A naukri.com initiative
ML News
Arxiv
6d
32
Image Credit: Arxiv
Variational quantum and neural quantum states algorithms for the linear complementarity problem
- Variational quantum algorithms (VQAs) and their classical counterparts, such as VQLS and VNLS, have been used to solve real-world problems on current noisy intermediate-scale quantum (NISQ) hardware.
- A novel application of VQLS and VNLS has been demonstrated in a minimum map Newton solver for a complementarity-based rigid body contact model.
- The results show that the VNLS accurately simulates the dynamics of rigid spherical bodies during collision events.
- These findings suggest that quantum and quantum-inspired linear algebra algorithms can be viable alternatives to standard linear algebra solvers for modeling certain physical systems.
Read Full Article
1 Like
Arxiv
6d
63
Image Credit: Arxiv
Orchestrating Agents and Data for Enterprise: A Blueprint Architecture for Compound AI
- Large language models (LLMs) are gaining interest in industry for their impressive capabilities.
- The adoption of LLMs presents challenges in integration, utilization of proprietary data and models, and meeting requirements.
- A shift towards compound AI systems, with a blueprint architecture for orchestrating agents and data in enterprise applications, is proposed.
- The architecture involves coordinating data and instructions among agents using streams, mapping models and APIs to agents, and utilizing proprietary data through a data registry.
Read Full Article
3 Likes
Arxiv
6d
180
Image Credit: Arxiv
External-Wrench Estimation for Aerial Robots Exploiting a Learned Model
- This paper presents an external wrench estimator for aerial robots.
- The estimator uses a hybrid dynamics model that combines a first-principles model and a neural network.
- The framework addresses the limitations of state-of-the-art model-based wrench observers.
- The proposed method improves wrench estimation by reducing the contributions from residual dynamics and focusing more on the external wrench.
Read Full Article
10 Likes
Arxiv
6d
0
Image Credit: Arxiv
Efficient measurement of neutral-atom qubits with matched filters
- Traditional approaches for high-fidelity measurement of neutral-atom qubits suffer from readout crosstalk in tightly spaced arrays.
- A new study presents two simpler and scalable machine learning algorithms for qubit readout problem.
- The algorithms use matched filters and demonstrate error reduction of up to 32% and 43% for site and array models.
- The proposed approaches require fewer parameters and computational resources compared to conventional methods.
Read Full Article
Like
Arxiv
6d
32
Image Credit: Arxiv
A Piecewise Lyapunov Analysis of sub--quadratic SGD: Applications to Robust and Quantile Regression
- Motivated by robust and quantile regression problems, a study investigates the stochastic gradient descent (SGD) algorithm for minimizing an objective function with a sub--quadratic tail.
- The study introduces a novel piecewise Lyapunov function that can handle functions with only first-order differentiability, including popular loss functions such as Huber loss.
- Finite-time moment bounds are derived for general diminishing stepsizes and constant stepsizes, and weak convergence, central limit theorem, and bias characterization are established for constant stepsize.
- The results have wide applications, specifically in online robust regression and online quantile regression.
Read Full Article
1 Like
Arxiv
6d
135
Image Credit: Arxiv
Particle Hit Clustering and Identification Using Point Set Transformers in Liquid Argon Time Projection Chambers
- Liquid argon time projection chambers are often used in neutrino physics and dark-matter searches because of their high spatial resolution.
- Traditional machine learning methods such as convolutional neural networks (CNNs) cannot operate directly on the sparse matrix representation of the detector data.
- A machine learning model using a point set neural network is proposed, which greatly improves processing speed and accuracy over methods that instantiate the dense matrix.
- Compared to competing methods, the proposed model improves classification and segmentation performance while significantly reducing time and memory requirements.
Read Full Article
8 Likes
Arxiv
6d
185
Image Credit: Arxiv
Neural Encoding and Decoding at Scale
- Recent work has focused on predicting neural activity from behavior (encoding) or predicting behavior from neural activity (decoding).
- A new multimodal, multi-task model called NEDS enables simultaneous Neural Encoding and Decoding at Scale.
- NEDS achieves state-of-the-art performance for both encoding and decoding when pretrained on multi-animal data.
- NEDS's learned embeddings are highly predictive of the brain regions in each recording without explicit training.
Read Full Article
11 Likes
Arxiv
6d
209
Image Credit: Arxiv
DRAFT-ing Architectural Design Decisions using LLMs
- Architectural Knowledge Management (AKM) is a challenge in software development due to the lack of standardization and manual effort involved.
