Loss In Classification Accuracy Research Articles

MixTrain: accelerating DNN training via input mixing.

Training Deep Neural Networks (DNNs) places immense compute requirements on the underlying hardware platforms, expending large amounts of time and energy. An important factor contributing to the long training times is the increasing dataset complexity required to reach state-of-the-art performance in real-world applications. To address this challenge, we explore the use of input mixing, where multiple inputs are combined into a single composite input with an associated composite label for training. The goal is for training on the mixed input to achieve a similar effect as training separately on each the constituent inputs that it represents. This results in a lower number of inputs (or mini-batches) to be processed in each epoch, proportionally reducing training time. We find that naive input mixing leads to a considerable drop in learning performance and model accuracy due to interference between the forward/backward propagation of the mixed inputs. We propose two strategies to address this challenge and realize training speedups from input mixing with minimal impact on accuracy. First, we reduce the impact of inter-input interference by exploiting the spatial separation between the features of the constituent inputs in the network's intermediate representations. We also adaptively vary the mixing ratio of constituent inputs based on their loss in previous epochs. Second, we propose heuristics to automatically identify the subset of the training dataset that is subject to mixing in each epoch. Across ResNets of varying depth, MobileNetV2 and two Vision Transformer networks, we obtain upto 1.6 × and 1.8 × speedups in training for the ImageNet and Cifar10 datasets, respectively, on an Nvidia RTX 2080Ti GPU, with negligible loss in classification accuracy.

Memory-bound k-mer selection for large and evolutionarily diverse reference libraries.

Using k-mers to find sequence matches is increasingly used in many bioinformatic applications, including metagenomic sequence classification. The accuracy of these downstream applications relies on the density of the reference databases, which are rapidly growing. Although the increased density provides hope for improvements in accuracy, scalability is a concern. Reference k-mers are kept in the memory during the query time, and saving all k-mers of these ever-expanding databases is fast becoming impractical. Several strategies for subsampling have been proposed, including minimizers and finding taxon-specific k-mers. However, we contend that these strategies are inadequate, especially when reference sets are taxonomically imbalanced, as are most microbial libraries. In this paper, we explore approaches for selecting a fixed-size subset of k-mers present in an ultra-large data set to include in a library such that the classification of reads suffers the least. Our experiments demonstrate the limitations of existing approaches, especially for novel and poorly sampled groups. We propose a library construction algorithm called k-mer RANKer (KRANK) that combines several components, including a hierarchical selection strategy with adaptive size restrictions and an equitable coverage strategy. We implement KRANK in highly optimized code and combine it with the locality-sensitive hashing classifier CONSULT-II to build a taxonomic classification and profiling method. On several benchmarks, KRANK k-mer selection significantly reduces memory consumption with minimal loss in classification accuracy. We show in extensive analyses based on CAMI benchmarks that KRANK outperforms k-mer-based alternatives in terms of taxonomic profiling and comes close to the best marker-based methods in terms of accuracy.

Memory-bound k -mer selection for large and evolutionary diverse reference libraries.

Using k -mers to find sequence matches is increasingly used in many bioinformatic applications, including metagenomic sequence classification. The accuracy of these down-stream applications relies on the density of the reference databases, which, luckily, are rapidly growing. While the increased density provides hope for dramatic improvements in accuracy, scalability is a concern. Reference k -mers are kept in the memory during the query time, and saving all k -mers of these ever-expanding databases is fast becoming impractical. Several strategies for subsampling have been proposed, including minimizers and finding taxon-specific k -mers. However, we contend that these strategies are inadequate, especially when reference sets are taxonomically imbalanced, as are most microbial libraries. In this paper, we explore approaches for selecting a fixed-size subset of k -mers present in an ultra-large dataset to include in a library such that the classification of reads suffers the least. Our experiments demonstrate the limitations of existing approaches, especially for novel and poorly sampled groups. We propose a library construction algorithm called KRANK (K-mer RANKer) that combines several components, including a hierarchical selection strategy with adaptive size restrictions and an equitable coverage strategy. We implement KRANK in highly optimized code and combine it with the locality-sensitive-hashing classifier CONSULT-II to build a taxonomic classification and profiling method. On several benchmarks, KRANK k -mer selection dramatically reduces memory consumption with minimal loss in classification accuracy. We show in extensive analyses based on CAMI benchmarks that KRANK outperforms k -mer-based alternatives in terms of taxonomic profiling and comes close to the best marker-based methods in terms of accuracy.

