Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 20 of 43 for “"Quantization Techniques"”.
-
Efficient quantization techniques for Large Language Models
Quantization has emerged as a critical technique in optimizing neural networks for efficient deployment across various hardware platforms. As Large Language Models (LLMs), which contain billions of parameters, continue to grow, quantization significantly reduces computational requirements, memory …
-
Advanced Quantization Techniques for Communication Efficiency and Privacy in Federated Learning, and Memory-efficient Fine-tuning of LLMs
Quantization techniques play a crucial role in developing communication-efficient Federated Learning (FL) algorithms, as well as in optimizing training for Large Language Models (LLMs). While it is evident that quantization can significantly enhance the efficiency of modern machine learning systems …
-
Vector Quantization of Deep Convolutional Neural Networks with Learned Codebook
… focus on compressing deep CNNs based on vector quantization techniques. The first part of this thesis summarizes some basic concepts in machine learning and popular techniques on model compression, including pruning, quantization, low-rank factorization and knowledge distillation approaches. Our …
-
Advanced wireless communications using large numbers of transmit antennas and receive nodes
… training phase. We then discuss efficient CSI quantization techniques using a trellis search. The proposed CSI quantization techniques can be implemented with a complexity that only grows linearly with the number of transmit antennas while the performance is close to the optimal case. We also …
-
Asymptotic analysis of digital transmission systems for a first order Gauss-Markov process
… optimum sampling rate and the optimum number of quantization levels. The performance of different reconstruction filters is first examined. Optimal nonlinear reconstruction filters are approximated by linear filters. Fine quantization techniques are used to analyze the performance of different …
-
Energy and nature based split multiple transform domain split vector quantization for speech coding
Recently vector quantization techniques in multiple transform domains for waveform signal characterization that improved signal-coding performance have been reported. Windowed portions of the signal are transformed into vectors that are encoded in multiple domains. The decoder selects the transform …
-
Wavelet Based Approach to 3D Geometry Point Cloud Compression
… proposed here hopes to improve on existing quantization techniques by segmenting each coordinate axis into continuous regions of arbitrary size and preforming a wavelet transformation on these regions. This Thesis describes a means of finding these regions, taking the transform, quantizing, …
-
Efficient Intelligence Towards Real-Time Precision Medicine With Systematic Pruning and Quantization
… investigates the application of pruning and quantization techniques to optimize CNNs for arrhythmia classification using the MIT-BIH Arrhythmia Database. By combining magnitude-based pruning, regularization-based pruning, filter map-based pruning, and quantization at different bit-widths …
-
Optimal chaos control through reinforcement learning
… of the reinforcement learning problem, vector quantization techniques are applied and the control policy is approximated in the reduced space. To find minimal sufficient codebooks, we suggest a fixed-point sensitive growing chaos controller which combines a modification of Fritzke's Growing …
-
Application of Outlier Detection for Volvo Trucks Safety Scoring: Classification of drivers based on driving behaviour and assign safety score using unsupervised learning and object detection.
… advanced data analysis and machine learning techniques. This thesis explores the application of outlier detection methods to evaluate and improve safety scoring of Volvo truck drivers based on the driver’s behavior, braking and acceleration patterns, and contextual traffic conditions. The …
-
Lossy compression of scientific data via wavelets and vector quantization
… capabilities for archiving. Lossy compression techniques can be used to compress the data, although by definition the original data cannot be recovered. The compression techniques used must ensure that the compressed data have both high visual fidelity and that calculations using the compressed …
-
Linear Precoding in Wireless Networks with Channel State Information Feedback
… In the first half of this thesis, memoryless quantization codebooks are designed and incremental vector quantization techniques are developed for the representation of CSI in MIMO point-to-point links and isolated (single-cell) downlink networks. The second half of the thesis seeks to design …
-
Hamiltonian magnetic monopoles and an integral transform for kinetic plasmas
… can never coincide or we must resort to new quantization techniques. The linearized Vlasov-Poisson system can be exactly solved using the G-transform, an integral transform that removes the electric field term. We apply the G-transform to several applications. By investigating the interaction …
-
Optimization of vector quantization in Hybrid Vector Scalar Quantization (HVSQ)
… compression schemes featuring new concepts and techniques have been proposed to yield superior compression quality. Hybrid Vector Scalar Quantization (HVSQ) is one such novel compression scheme used for compressing high resolution cervigram image archives of the US National Library of Medicine. …
-
Quantization of black holes and singularity resolution in loop quantum gravity.
… Loop quantum gravity is based on the canonical quantization of holonomies and fluxes of densitized triads. In loop quantum cosmology (LQC), the effective Hamiltonian can be obtained from the classical Hamiltonian by polymerization. The interior of Schwarzschild black hole is isometric to …
-
Energy-Efficient Hardware Acceleration of Transformer-Based Models
… transformer models on FPGAs. It introduces novel quantization techniques tailored for both BERT and generative LLMs. For weight quantization, the proposed Hessian-based parameter-wise method assigns optimal bit precision based on sensitivity analysis, while the proposed row-wise quantization …
-
Resistivity anomalies for ferromagnetic metals at Curie points
… the spin correlation function. Using the second quantization techniques of field theory, the correlation function is found. Thence, the scattering cross section is obtained. Based on transport theory, the relaxation time and the resistivity are calculated for both the paramagnetic and the …
-
Towards Green AI: Evaluating Energy Efficiency for Quantized LLM Inference
While it is known that quantization of large language models (LLMs) reduces memory usage via lower-bit weights, studies quantifying the resulting impact on energy usage and carbon intensity are scarce. We present the first unified evaluation of weight-only quantization strategies for LLM inference …
-
A Talking Cart
… strategies for edge deployment, incorporating quantization techniques to ensure computational efficiency without compromising accuracy. The system's performance was validated using metrics such as mAP and latency, showcasing its ability to operate effectively in real-time conditions. Further …
Page 1 of 3