A5102792137- Robot Manipulation and Learning
- Reinforcement Learning in Robotics
- Recommender Systems and Techniques
- Semiconductor materials and devices
- Advancements in Semiconductor Devices and Circuit Design
- Internet of Things and AI
- Web Data Mining and Analysis
- Advanced Bandit Algorithms Research
- Robotic Path Planning Algorithms
- Data Management and Algorithms
- Machine Learning and Algorithms
- Adversarial Robustness in Machine Learning
- Topic Modeling
- Advanced Wireless Communication Technologies
- Nanowire Synthesis and Applications
- Data Stream Mining Techniques
- PAPR reduction in OFDM
- Multimodal Machine Learning Applications
- Teleoperation and Haptic Systems
- Network Security and Intrusion Detection
- Photovoltaic System Optimization Techniques
- Domain Adaptation and Few-Shot Learning
- Advanced Photonic Communication Systems
- Semantic Web and Ontologies
- Robotics and Sensor-Based Localization
Carnegie Mellon University
2018-2024
Birla Institute of Technology and Science, Pilani
2024
Medical University of Warsaw
2021-2024
Vivekananda Global University
2022-2024
Institute of Technology Management
2024
Technology Innovation Institute
2023-2024
Malopolska Higher Vocational School of J. Dietl in Kraków
2022-2024
Galgotias University
2022-2023
Jain University
2023
Teerthanker Mahaveer University
2023
In recent years, there has been a substantial rise in the consumption of news via online platforms. The ease publication and lack editorial rigour some these platforms have further led to proliferation fake news. this paper, we study problem detecting on FakeNewsNet repository, collection full length articles along with associated images. We present SpotFake+, multimodal approach that leverages transfer learning capture semantic contextual information from its images achieves better accuracy...
Large, high-capacity models trained on diverse datasets have shown remarkable successes efficiently tackling downstream applications. In domains from NLP to Computer Vision, this has led a consolidation of pretrained models, with general backbones serving as starting point for many Can such happen in robotics? Conventionally, robotic learning methods train separate model every application, robot, and even environment. we instead generalist X-robot policy that can be adapted new robots,...
Understanding the search tasks and behavior of users is necessary for optimizing engine results. While much work has been done on understanding in Web search, little knowledge available about E-Commerce (E-Com) applications. In this paper, we share first empirical study queries E-Com by analyzing log from a major engine. The analysis results show that can be categorized into five categories, each with distinctive behaviors: (1) Shallow Exploration Queries are short vague user may use...
We introduce STRING: Separable Translationally Invariant Position Encodings. STRING extends Rotary Encodings, a recently proposed and widely used algorithm in large language models, via unifying theoretical framework. Importantly, still provides exact translation invariance, including token coordinates of arbitrary dimensionality, whilst maintaining low computational footprint. These properties are especially important robotics, where efficient 3D representation is key. integrate into Vision...
Visuomotor policies trained via imitation learning are capable of performing challenging manipulation tasks, but often extremely brittle to lighting, visual distractors, and object locations. These vulnerabilities can depend unpredictably on the specifics training, expose without time-consuming expensive hardware evaluations. We propose problem predictive red teaming: discovering a policy with respect environmental factors, predicting corresponding performance degradation evaluations in...
The use of imitation learning to learn a single policy for complex task that has multiple modes or hierarchical structure can be challenging. In fact, previous work shown when the are known, separate policies each mode sub-task greatly improve performance learning. this work, we discover interaction between sub-tasks from their resulting state-action trajectory sequences using directed graphical model. We propose new algorithm based on generative adversarial framework which automatically...
In this paper, we present IVA-HD, a true multistandard, programmable, full HD video coding engine which adopts optimal hardware-software partitioning to achieve the low-power and area requirements of OMAP 4 processor. Unlike approach using separate IPs for encoder decoder, IVA-HD uses an integrated codec is efficient, as most decoder logic reused encoder. architected perform stream-rate pixel- rate processing in single pipeline (that processes one 16x16 macroblock at time), so support...
Recommending new items to existing users has remained a challenging problem due absence of user's past preferences for these items. The user personalized non-collaborative methods based on item features can be used address this cold-start problem. These rely similarities between the target and previous preferred While computing features, overlook interactions among consider them independently. Modeling helpful as some when considered together, provide stronger signal relevance an compared...