Automatic Identification and Severity Classification of Retinal Biomarkers in SD-OCT Using Dilated Depthwise Separable Convolution ResNet with SVM Classifier

Purpose Diagnosis of Uveitic Macular Edema (UME) using Spectral Domain OCT (SD-OCT) is a promising method for early detection and monitoring of sight-threatening visual impairment. Viewing multiple B-scans and identifying biomarkers is challenging and time-consuming for clinical practitioners. To overcome these challenges, this paper proposes an image classification hybrid framework for predicting the presence of biomarkers such as intraretinal cysts (IRC), hyperreflective foci (HRF), hard exudates (HE) and neurosensory detachment (NSD) in OCT B-scans along with their severity. Methods A dataset of 10880 B-scans from 85 Uveitic patients is collected and graded by two board-certified ophthalmologists for the presence of biomarkers. A novel image classification framework, Dilated Depthwise Separable Convolution ResNet (DDSC-RN) with SVM classifier, is developed to achieve network compression with a larger receptive field that captures both low and high-level features of the biomarkers without loss of classification accuracy. The severity level of each biomarker is predicted from the feature map, extracted by the proposed DDSC-RN network. Results The proposed hybrid model is evaluated using ground truth labels from the hospital. The deep learning model initially, identified the presence of biomarkers in B-scans. It achieved an overall accuracy of 98.64%, which is comparable to the performance of other state-of-the-art models, such as DRN-C-42 and ResNet-34. The SVM classifier then predicted the severity of each biomarker, achieving an overall accuracy of 89.3%. Conclusions A new hybrid model accurately identifies four retinal biomarkers on a tissue map and predicts their severity. The model outperforms other methods for identifying multiple biomarkers in complex OCT B-scans. This helps clinicians to screen multiple B-scans of UME more effectively, leading to better treatment outcomes.

Testing the accuracy of the SexEst software for sex estimation in a modern Greek sample

AbstractSex estimation from human skeletal remains is fundamental in osteoarcheology and forensic anthropology. The increasing availability of reference skeletal collections across the world has allowed the development of morphological and metric methods for skeletal sex estimation, some of which may be implemented in specialized computer software. The present study aims to evaluate the freely available SexEst software, which utilizes cranial and postcranial measurements, and different classification models for sex estimation, on a contemporary Greek population comprising of 227 (126 males and 101 females) adult individuals. After the calculation of intra‐observer error to assess the repeatability of the measurements, the proposed variables were tested for classification accuracy individually and in different combinations. Based on the results, the postcranial models outperformed the cranial ones in all cases and can be adequately applied on a Greek population sample. The light gradient boosting (LGB) algorithm yielded the highest correct classification rates when no missing values exist, while the linear discriminant analysis (LDA) models should only be used when dealing with missing data. The highest classification accuracy for a 0.65 posterior probability threshold was reached when utilizing a combination of postcranial variables (89.67%), while the lowest was achieved with the cranial measurement “Glabella‐occipital length” (45.00%). The same models yielded the highest and lowest accuracy for a 0.5 probability threshold, with values of 92.96% and 67.73%, respectively. Combining variables yielded higher accuracies in both skeletal regions, suggesting that the software would be more helpful in cases of intact skeletons. The loss of classification accuracy due to population specificity further corroborates the need to include different ancestries in sex estimation software.

