A naukri.com initiative
ML News
Arxiv
1d
289
Image Credit: Arxiv
AI in Lung Health: Benchmarking Detection and Diagnostic Models Across Multiple CT Scan Datasets
- Lung cancer remains the leading cause of cancer-related mortality worldwide.
- AI models are being integrated into medical imaging for the early detection of lung cancer.
- Duke Lung Cancer Screening (DLCS) Dataset, a large open-access dataset with over 2,000 scans and 3,000 expert-verified nodules, is introduced.
- Benchmarking of deep learning models for nodule detection and lung cancer classification is conducted across multiple datasets, demonstrating strong generalizability.
Read Full Article
17 Likes
Arxiv
1d
63
Image Credit: Arxiv
RACH Traffic Prediction in Massive Machine Type Communications
- Traffic pattern prediction in massive machine-type communication (mMTC) networks is challenging due to the inherent randomness of events and bursty traffic.
- A machine learning-based framework using long-term short-term memory (LSTM) and DenseNet with feed-forward neural network (FFNN) layers is proposed for forecasting bursty traffic in multi-channel slotted ALOHA networks.
- The framework includes a low-complexity online prediction algorithm that updates the states of the LSTM network using frequently collected data from the mMTC network.
- Simulation results show that the proposed framework achieves a 52% higher accuracy in long-term predictions compared to traditional methods, without imposing additional processing load.
Read Full Article
3 Likes
Arxiv
1d
323
Image Credit: Arxiv
Discrete Cosine Transform Based Decorrelated Attention for Vision Transformers
- Researchers propose a discrete cosine transform (DCT)-based approach for initializing and compressing the attention mechanism in Vision Transformers.
- The DCT-based attention initialization method offers improved accuracy in classification tasks for Vision Transformers.
- DCT effectively decorrelates image information in the frequency domain, enabling the compression of higher-frequency components.
- The DCT-based compression technique reduces the size of weight matrices for queries, keys, and values, resulting in decreased computational overhead.
Read Full Article
19 Likes
Arxiv
1d
218
Image Credit: Arxiv
Siren -- Advancing Cybersecurity through Deception and Adaptive Analysis
- Siren is a cybersecurity project that integrates deception, machine learning, and proactive threat analysis.
- The system lures potential threats into controlled environments, employing a dynamic machine learning model for real-time analysis and classification.
- It includes a link monitoring proxy, a purpose-built machine learning model for dynamic link analysis, and a honeypot with simulated user interactions.
- Siren transforms traditional defense mechanisms into proactive systems that actively engage and learn from potential adversaries.
Read Full Article
13 Likes
Arxiv
1d
259
Image Credit: Arxiv
RSEND: Retinex-based Squeeze and Excitation Network with Dark Region Detection for Efficient Low Light Image Enhancement
- A new Retinex-based Squeeze and Excitation Network with Dark Region Detection (RSEND) is proposed for efficient low light image enhancement.
- RSEND divides the low-light image into the illumination map and reflectance map, then captures important details in the illumination map for light enhancement.
- The framework utilizes Squeeze and Excitation network to better capture details and performs well in complicated datasets like LOL-v2.
- RSEND outperforms other CNN-based models and achieves significant improvement in peak signal-to-noise ratio (PSNR) in different datasets.
Read Full Article
15 Likes
Arxiv
1d
240
Image Credit: Arxiv
ReaL: Efficient RLHF Training of Large Language Models with Parameter Reallocation
- Reinforcement Learning from Human Feedback (RLHF) is a pivotal technique for empowering large language model (LLM) applications.
- The RLHF training process for LLMs requires sophisticated parallelization strategies to improve training efficiency.
- To address this, a novel technique called parameter ReaLlocation is proposed, which dynamically adapts parallelization strategies during training.
- The ReaL system achieves significant speedups and performance improvement compared to baseline methods for RLHF training.
Read Full Article
14 Likes
Arxiv
1d
169
Image Credit: Arxiv
DDU-Net: A Domain Decomposition-Based CNN for High-Resolution Image Segmentation on Multiple GPUs
- A novel approach, DDU-Net, is proposed for high-resolution image segmentation.
- DDU-Net combines encoder-decoder architectures with domain decomposition.
- It partitions input images into non-overlapping patches for independent processing.
- Experimental results show improved performance in segmenting ultra-high-resolution images.
Read Full Article
10 Likes
Arxiv
1d
338
Image Credit: Arxiv
Set2Seq Transformer: Temporal and Positional-Aware Set Representations for Sequential Multiple-Instance Learning
- Sequential multiple-instance learning involves learning representations of sets distributed across discrete timesteps.
- Existing methods either focus on learning set representations at a static level, ignoring temporal dynamics, or treat sequences as ordered lists of individual elements, lacking explicit mechanisms to represent sets.
