techniques early in training. In this article, we study network activations, rather than network weights, and find that hidden layer activations in overparameterized networks exist primarily in subspaces smaller than the actual model width. We further notice that these subspaces can be identified early in training. Based on these observations, we show how to efficiently find small networks that exhibit similar accuracy to their overparameterized counterparts after only a few training epochs. We term these network architectures Principal Component Networks (PCNs). PCNs compress individual layers by retaining only the high variance linear combinations of channels—defined by the principal components of the layer inputs—a key difference from structured pruning techniques which focus on individual channel pruning using localized channel measurements. We evaluate PCNs on CIFAR-10 and ImageNet for VGG and ResNet style architectures and compare against existing methods for subnetwork identification during early training. We find that PCNs consistently reduce parameter counts with little accuracy loss, thus providing the potential to reduce the computational costs of deep neural network training. Beyond model compression, we also connect our observation regarding hidden layer activations to the feature representations learned by neural networks and discuss areas for future work.

The principle of minimum pressure gradient: An alternative basis for physics-informed learning of incompressible fluid mechanics

Recent advances in the application of physics-informed learning in the field of fluid mechanics have been predominantly grounded in the Newtonian framework, primarily leveraging Navier–Stokes equations or one of their various derivatives to train a neural network. Here, we propose an alternative approach based on variational methods. The proposed approach uses the principle of minimum pressure gradient combined with the continuity constraint to train a neural network and predict the flow field in incompressible fluids. We describe the underlying principles of the proposed approach, then use a demonstrative example to illustrate its implementation, and show that it reduces the computational time per training epoch when compared to the conventional approach.

DN-DETR: Accelerate DETR Training by Introducing Query DeNoising.

We present in this paper a novel denoising training method to speed up DETR (DEtection TRansformer) training and offer a deepened understanding of the slow convergence issue of DETR-like methods. We show that the slow convergence results from the instability of bipartite graph matching which causes inconsistent optimization goals in early training stages. To address this issue, except for the Hungarian loss, our method additionally feeds GT bounding boxes with noises into the Transformer decoder and trains the model to reconstruct the original boxes, which effectively reduces the bipartite graph matching difficulty and leads to faster convergence. Our method is universal and can be easily plugged into any DETR-like method by adding dozens of lines of code to achieve a remarkable improvement. As a result, our DN-DETR results in a remarkable improvement ( +1.9AP) under the same setting and achieves 46.0 AP and 49.5 AP trained for 12 and 50 epochs with the ResNet-50 backbone. Compared with the baseline under the same setting, DN-DETR achieves comparable performance with 50% training epochs. We also demonstrate the effectiveness of denoising training in CNN-based detectors (Faster R-CNN), segmentation models (Mask2Former, Mask DINO), and more DETR-based models (DETR, Anchor DETR, Deformable DETR).

Weight Averaging and re-adjustment ensemble for QRCD

Question Answering (QA) is a prominent task in the field of Natural Language Processing (NLP) with extensive applications. Recently, there has been a notable surge in research interest concerning the development of QA systems for the Holy Qur’an, an Islamic religious text. The Qur’an Reading Comprehension Dataset (QRCD) Malhas and Elsayed (2020) is a highly commendable effort in this respect. It stands as the first benchmark dataset specifically designed for a set of directly answerable questions from the Qur’an. Each question in the dataset is meticulously labeled with all potential answers sourced from the Holy Qur’an. From our perspective, the main challenge in QRCD stems from the limited volume of training data it offers. As a solution we propose an innovative approach to build a Deep Neural Network (DNN) ensemble, centered around Ara-Electra model (Antoun et al., 2021), that we called Weight Averaging and Re-adjustment (WAR) model. The model is constructed by computing running averages of all model states that evolve during a single training session and ensuring that model weights are readjusted prior to each training epoch, in order to hold it back from over fitting the training data. The scheme results in a single standalone model that exhibits the benefits of multi-model ensembles. It is distinguished from other ensembles proposed for QRCD that accumulates outputs from multiple expert models and employs classic techniques like hard voting or score averaging on output probabilities to build unified results. Each expert model costs individual training time and compute resources. The WAR model outperforms existing systems with improved generalization over unseen data. It achieves F1, partial Reciprocal Rank (pRR), and exact-match (EM) scores of 0.567, 0.60 and 0.29 respectively, exceeding best reported QRCD scores by 3%, 1.5% and 0.69% respectively. Notably, we are comparing our results with the top scores from different models, highlighting our model’s consistent performance across all three metrics.