Development of Fabrication Techniques for Magneto-Optical Diffractive Deep Neural Networks

Magneto-optical diffractive deep neural network (MO-D <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> NN) is a type of diffractive deep neural network that utilizes the magneto-optical effect of magnetic materials. MO-D <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> NN is composed of magnetic material, which is non-volatile and allows neurons to be rewritten. In our work, we fabricated and evaluated MO-D <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> NN with 2 layers containing 100 × 100 magnetic domains with a domain size of 1 μm. Nd <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.5</sub> Bi <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2.5</sub> Fe <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4</sub> GaO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">12</sub> thin films were prepared on the both sides of a Gd <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sub> Ga <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">5</sub> O <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">12</sub> substrate and magnetic domain patterns were recorded using the magneto-optical recording technique. The recording accuracies of first and second layer were 79% and 70%, respectively. We believed that magneto-optical recording technology can achieve a recording accuracy of over 90% by optimizing the fabrication and recording conditions. The optical setup was built and handwritten digit classification was carried out using fabricated MO-D <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> NN, but the number of classifications was lower than in the case of simulation. We considered that recording errors were responsible for the low number of classifications and assessed the effect of recording errors on accuracy by simulation. When the recording error was lower than 10%, the loss in classification accuracy was found to be only ~3%. Therefore, we are convinced that MO-D <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> NN has high physical implementation potential.

Booth Encoding-Based Energy Efficient Multipliers for Deep Learning Systems

Artificial intelligence on edge is a growing research field. In this paper, we propose a novel re-encoding scheme for reducing the size of the weights of deep neural networks (DNNs). The proposed re-encoding scheme exploits the Booth encoding scheme and the power-of-two (PO2) quantization to allow for very low energy computations during the inference of the neural networks with minimal loss in classification accuracy. We demonstrate the advantages of the proposed re-encoding scheme by computing a convolutional neural network (CNN) and a linear neural network on the proposed Extended Exact Multiplier and the proposed PO2 Multiplier. Our proposed PO2 quantization and re-encoding method reduce the model size for the CNN by 30.77% and the model size of the linear neural network by 49.86%. Furthermore, our multipliers reduce the inference energy for CNN by 50.6% and for the linear neural network by 90.1%. The PO2 Multiplier is proposed for the sensor-end computation of the linear neural network with a 77.32% reduction in the area relative to an exact Booth multiplier and it reduces the inference energy consumption of the linear neural network by 93.2% when compared to the unmodified exact multiplier. Our proposed scheme can be used to improve the energy consumption during inference for most Booth multipliers with minor modifications to the re-encoding signal arrangements. We also demonstrate that the proposed re-encoding scheme paired with the proposed multipliers outperforms all the existing designs in terms of resource utilization with a minimal impact on the inference accuracy of the neural networks.

An Ensemble Learning Method Based on One-Class and Binary Classification for Credit Scoring

It is crucial to correctly assess whether a potential borrower can repay the loan in the credit scoring model. The credit loan data has a serious data imbalance because the number of defaulters is far less than the nondefaulters. However, most current methods for dealing with data imbalance are designed to improve the classification performance of minority data, which will reduce the performance of majority data. For a financial institution, the economic loss caused by the decrease in the classification performance of nondefaulters (majority data) cannot be ignored. This paper proposes an ensemble learning method based on one-class and binary classification (EMOBC) for credit scoring. The purpose is to improve the classification accuracy of the minority class while mitigating the loss of classification accuracy of the majority class as much as possible. EMOBC uses undersampling for the majority class (nondefault samples in credit scoring) and perform binary-class learning on the balanced data to improve the classification accuracy of the minority. To alleviate the decline in classification performance of the majority class, EMOBC uses one-class and binary collaborative classification to train classifiers. The classification result is determined by the average of one-class and binary-class classifiers. The experimental results show that EMOBC has good comprehensive performance compared with the existing methods.

A Technique for Approximate Communication in Network-on-Chips for Image Classification

Approximation is an emerging design methodology for reducing power consumption and latency of on-chip communication in many computing applications. However, existing approximation techniques either achieve modest improvements in these metrics or require retraining after approximation. Since classifying many images introduces intensive on-chip communication, reductions in both network latency and power consumption are highly desired. In this paper, we propose an approximate communication technique (ACT) to improve the efficiency of on-chip communications for image classification applications. The proposed technique exploits the error-tolerance of the image classification process to reduce power consumption and latency of on-chip communications, resulting in better overall performance for image classification. This is achieved by incorporating novel quality control and data approximation mechanisms that reduce the packet size. In particular, the proposed quality control mechanisms identify the error-resilient variables and automatically adjust the error thresholds of the variables based on the image classification accuracy. The proposed data approximation mechanisms significantly reduce packet size when the variables are transmitted. The proposed technique reduces the number of flits in each data packet as well as the on-chip communication while maintaining an excellent image classification accuracy. Cycle-accurate simulation results show that ACT achieves 23% in network latency reduction and 24% in dynamic power reduction as compared to existing approximate communication techniques with less than 0.99% classification accuracy loss.