- Set2Seq Transformer is a novel architecture that jointly models permutation-invariant set structure and temporal dependencies by learning temporal and positional-aware representations of sets within a sequence in an end-to-end multimodal manner.
- The Set2Seq Transformer significantly improves over traditional static multiple-instance learning methods by effectively learning permutation-invariant set, temporal, and positional-aware representations.
Read Full Article
20 Likes
Arxiv
1d
251
Image Credit: Arxiv
A causal viewpoint on prediction model performance under changes in case-mix: discrimination and calibration respond differently for prognosis and diagnosis predictions
- Prediction models need reliable performance for diagnosis, prognosis, and treatment planning.
- Changes in case-mix can impact model performance.
- Discrimination and calibration respond differently depending on the causal direction of the prediction task.
- Understanding the causal structure of the prediction task is important for developing and evaluating prediction models in different clinical settings.
Read Full Article
15 Likes
Arxiv
1d
199
Image Credit: Arxiv
Feature-to-Image Data Augmentation: Improving Model Feature Extraction with Cluster-Guided Synthetic Samples
- The study introduces FICAug, a feature-to-image data augmentation framework designed to improve model generalization under limited data conditions.
- FICAug operates in the feature space, where original data are clustered and synthetic data is generated through Gaussian sampling.
- These synthetic features are then projected back into the image domain using a generative neural network.
- Experimental results demonstrate that FICAug significantly improves classification accuracy, achieving a cross-validation accuracy of 84.09% in the feature space and 88.63% when training a ResNet-18 model on the reconstructed images.
Read Full Article
11 Likes
Arxiv
1d
334
Image Credit: Arxiv
Convergence of Diffusion Models Under the Manifold Hypothesis in High-Dimensions
- Denoising Diffusion Probabilistic Models (DDPM) are powerful state-of-the-art methods used to generate synthetic data from high-dimensional data distributions.
- This work studies DDPMs under the manifold hypothesis and proves that they achieve rates independent of the ambient dimension in terms of score learning.
- In terms of sampling complexity, the rates achieved by DDPMs are independent of the ambient dimension w.r.t. the Kullback-Leibler divergence and O(sqrt(D)) w.r.t. the Wasserstein distance.
- A new framework connecting diffusion models to the theory of extrema of Gaussian Processes is developed.
Read Full Article
20 Likes
Arxiv
1d
300
Image Credit: Arxiv
Selective Attention Improves Transformer
- Unneeded elements in the attention's context degrade performance.
- Selective Attention is introduced as a simple parameter-free change to the standard attention mechanism.
- Selective Attention consistently improves language modeling and downstream task performance.
- Selective Attention allows for meaningful reductions in memory and compute requirements during inference.
Read Full Article
18 Likes
Arxiv
1d
78
Image Credit: Arxiv
Parameter-Efficient Fine-Tuning in Large Models: A Survey of Methodologies
- Large models have made significant progress in natural language generation tasks, but their parameter scale poses challenges in fine-tuning.
- Parameter-Efficient Fine-Tuning (PEFT) offers a solution to efficiently adjust parameters of large pre-trained models for specific tasks.
- PEFT minimizes the introduction of additional parameters and reduces the computational resources required.
- This review provides an overview of PEFT, including its core principles, applications, and future research directions.
Read Full Article
4 Likes
Arxiv
1d
75
Image Credit: Arxiv
Improved implicit diffusion model with knowledge distillation to estimate the spatial distribution density of carbon stock in remote sensing imagery
- A study proposes an improved implicit diffusion model (IIDM) for estimating the spatial distribution density of carbon stock in remote sensing imagery.
- The study utilizes GF-1 WFV satellite imagery to estimate carbon stock in Huize County, Qujing City, Yunnan Province, China.
- The IIDM model demonstrates the highest estimation accuracy with an RMSE of 12.17%, outperforming other models by 41.69% to 42.33%.
- The 16-meter resolution estimates can enhance regional carbon stock management and inform forest carbon sink regulations.
Read Full Article
4 Likes
Arxiv
1d
202
Image Credit: Arxiv
Machine Learning-Based Automated Assessment of Intracorporeal Suturing in Laparoscopic Fundoplication
- Automated assessment of surgical skills using artificial intelligence (AI) provides trainees with instantaneous feedback.
- Researchers developed an AI-based tool tracking model to eliminate the need for human annotators in laparoscopic suturing tasks.
- The study evaluated the usefulness of the tool tracking model in automated assessment during a laparoscopic suturing task in the fundoplication procedure.
- The AI model achieved an accuracy of 0.817 and an F1 score of 0.806, demonstrating its capability of automated performance classification.
Read Full Article
12 Likes
For uninterrupted reading, download the app