- Architecture Decision Records (ADRs) offer a structured approach to capture Architecture Design Decisions (ADDs), but their adoption is limited due to manual effort and insufficient tool support.
- Research has explored using Large Language Models (LLMs) to generate ADDs, but there is a need to enhance the quality of generated ADDs.
- The proposed approach, called DRAFT, combines few-shot, retrieval-augmented generation (RAG) and fine-tuning techniques to generate more effective ADDs while addressing privacy and resource challenges.
Read Full Article
12 Likes
Arxiv
6d
312
Image Credit: Arxiv
Optimizing Power Grid Topologies with Reinforcement Learning: A Survey of Methods and Challenges
- Reinforcement Learning (RL) is a promising approach to power network control (PNC) for optimizing power grid topologies.
- The Learning To Run a Power Network (L2RPN) competitions have accelerated research in RL-based methods for power grid optimization.
- This survey provides a comprehensive overview of RL applications for power grid topology optimization, categorizing existing techniques and highlighting key design choices.
- The survey also presents a comparative study evaluating the practical effectiveness of commonly applied RL-based methods and identifies open research challenges.
Read Full Article
18 Likes
Arxiv
6d
168
Image Credit: Arxiv
Bringing Structure to Naturalness: On the Naturalness of ASTs
- Source code comes in different shapes and forms, and previous research has shown code to be more predictable than natural language.
- The structure of code has been successfully used to improve the state-of-the-art on numerous tasks, such as code suggestion and code summarization.
- The Structured Naturalness Hypothesis suggests that a structured view of code is also natural, and evidence is provided in the case of trees.
- Naturalness signals in code can be employed for near state-of-the-art results on defect prediction, reducing the need for manual feature engineering.
Read Full Article
10 Likes
Arxiv
6d
304
Image Credit: Arxiv
Spectral Normalization for Lipschitz-Constrained Policies on Learning Humanoid Locomotion
- Reinforcement learning has shown potential in training legged robots for complex locomotion behaviors.
- Policies trained in simulation struggle to transfer to real-world robots due to unrealistic assumptions.
- Traditional methods penalize aggressive motions, but require extensive tuning.
- This work proposes Spectral Normalization as an efficient method to enforce Lipschitz continuity and reduce memory usage.
Read Full Article
18 Likes
Arxiv
6d
316
Image Credit: Arxiv
Neural Network-assisted Interval Reachability for Systems with Control Barrier Function-Based Safe Controllers
- Researchers propose a computationally efficient interval reachability method for performance verification of systems with optimization-based controllers.
- The method involves approximating the optimization-based controller using a pre-trained neural network and leveraging state-of-the-art neural network verification algorithms.
- The approach provides an over-approximation of the reachable set of the system with optimization-based controllers.
- Numerical results are presented to support the technical findings.
Read Full Article
19 Likes
Arxiv
6d
12
Image Credit: Arxiv
Understanding the Impact of Data Domain Extraction on Synthetic Data Privacy
- Privacy attacks, specifically membership inference attacks (MIAs), are commonly used to evaluate the privacy of generative models for tabular synthetic data.
- This paper highlights the importance of data domain extraction in generative models and its impact on privacy attacks.
- Three strategies for defining the data domain are examined: using an externally provided domain, extracting it directly from the input data, and extracting it with differential privacy (DP) mechanisms.
- The study shows that using the second approach of extracting the data domain directly from the input data can compromise end-to-end DP guarantees and make models vulnerable to privacy attacks.
Read Full Article
Like
Arxiv
6d
382
Image Credit: Arxiv
ELSA: A Style Aligned Dataset for Emotionally Intelligent Language Generation
- Advancements in emotion aware language processing increasingly shape vital NLP applications.
- ELSA Emotion and Language Style Alignment Dataset (ELSA Dataset) is introduced to bridge the gap between emotional granularity and style diversity.
- The dataset comprises emotionally nuanced variations of original sentences regenerated across different contextual styles.
- The dataset's emotional authenticity, linguistic fluency, and textual diversity are validated through rigorous computational evaluation.
Read Full Article
23 Likes
Arxiv
6d
312
Image Credit: Arxiv
MedRep: Medical Concept Representation for General Electronic Health Record Foundation Models
- Researchers propose MedRep, a solution for electronic health record (EHR) foundation models.
- MedRep addresses the limitation of processing unseen medical codes out of the vocabulary.
- It provides integrated medical concept representations and data augmentation strategies for patient trajectories.
- EHR foundation models trained with MedRep show improved prediction performance in external datasets.
Read Full Article
18 Likes
For uninterrupted reading, download the app