Software Development Life Cycle (SDLC) is an important concept used in software engineering to describe a procedure for planning, creating, coding, testing and implementation of user requirement specification. development life cycle applies range hardware configurations. SDLC step by process creating quality users. It involves different phases that are followed one after one, essential engineers such as analysis, design, implementation. In the early years, was costly relatively cheap....
Cutting is a common form of manipulation when working with divisible objects such as food, rope, or clay. Cooking in particular relies heavily on cutting to divide food items into desired shapes. However, challenging task due the wide range material properties exhibited by items. Due this variability, same motions cannot be used for all Sensations from contact events, e.g., placing knife item, will also vary depending properties, and robot need adapt accordingly. In paper, we propose using...
Robots deployed in many real-world settings need to be able acquire new skills and solve tasks over time. Prior works on planning with often make assumptions the structure of tasks, such as subgoal skills, shared skill implementations, or task-specific plan skeletons, which limit adaptation tasks. By contrast, we propose doing task by jointly searching space parameterized using high-level effect models learned simulation. We use an iterative training procedure efficiently generate relevant...
Non-Orthogonal multiple access (NOMA) and Cognitive radio (Cr) are seen as one of the most promising techniques, which improves utilization spectrum in 5G. The expanding number wireless applications like new gadgets, IOT brought about developing a block ISM groups. FCC requested to permit unlicensed clients work void area without obstruction an authorized guest. Cr gives answer for extra range prerequisite issue productive usage. foremost condition permitting CRs utilize is not causing...
The main characteristic of the next-generation wire-less communications are seamless connectivity, large bandwidth, and high spectral energy efficiencies. To acheive these objectives, reconfigurable intelligent surface (RIS), terahertz (THz) communication, user pairing nonorthogonal multiple access (NOMA) have been recognized as key enabling technologies. In this paper, we propose a RIS-assisted THz-NOMA system, which is referred to RTHz-NOMA. performed based on distance between users power...
Non-orthogonal multiple access (NOMA) is gaining considerable attention due to its features, such as low out-of-band radiation, signal detection capability, high spectrum gain, fast data rate, and massive D2D connectivity. It may be considered for 5G networks. However, the peak-to-average power ratio (PAPR) viewed a significant disadvantage of NOMA waveform, it weakens quality signals throughput scheme. In this article, we introduce modified system by employing block wavelet transform, an...
As the role of remote vehicle operation gains prominence, ensuring robust wireless communication systems is crucial. Prior studies have delved into various aspects teleoperation, including its legal framework and workforce implications. Some even formulated models to estimate number operators for large automated fleets. These suggest that vehicles overseen by could supplant a significant portion current driving jobs in United States. Such findings underscore need governmental review policy...
Abstract: Traffic sign recognition is a crucial component in advanced driver assistance systems and autonomous vehicles, enhancing road safety overall transportation efficiency. Deep learning, specifically convolutional neural networks (CNNs), has emerged as powerful tool for image based tasks. The proposed deep learning-based traffic system exhibits promising results, providing foundation the development of intelligent systems. trained model demonstrates remarkable accuracy recognizing wide...
Recommender systems are widely used to recommend the most appealing items users. These recommendations can be generated by applying collaborative filtering methods. The low-rank matrix completion method is state-of-the-art method. In this work, we show that skewed distribution of ratings in user-item rating real-world datasets affects accuracy matrix-completion-based approaches. Also, number an item or a user has positively correlates with ability approaches predict for accurately....
Reliability and sustainability of power supply between already existing network Microgrid (MG) having DGs is ensured by both the grid connected islanded mode operations. The selection operation a MG based on technical economic factors. intentional islanding depends prevailing operating condition MG. These conditions are sensed at point common coupling (PCC) accordingly switching to performed. This paper presents detection evaluation PCC, Islanded need. have been categorized for decision...
In this work, we present an interaction-based approach to learn semantically rich representations for the task of slicing vegetables. Unlike previous approaches, focus on object-centric and use auxiliary tasks using a two-step process. First, simple tasks, such as predicting thickness cut slice, embedding space which captures object properties that are important second step, these learned latent embeddings forward model. Learning model affords us plan online in forces our improve its while...
E-commerce search engines can fail to retrieve results that satisfy a query's product intent because: (i) conventional retrieval approaches, such as BM25, may ignore the important terms in queries owing their low "inverse document frequency" " (IDF), and (ii) for long queries, is usually case rare (i.e., tail queries), they determine relevant are representative of intent. In this paper, we leverage historical query reformulation logs large e-retailer (walmart.com) develop...