Bayesian Photonic Accelerators for Energy Efficient and Noise Robust Neural Processing

Artificial neural networks are efficient computing platforms inspired by the brain. Such platforms can tackle a vast area of real-life tasks ranging from image processing to language translation. Silicon photonic integrated chips (PICs), by employing coherent interactions in Mach-Zehnder interferometers, are promising accelerators offering record low power consumption and ultra-fast matrix multiplication. Such photonic accelerators, however, suffer from phase uncertainty due to fabrication errors and crosstalk effects that inhibit the development of high-density implementations. In this work, we present a Bayesian learning framework for such photonic accelerators. In addition to the conventional log-likelihood optimization path, two novel training schemes are derived, namely a regularized version and a fully Bayesian learning scheme. They are applied on a photonic neural network with 512 phase shifters targeting the MNIST dataset. The new schemes, when combined with a pre-characterization stage that provides the passive offsets, are able to dramatically decrease the operational power of the PIC beyond 70%, with just a slight loss in classification accuracy. The full Bayesian scheme, apart from this energy reduction, returns information with respect to the sensitivity of the phase shifters. This information is used to de-activate 31% of the phase actuators and, thus, significantly simplify the driving system.

Acoustic scene analysis using analog spiking neural network

Sensor nodes in a wireless sensor network for security surveillance applications should preferably be small, energy-efficient, and inexpensive with in-sensor computational abilities. An appropriate data processing scheme in the sensor node reduces the power dissipation of the transceiver through the compression of information to be communicated. This study attempted a simulation-based analysis of human footstep sound classification in natural surroundings using simple time-domain features. The spiking neural network (SNN), a computationally low-weight classifier derived from an artificial neural network (ANN), was used to classify acoustic sounds. The SNN and required feature extraction schemes are amenable to low-power subthreshold analog implementation. The results show that all analog implementations of the proposed SNN scheme achieve significant power savings over the digital implementation of the same computing scheme and other conventional digital architectures using frequency-domain feature extraction and ANN-based classification. The algorithm is tolerant of the impact of process variations, which are inevitable in analog design, owing to the approximate nature of the data processing involved in such applications. Although SNN provides low-power operation at the algorithm level, ANN to SNN conversion leads to an unavoidable loss of classification accuracy of ∼5%. We exploited the low-power operation of the analog processing SNN module by applying redundancy and majority voting, which improved the classification accuracy, taking it close to the ANN model.

A Hybrid Privacy-Preserving Deep Learning Approach for Object Classification in Very High-Resolution Satellite Images

Deep learning (DL) has shown outstanding performances in many fields, including remote sensing (RS). DL is turning into an essential tool for the RS research community. Recently, many cloud platforms have been developed to provide access to large-scale computing capacity, consequently permitting the usage of DL architectures as a service. However, this opened the door to new challenges associated with the privacy and security of data. The RS data used to train the DL algorithms have several privacy requirements. Some of them need a high level of confidentiality, such as satellite images related to public security with high spatial resolutions. Moreover, satellite images are usually protected by copyright, and the owner may strictly refuse to share them. Therefore, privacy-preserving deep learning (PPDL) techniques are a possible solution to this problem. PPDL enables training DL on encrypted data without revealing the original plaintext. This study proposes a hybrid PPDL approach for object classification for very-high-resolution satellite images. The proposed encryption scheme combines Paillier homomorphic encryption (PHE) and somewhat homomorphic encryption (SHE). This combination aims to enhance the encryption of satellite images while ensuring a good runtime and high object classification accuracy. The method proposed to encrypt images is maintained through the public keys of PHE and SHE. Experiments were conducted on real-world high-resolution satellite images acquired using the SPOT6 and SPOT7 satellites. Four different CNN architectures were considered, namely ResNet50, InceptionV3, DenseNet169, and MobileNetV2. The results showed that the loss in classification accuracy after applying the proposed encryption algorithm ranges from 2% to 3.5%, with the best validation accuracy on the encrypted dataset reaching 92%.