Prediction of constrained modulus for granular soil using 3D discrete element method and convolutional neural networks

To efficiently predict the mechanical parameters of granular soil based on its random micro-structure, this study proposed a novel approach combining numerical simulation and machine learning algorithms. Initially, 3500 simulations of one-dimensional compression tests on coarse-grained sand using the three-dimensional (3D) discrete element method (DEM) were conducted to construct a database. In this process, the positions of the particles were randomly altered, and the particle assemblages changed. Interestingly, besides confirming the influence of particle size distribution parameters, the stress-strain curves differed despite an identical gradation size statistic when the particle position varied. Subsequently, the obtained data were partitioned into training, validation, and testing datasets at a 7:2:1 ratio. To convert the DEM model into a multi-dimensional matrix that computers can recognize, the 3D DEM models were first sliced to extract multi-layer two-dimensional (2D) cross-sectional data. Redundant information was then eliminated via gray processing, and the data were stacked to form a new 3D matrix representing the granular soil's fabric. Subsequently, utilizing the Python language and Pytorch framework, a 3D convolutional neural networks (CNNs) model was developed to establish the relationship between the constrained modulus obtained from DEM simulations and the soil's fabric. The mean squared error (MSE) function was utilized to assess the loss value during the training process. When the learning rate (LR) fell within the range of 10-5–10-1, and the batch sizes (BSs) were 4, 8, 16, 32, and 64, the loss value stabilized after 100 training epochs in the training and validation dataset. For BS = 32 and LR = 10-3, the loss reached a minimum. In the testing set, a comparative evaluation of the predicted constrained modulus from the 3D CNNs versus the simulated modulus obtained via DEM reveals a minimum mean absolute percentage error (MAPE) of 4.43% under the optimized condition, demonstrating the accuracy of this approach. Thus, by combining DEM and CNNs, the variation of soil's mechanical characteristics related to its random fabric would be efficiently evaluated by directly tracking the particle assemblages.

DEW: A wavelet approach of rare sound event detection.

This paper presents a novel sound event detection (SED) system for rare events occurring in an open environment. Wavelet multiresolution analysis (MRA) is used to decompose the input audio clip of 30 seconds into five levels. Wavelet denoising is then applied on the third and fifth levels of MRA to filter out the background. Significant transitions, which may represent the onset of a rare event, are then estimated in these two levels by combining the peak-finding algorithm with the K-medoids clustering algorithm. The small portions of one-second duration, called 'chunks' are cropped from the input audio signal corresponding to the estimated locations of the significant transitions. Features from these chunks are extracted by the wavelet scattering network (WSN) and are given as input to a support vector machine (SVM) classifier, which classifies them. The proposed SED framework produces an error rate comparable to the SED systems based on convolutional neural network (CNN) architecture. Also, the proposed algorithm is computationally efficient and lightweight as compared to deep learning models, as it has no learnable parameter. It requires only a single epoch of training, which is 5, 10, 200, and 600 times lesser than the models based on CNNs and deep neural networks (DNNs), CNN with long short-term memory (LSTM) network, convolutional recurrent neural network (CRNN), and CNN respectively. The proposed model neither requires concatenation with previous frames for anomaly detection nor any additional training data creation needed for other comparative deep learning models. It needs to check almost 360 times fewer chunks for the presence of rare events than the other baseline systems used for comparison in this paper. All these characteristics make the proposed system suitable for real-time applications on resource-limited devices.

DenseNet Architecture for Efficient and Accurate Recognition of Javanese Script Hanacaraka Character

This study introduced a specifically optimized DenseNet architecture for recognizing Javanese Hanacaraka characters, focusing on enhancing efficiency and accuracy. The research aimed to preserve and celebrate Java’s rich cultural heritage and historical significance through the development of precise character recognition technology. The method used advanced techniques within convolutional neural networks (CNN) to integrate feature extraction across densely connected layers efficiently. The result of this study was that the developed model achieved a training accuracy of 100% and a validation accuracy of approximately 99.50% after 30 training epochs. Furthermore, when tested on previously unseen datasets, the model exhibited exceptional accuracy, precision, recall, and F1-score, reaching 100%. These findings underscored the remarkable capability of DenseNet architecture in character recognition, even across novel datasets, suggesting significant potential for automating Javanese Hanacaraka text processing across various applications, ranging from text recognition to digital archiving. The conclusion drawn from this study suggests that optimizing DenseNet architecture can be a significant step in preserving and developing character recognition technology for Javanese

Control-Etched Ti3C2Tx MXene Nanosheets for a Low-Voltage-Operating Flexible Memristor for Efficient Neuromorphic Computation.

