Search Results for "space"

...Hora implements multiple efficient indexing algorithms that allow systems to rapidly search through high-dimensional vectors produced by machine learning models. These vectors are commonly generated by neural networks to represent images, text, audio, or other data types in a mathematical embedding space. The library is written in Rust and emphasizes performance, safety, and efficient memory management, making it suitable for production-grade applications requiring low latency and high throughput.

...The project implements the architecture introduced in the CVPR research paper on Adversarial Latent Autoencoders, which focuses on improving generative modeling by learning latent representations aligned with adversarial training objectives. Unlike traditional GANs that directly generate images from random noise, ALAE uses an encoder-decoder architecture that maps images into a structured latent space and then reconstructs them through adversarial training. This design allows the model to learn interpretable latent representations that can be manipulated to control generated image attributes.

Siamese and triplet learning is a PyTorch implementation of Siamese and triplet neural network architectures designed for learning embedding representations in machine learning tasks. These types of networks learn to map images into a compact feature space where the distance between vectors reflects the similarity between inputs. Such embeddings are commonly used in applications like face recognition, image similarity search, and few-shot learning. The repository demonstrates how to train these models using contrastive loss and triplet loss functions, which encourage embeddings of similar samples to be close while pushing dissimilar samples farther apart. ...

...The project implements the system described in the research paper Neural Photo Editing with Introspective Adversarial Networks, which introduces a generative model capable of modifying images in semantically meaningful ways. Instead of editing images by directly manipulating pixels, the software allows users to influence changes in the latent space of a trained generative model. This approach enables large and coherent modifications to images while preserving visual realism. The system relies on an Introspective Adversarial Network, a hybrid architecture combining elements of variational autoencoders and generative adversarial networks to improve reconstruction accuracy and generative quality.

This site contains four packages of Mass and mass-based density estimation. 1. The first package is about the basic mass estimation (including one-dimensional mass estimation and Half-Space Tree based multi-dimensional mass estimation). This packages contains the necessary codes to run on MATLAB. 2. The second package includes source and object files of DEMass-DBSCAN to be used with the WEKA system. 3. The third package DEMassBayes includes the source and object files of a Bayesian classifier using DEMass. DEMassBayes.7z has jar file to be used with WEKA and a readme file listing parameters used. ...

A proximity forest is a data structure that allows for efficient computation of approximate nearest neighbors of arbitrary data elements in a metric space. See: O'Hara and Draper, "Are You Using the Right Approximate Nearest Neighbor Algorithm?", WACV 2013 (best student paper award). One application of a ProximityForest is given in the following CVPR publication: Stephen O'Hara and Bruce A. Draper, "Scalable Action Recognition with a Subspace Forest," IEEE Conference on Computer Vision and Pattern Recognition, 2012. ...

...Much of the search space is pruned early on by a user-selectable fitness function. Within a few epochs, reactor output quickly converges to several sub-optimal, yet high-output reactors. Given enough time, the idea is that the optimal reactor configuration will be yielded.