Accelerating 3D Convolutional Neural Network with Channel Bottleneck Module for EEG-Based Emotion Recognition.

Deep learning-based emotion recognition using EEG has received increasing attention in recent years. The existing studies on emotion recognition show great variability in their employed methods including the choice of deep learning approaches and the type of input features. Although deep learning models for EEG-based emotion recognition can deliver superior accuracy, it comes at the cost of high computational complexity. Here, we propose a novel 3D convolutional neural network with a channel bottleneck module (CNN-BN) model for EEG-based emotion recognition, with the aim of accelerating the CNN computation without a significant loss in classification accuracy. To this end, we constructed a 3D spatiotemporal representation of EEG signals as the input of our proposed model. Our CNN-BN model extracts spatiotemporal EEG features, which effectively utilize the spatial and temporal information in EEG. We evaluated the performance of the CNN-BN model in the valence and arousal classification tasks. Our proposed CNN-BN model achieved an average accuracy of 99.1% and 99.5% for valence and arousal, respectively, on the DEAP dataset, while significantly reducing the number of parameters by 93.08% and FLOPs by 94.94%. The CNN-BN model with fewer parameters based on 3D EEG spatiotemporal representation outperforms the state-of-the-art models. Our proposed CNN-BN model with a better parameter efficiency has excellent potential for accelerating CNN-based emotion recognition without losing classification performance.

SCANN: Synthesis of Compact and Accurate Neural Networks

Deep neural networks (DNNs) have become the driving force behind recent artificial intelligence (AI) research. With the help of a vast amount of training data, neural networks can perform better than traditional machine learning algorithms in many applications. An important problem with implementing a neural network is the design of its architecture. Typically, such an architecture is obtained manually by exploring its hyperparameter space and kept fixed during training. This approach is both time consuming and inefficient. Another issue is that modern neural networks often contain millions of parameters, whereas many applications require small inference models due to imposed resource constraints, such as energy constraints on battery-operated devices. However, efforts to migrate DNNs to such devices typically entail a significant loss of classification accuracy. To address these challenges, we propose a two-step neural network synthesis methodology, called DR+SCANN, that combines two complementary approaches to design compact and accurate DNNs. At the core of our framework is the SCANN methodology that uses three basic architecture-changing operations, namely, connection growth, neuron growth, and connection pruning, to synthesize feedforward architectures with arbitrary structure. These neural networks are not limited to the multilayer perceptron structure. SCANN encapsulates three synthesis methodologies that apply a repeated grow-and-prune paradigm to three architectural starting points. DR+SCANN combines the SCANN methodology with dataset dimensionality reduction to alleviate the curse of dimensionality. We demonstrate the efficacy of SCANN and DR+SCANN on various image and nonimage datasets. We evaluate SCANN on MNIST, CIFAR-10, and ImageNet benchmarks. Without any loss in accuracy, SCANN generates a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$46.3\times $ </tex-math></inline-formula> smaller network than the LeNet-5 Caffe model. We also compare SCANN-synthesized networks with a state-of-the-art fully connected (FC) feedforward model for MNIST, and show <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$20\times $ </tex-math></inline-formula> ( <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$19.9\times $ </tex-math></inline-formula> ) reduction in the number of parameters (floating-point operations) with little drop in accuracy. For the CIFAR-10 dataset, we target AlexNet and VGG-16 baseline architectures. SCANN reduces the number of parameters in AlexNet by <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$10.1\times $ </tex-math></inline-formula> without any drop in accuracy. It reduces the number of parameters in the FC layers of VGG-16 to only 2.5k while increasing accuracy by 1.05%. On the ImageNet dataset, for the VGG-16 and MobileNetV2 architectures, we reduce network parameters by <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$8.0\times $ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$1.3\times $ </tex-math></inline-formula> , respectively, with a similar or improved performance over their respective baselines. We also evaluate the efficacy of using dimensionality reduction alongside SCANN (DR+SCANN) on nine small-to-medium-size datasets. Using this methodology enables us to reduce the number of connections in the network by up to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$5078.7\times $ </tex-math></inline-formula> (geometric mean: <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$82.1\times $ </tex-math></inline-formula> ), with little to no drop in accuracy. On seven out of nine datasets, we show 0.41%–10.09% accuracy improvements over the FC baseline models. We also show that our synthesis methodology yields neural networks that are much better at navigating the accuracy versus energy efficiency space. This can enable neural network-based inference even on Internet-of-Things sensors.