Hardware neural networks with mechanical flexibility are promising next-generation computing systems for smart wearable electronics. Overcoming the challenge of developing a fully synaptic plastic network, we demonstrate a low-operating-voltage PET/ITO/p-MXene/Ag flexible memristor device by controlling the etching of aluminum metal ions in Ti3C2Tx MXene. The presence of a small fraction of Al ions in partially etched MXene (p-Ti3C2Tx) significantly suppresses the operating voltage to 1 V compared to 7 V from fully Al etched MXene (f-Ti3C2Tx)-based devices. Former devices exhibit excellent non-volatile data storage properties, with a robust ∼103 ON/OFF ratio, high endurance of ∼104 cycles, multilevel resistance states, and long data retention measured up to ∼106 s. High mechanical stability up to ∼73° bending angle and environmental robustness are confirmed with consistent switching characteristics under increasing temperature and humid conditions. Furthermore, a p-Ti3C2Tx MXene memristor is employed to mimic the biological synapse by measuring the learning-forgetting pattern for ∼104 cycles as potentiation and depression. Spike time-dependent plasticity (STDP) based on Hebb's Learning rules is also successfully demonstrated. Moreover, a remarkable accuracy of ∼95% in recognizing modified patterns from the National Institute of Standards and Technology (MNIST) data set with just 29 training epochs is achieved in simulation. Ultimately, our findings underscore the potential of MXene-based flexible memristor devices as versatile components for data storage and neuromorphic computing.

Deep learning for predictive window operation modeling in open-plan offices

This study explores how the past, both short and long-term, affects the predictive window operation modeling in open-plan offices. To achieve this, the study proposes a deep learning method that uses long short-term memory (LSTM) artificial neural networks. The proposed model is an ensemble of LSTM networks that utilize relevant indoor environmental data from multiple sensory systems and weather data from a nearby weather station to predict the window state with different predictive horizons. The drawbacks of such networks are extensive training time and the lack of interpretability. Hence, Bayesian optimization is proposed for efficient hyperparameter tuning of the network. Furthermore, explainability through a model agnostic approach is utilized for feature importance ranking.The performance of the LSTM networks was compared to a dense feedforward neural network (DFNN) on an imbalanced dataset to evaluate the effect of the long-term past. Handling imbalanced window operation data was improved by employing class weight and early stopping, leading to a higher true positive rate, marginally better F1-score, and fewer network training epochs. The results indicate that a stateful bi-directional LSTM (Bi-LSTM) network models the data sequence more efficiently in aggregated comparison than other LSTM types and DFNN. However, the DFNN slightly outperformed LSTM networks in interval-by-interval comparison, with an F1-score of 0.88 when predicting the window state 10 min ahead, shifting to 0.82 for a 60-minute predictive horizon. Additionally, the study showed that the ensemble model performed better in terms of F1-score than the individual models.

United We Stand: Using Epoch-Wise Agreement of Ensembles to Combat Overfit

Deep neural networks have become the method of choice for solving many classification tasks, largely because they can fit very complex functions defined over raw data. The downside of such powerful learners is the danger of overfit. In this paper, we introduce a novel ensemble classifier for deep networks that effectively overcomes overfitting by combining models generated at specific intermediate epochs during training. Our method allows for the incorporation of useful knowledge obtained by the models during the overfitting phase without deterioration of the general performance, which is usually missed when early stopping is used. To motivate this approach, we begin with the theoretical analysis of a regression model, whose prediction - that the variance among classifiers increases when overfit occurs - is demonstrated empirically in deep networks in common use. Guided by these results, we construct a new ensemble-based prediction method, where the prediction is determined by the class that attains the most consensual prediction throughout the training epochs. Using multiple image and text classification datasets, we show that when regular ensembles suffer from overfit, our method eliminates the harmful reduction in generalization due to overfit, and often even surpasses the performance obtained by early stopping. Our method is easy to implement and can be integrated with any training scheme and architecture, without additional prior knowledge beyond the training set. It is thus a practical and useful tool to overcome overfit.

Quality Metrics of Automated Machinery in Potato Plant Cultivation for Breeding and Seed Production

The paper notes the significance of promptly identifying infected plants when cultivating potatoes for breeding and seed production. Consequently, there is a need to undertake a series of initiatives aimed at developing a digital system for automated detection and recognition of both healthy and infected plants. (Research purpose) The research aims to determine the patterns of changes in the quality indicators of the machinery employed in cultivating potato plants. (Materials and methods) The research was carried out on the area of the selection-experimental plot. A system of criteria was developed to evaluate the identification of infected plants. (Results and discussion) The research assisted in identifying the required reliability of the measuring operation for the machine vision system and aided in predicting its current state for identifying infected plants. This was achieved by analyzing statistical data on the distribution of the indirect parameter (indications of infection on the inside of the plant leaf) and considering the margin of error in its measurements. The reliability of the system for identifying infected plants depends on the precision of technical instruments used to gauge the plant’s condition, the methodologies employed in measurement, the software utilized for processing the obtained data, and other parameters. (Conclusions) Measurement information management involves making a judicious selection of an indirect parameter that guarantees the precision of identifying infected plants with a confidence interval of 0.95. It is revealed that in the initial training epoch of the infected plant identification system, the accuracy of plant classification stood at 0.797, equivalent to 79.7 percent for all plants. The correctness of infected plant recognition was 0.607 or 60.7 percent. Moreover, the accuracy of correctly identifying infected plants was determined to be 0.607, or 60.7 percent. Notably, by this epoch, the accuracy of recognizing healthy plants had already reached 99.9 percent.