Computer Science and Automation (CSA)
Recent Submissions
-
A framework for timing analysis of event-driven applications
Event-driven applications, particularly those based on the publish–subscribe communication model are widely adopted to build responsive and decoupled applications in domains such as robotics, the Internet of Things ... -
Multi-timescale and Multi-agent Reinforcement Learning Algorithms
This thesis presents six novel works involving several research domains, such as reinforcement learning (RL)– both with or without function approximators including deep neural networks, multi-agent RL, stochastic optimization, ... -
CHARGE: Accelerating GNN Training via CPU Sampling in Heterogeneous CPU–GPU Environment
Graph Neural Networks (GNNs) have demonstrated exceptional performance across a wide range of applications, driving their widespread adoption. Current frameworks employ CPU and GPU resources—either in isolation or ... -
On Orbits and Border of Constant Read Circuits and Lower Bounds for Constant Depth Circuits
Two of the most common ways in which arithmetic circuits can be restricted is by requiring that they be constant read or constant depth. Constant depth circuits have received a lot of attention in the arithmetic circuit ... -
Design of AI-based Computational Framework for Accurate Detection of Polycystic Ovarian Disease and Ovarian Cancer Using Ultrasound, CT and Histopathology Images
Polycystic Ovarian Disease (PCOD) and Ovarian Cancer (OC) represent critical health challenges affecting millions of women globally. Early and accurate detection of these conditions is essential for effective clinical ... -
Controller Synthesis Techniques for Infinite-State Reactive Systems
Reactive systems are ubiquitous, with various use cases ranging from controllers for cyber-physical systems to programs that run on multipurpose computers. A system is reactive if it continually behaves in a specific manner ... -
IEDFuRL: A Black-box Fuzz Tester for IEC61850-based Intelligent Electronic Devices using Reinforcement Learning
Intelligent Electronic Devices (IEDs) are essential components of modern power grids, functioning as microprocessor-based controllers that facilitate communication, monitoring, protection, and control within Supervisory ... -
Mitigating Bias via Algorithms with Fairness Guarantees
The rapid integration of automated decision-making systems in critical domains such as resume screening, loan approval, content recommendation, and disaster containment has raised significant concerns regarding biases in ... -
Reasoning about Safety of Camera-Based Neural Network Controlled Systems
Autonomous technologies are becoming increasingly prevalent due to the numerous benefits they offer, including improved safety and security, enhanced accessibility for mobility-challenged individuals, and the ability to ... -
Algorithms for various cost criteria in Reinforcement Learning
In this thesis we will look at various Reinforcement Learning algorithms. We will look at algorithms for various cost criteria or reward objectives namely Finite Horizon, Discounted Cost, Risk-Sensitive Cost. For Finite ... -
Combinatorial Problems arising in Quantum Physics and Model counting
This thesis investigates some combinatorial problems arising in quantum physics and model counting. In particular, we develop new graph-theoretic techniques to resolve some open questions from quantum photonics. We also ... -
Feature Tracking and Visual Analysis of Temporal Scalar Fields in the Ocean
The Bay of Bengal (BoB) has maintained its salinity distribution over the years despite a continuous flow of fresh water entering it through rivers on the northern coast, which can dilute its salinity. This can be ... -
Loop Transformations for Multi-/Many-Core Architectures using Machine Learning
Loop transformation techniques such as loop tiling, loop interchange and unroll-and-jam help expose better coarse-grain and fine-grain data-level parallelisms as well as exploit data locality. These transformations are ... -
Fair and Efficient Dynamic Memory De-bloating
The virtual memory abstraction simplifies programming and enhances portability but requires the processor to translate virtual addresses to physical addresses which can be expensive. To speed up the virtual-to-physical ... -
PyGraph: Compiler Support for Efficient and Transparent Use of CUDA Graphs
CUDA Graphs --- a recent hardware feature introduced for NVIDIA GPUs --- aim to reduce CPU launch overhead by capturing and launching a series of GPU tasks (kernels) as a DAG. However, deploying CUDA Graphs faces several ... -
Advancing the Communication Complexity Landscape of Perfectly Secure Multiparty Computation
Secure multiparty computation (MPC) allows n distrustful parties to jointly compute a function on their inputs while keeping their inputs private. The distrust is modelled as an adversary that controls up to t parties and ... -
Zero-Knowledge Proofs with Enhanced Deniability
In cryptography, deniability is a crucial concept that allows a participant to plausibly deny taking part in executing a scheme or protocol. Non-interactive zero-knowledge (NIZK) proofs enable a party (the prover) to ... -
Ankora: Notions of Multi-party Computation and Zero-knowledge Beyond Conventional Models
In the era of digitalization, data privacy and integrity are of utmost importance. Secure multiparty computation (MPC) and zero-knowledge (ZK) facilitate data privacy and integrity. MPC enables privacy-preserving collaborative ... -
Single and Multi-Agent Finite Horizon Reinforcement Learning Algorithms for Smart Grids
In this thesis, we study sequential decision-making under uncertainty in the context of smart grids using reinforcement learning. The underlying mathematical model for reinforcement learning algorithms are Markov Decision ... -
Analysis, Design, and Acceleration of DNN Training Preprocessing Pipelines
The performance of deep neural network (DNN) training is a function of both the training compute latency and the latency to fetch and preprocess the input data needed to train the model. With advance- ments in GPUs and ...
RSS 1.0