A photosensor employing data-driven binning for ultrafast image recognition

Pixel binning is a technique, widely used in optical image acquisition and spectroscopy, in which adjacent detector elements of an image sensor are combined into larger pixels. This reduces the amount of data to be processed as well as the impact of noise, but comes at the cost of a loss of information. Here, we push the concept of binning to its limit by combining a large fraction of the sensor elements into a single “superpixel” that extends over the whole face of the chip. For a given pattern recognition task, its optimal shape is determined from training data using a machine learning algorithm. We demonstrate the classification of optically projected images from the MNIST dataset on a nanosecond timescale, with enhanced dynamic range and without loss of classification accuracy. Our concept is not limited to imaging alone but can also be applied in optical spectroscopy or other sensing applications.

A Neuromorphic Processing System With Spike-Driven SNN Processor for Wearable ECG Classification.

This paper presents a neuromorphic processing system with a spike-driven spiking neural network (SNN) processor design for always-on wearable electrocardiogram (ECG) classification. In the proposed system, the ECG signal is captured by level crossing (LC) sampling, achieving native temporal coding with single-bit data representation, which is directly fed into an SNN in an event-driven manner. A hardware-aware spatio-temporal backpropagation (STBP) is suggested as the training scheme to adapt to the LC-based data representation and to generate lightweight SNN models. Such a training scheme diminishes the firing rate of the network with little plenty of classification accuracy loss, thus reducing the switching activity of the circuits for low-power operation. A specialized SNN processor is designed with the spike-driven processing flow and hierarchical memory access scheme. Validated with field programmable gate arrays (FPGA) and evaluated in 40 nm CMOS technology for application-specific integrated circuit (ASIC) design, the SNN processor can achieve 98.22% classification accuracy on the MIT-BIH database for 5-category classification, with an energy efficiency of 0.75 μJ/classification.

Abstract 5159: Towards population-scale screening of human cancer using genome-wide fragmentation profiles of cell-free DNA

Abstract Introduction: Earlier detection is a critical clinical intervention to reduce cancer-related mortality. The DELFI liquid biopsy approach (DNA evaluation of fragments for early interception) utilizing low coverage (1-2x) whole genome sequencing (WGS) to analyze cell-free DNA (cfDNA) fragmentation provides a promising avenue for cancer detection. As sequencing costs remain a barrier to adoption of liquid biopsy approaches for early detection, we evaluated WGS for DELFI using 2-channel Illumina NovaSeq sequencing as a more affordable (~7-fold cost savings) alternative to 4-channel HiSeq instruments. Methods: We performed WGS on the prospectively collected LUCAS cohort of 365 individuals at risk for lung cancer using both HiSeq and NovaSeq platforms (Mathios et. al., Nature Communications 2021). Genome-wide fragmentation was summarized in non-overlapping 5 Mb bins by ratio of short (100-150 bp) to long (151-220 bp) fragments.To measure within-sequencer repeatability, we compared fragmentation profiles of non-cancer individuals to the median non-cancer fragmentation profile by Spearman correlation. Principal component analyses were performed to assess the extent to which the sequencer explains variation of fragmentation profiles across samples.For cancer prediction, we used a penalized logistic regression model with fragmentation profiles and other genome-wide characteristics as features. Machine learning performance was assessed by cross-validation and area under the receiver operator characteristic curve (AUC). To evaluate whether we could have developed the classifier from a combination of NovaSeq with HiSeq sequenced samples, we evaluated performance trained on 90:10%, 75:25%, 50:50%, 25:75%, and 10:90% HiSeq:NovaSeq mixtures, respectively. Results: cfDNA fragmentation profiles were highly concordant among non-cancer individuals for both platforms with median correlations of 0.96 (IQR: 0.95 - 0.97) and 0.95 (IQR: 0.94 - 0.96). Visualization of fragmentation principal components did not reveal separation by sequencing platform. The DELFI approach applied to samples sequenced by NovaSeq recapitulated previously published performance measures based on HiSeq (AUC 0.90, 95% CI 0.86 - 0.94). In simulations of mixed-platform datasets, we found the same qualitative performance (AUC range: 0.893-0.902). Conclusions: cfDNA fragmentation profiles were similar between HiSeq and NovaSeq platforms, and classification accuracies from machine learning models trained on these platforms were equivalent. Our results indicate HiSeq and NovaSeq sequenced samples can be combined in models with no discernible loss in classification accuracy provided a balance of non-cancers and cancers are sequenced on both platforms. The lower cost of NovaSeq sequencing may enable wider adoption of genome-wide fragmentation-based approaches for cancer detection. Citation Format: Akshaya V. Annapragada, Dimitrios Mathios, Stephen Cristiano, Jamie E. Medina, Vilmos Adleff, Noushin Niknafs, Jacob Carey, Nic Dracopoli, Peter Bach, Jillian Phallen, Victor E. Velculescu, Robert B. Scharpf. Towards population-scale screening of human cancer using genome-wide fragmentation profiles of cell-free DNA [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 5159.

