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 21 for “"weight decay"”.
-
Konvoliucinių neuroninių tinklų veikimas vaizdų klasifikavime su mažu duomenų kiekiu /
… overfitting prevention methods that were used: Weight decay, Dropout, DisturbLabel produced results which have shown that regardless of small training image data amount an improved accuracy can be achieved using any of these methods. The best classification accuracy improvement was achieved …
-
Mathematical Guarantees of Near-Optimal Cross-Entropy Loss through Neural Collapse
… up to scaling, 3) last-layer class features and weights becomes equal up the scaling, and 4) classification behavior collapses to the nearest class center (NCC) decision rule. This thesis investigates the emergence of Neural Collapse for unbiased neural network classifiers when the training …
-
Decay Makes Supervised Predictive Coding Generative
… constraint on the state nodes and the synaptic weights of a Predictive Coding network, the solution space collapses to a unique solution that is close to the training inputs. This minimum-norm constraint can be enforced by adding decay to the Predictive Coding equations. Decay is implemented in …
-
Adaptive Analog VLSI Signal Processing and Neural Networks
… case of correlated inputs required dealing with weight decay naturally exhibited by the learning rule. We introduced a source-follower floating-gate synapse as an improvement over our earlier source-degenerated floating-gate synapse in terms of relative weight decay cite{kn:Dugger2004}. A larger …
-
DRAG: Diversity in Retrieval Augmented Generation through the Application of Submodular Functions
… a one-hot feature based function) with manual weights and Facility Location, and 3) One-Hot Quantization with exponential weight decay and Facility Location. We perform hyperparameter selection for the submodular functions and for MMR. We evaluate these on five datasets designed for …
-
Įterptinių instrukcijų atakų (XSS) aptikimas naudojant dirbtinius neuroninius tinklus /
… optimal epoch number, batch size, learning rate, weight decay rate, dropout rate and best vectorization, tokenization methods. Lastly, a created model was implemented in logging. A website was created that sends requests to "Seq" server, "Seq" server logs the requests, groups possible attacks and …
-
Hyperspectral unmixing and band weighting for multiple endmember sets
… of spectral unmixing methods. As a result, band weighting can aid hyperspectral unmixing and endmember estimation methods. 'Band Weighting' refers to a method for assigning weights to each band in hyperspectral data that indicate the relative degree of importance during unmixing. Many …
-
MELANOMA DETECTION BASED ON DEEP LEARNING NETWORKS
… pretrained model with learning rates and weight decay. This model consists of 50 layers organized into blocks that include batch normalization and skip connections (known as connections). We adjusted the depth of the model to improve its accuracy. Our experimental results demonstrate that …
-
Single and Multi-Label Environmental Sound Classification Using Convolutional Neural Networks
… mel-spectrogram of the signals. Dropout and weight decay regularization methods are applied and the cross-entropy loss is optimized using Adam algorithm. Results show that these systems are very sensitive to noise and level corruptions of the inputs. Techniques like data augmentation and …
-
Implementación de un modelo inteligente para identificar eventos sospechosos a través de YOLO
… scheduler, explicit warmup epochs, and increased weight decay. The architecture was kept fixed in both cases, so that any difference in performance could be attributed exclusively to the training regime. Results showed that both models achieved a mAP@0.5 of 0.955, but multi-scenario evaluation …
-
Molecular Weight and Degree of Acetylation of Ultrasonicated Chitosan
… activity, are related to its molecular weight and degree of acetylation (DA). High intensity ultrasonication has the potential to modify molecular weight of chitosan and thus alter or improve chitosan functional properties. The objective of this research was to determine the DA and …
-
Topographic mappings and feed-forward neural networks
… good generalisation as a result of implicit weight decay in the training process. It is further argued that for ideal structure retention, the network transformation should be perfectly smooth for all inter-data directions in input space. Finally, there is a critique of optimisation …
-
Understanding Deep Learning via Analyzing Training Dynamics
… linear network. Our analysis reveals the role of weight decay in discarding the nuisance features and keeping the robust features. Then we show there will be a long plateau in both the loss and accuracy interpolation (between a random initialization with the minimizer it converges to) if different …
-
Reduced hyperBF networks : practical optimization, regularization, and applications in bioinformatics.
… activation function is a radial function of a weighted distance. The local weighting of the distance accounts for the variation in local scaling and discriminative power along each feature. Such generalization makes HyperBF networks capable of interpolating decision functions with high …
-
Early Detection of Oak Wilt using Unmanned Aerial Vehicles (UAV) & Computer Vision
… vision. Expanding on earlier work with a lightweight CNN achieving an F1-score of 79.8%, this study compares twelve state-of-the-art architectures—traditional CNNs, mobile-optimized models, hybrid networks, and vision transformers—across multiple evaluation metrics. A UAV-acquired dataset of …
-
Optimization Dynamics in Mildly Overparametrized Models
… neural networks with positive second-layer weights using gradient descent, and (2) two-layer neural networks with standard second-layer weights using gradient descent with weight decay, can efficiently recover the target network. These results also suggest a strong form of feature learning, …
-
Generalisation and optimisation in neural networks
… DNF length), which maps to the network's minimum weight norm. We show the prior over functions, <em>P</em>(<em>f</em>), exponentially favours functions with low <em>K<sub>DNF</sub></em>(<em>f</em>) Consequently, low-complexity functions are learnable, while …
-
Advances in Microbiome Analysis: From the Variance Component Model to Deep Learning
… curve (DOC) microbiome dynamics analysis, weighted Spearman correlation analysis, network analysis, etc. We find that previously defined microbial taxa have different effect on inflammation markers and lung function in an HIV positive population. The dynamics of post-ART pneumotype SPT was …
-
Phonetische Transkription für ein multilinguales Sprachsynthesesystem
… wobei auf die Gewichte dieser Verbindung ein Weight Decay (Gewichtezerfall) angewendet wird. Damit erreicht man eine Bewertung der Eingabeinformation während des Trainings. Eingabeknoten mit einem großen Informationsgehalt werden verstärkt, während weniger interessante Knoten abgeschwächt …
-
Iterative Data Augmentation For Enhancing Deep Learning Performance With Limited Training Data
… such as batch size, dropout rate, and weight decay in response to stagnation in validation accuracy. When validation accuracy fails to improve over multiple epochs, these parameters are adjusted to promote generalization and mitigate overfitting. Experimental validation across CIFAR-10, …
Page 1 of 2