Bioinspired random projections for robust, sparse classification

Inspired by the use of random projections in biological sensing systems, we present a new algorithm for processing data in classification problems. This is based on observations of the human brain and the fruit fly's olfactory system and involves randomly projecting data into a space of greatly increased dimension before applying a cap operation to truncate the smaller entries. This leads to an algorithm that achieves a sparse representation with minimal loss in classification accuracy and is also more robust in the sense that classification accuracy is improved when noise is added to the data. This is demonstrated with numerical experiments, which supplement theoretical results demonstrating that the resulting signal transform is continuous and invertible, in an appropriate sense.

Evaluation and Stratification for Chinese International Education Quality with Deep Learning Model.

In the process of human communication, language learning and communication play a fundamental, leading, broad, and long-lasting role. It serves as a link and a bridge between countries and peoples, allowing for greater understanding and camaraderie. The significance of Chinese in international commercial and cultural exchanges has been more obvious in the current era, as China's comprehensive strength continues to improve. Its cultural value and practical value have been continuously improved, the international community has an increasing demand for learning Chinese, and the cause of international Chinese language education has developed rapidly. China must strengthen the international dissemination of Chinese, so that the world can better understand and accept China, so that China can better integrate into the world. In this context, how to evaluate and stratify the quality of Chinese international education has become an important research topic. Relying on the hot deep learning technology in recent years, this work designs a neural network for evaluating the quality of international Chinese education. The content of this work is as follows: aiming at the serious defect of the current mainstream feature classification network that only uses the top-level features extracted by a single convolution layer to classify, which leads to the loss of classification accuracy. A multiscale feature pyramid fusion network is built in this paper, starting with the working mechanism of a convolutional neural network. It is capable of fully extracting and combining the representations of the network's shallow and deep outputs, based on first- and second-order characteristics of global and local discriminative region information. Second, the network structure has the bottleneck layer module and the batch normalization layer module, both of which are made up of varying numbers of 1 × 1 convolution kernels.

One Parameter Defense—Defending Against Data Inference Attacks via Differential Privacy

Machine learning models are vulnerable to data inference attacks, such as membership inference and model inversion attacks. In these types of breaches, an adversary attempts to infer a data record’s membership in a dataset or even reconstruct this data record using a confidence score vector predicted by the target model. However, most existing defense methods only protect against membership inference attacks. Methods that can combat both types of attacks require a new model to be trained, which may not be time-efficient. In this paper, we propose a differentially private defense method that handles both types of attacks in a time-efficient manner by tuning only one parameter, the privacy budget. The central idea is to modify and normalize the confidence score vectors with a differential privacy mechanism which preserves privacy and obscures membership and reconstructed data. Moreover, this method can guarantee the order of scores in the vector to avoid any loss in classification accuracy. The experimental results show the method to be an effective and timely defense against both membership inference and model inversion attacks with no reduction in accuracy.