Example: The crux of interpreting the latent space of GANs is to find the meaningful directions in the latent space corresponding to the human-understandable concepts.
ru: суть, главный вопрос
downstream task
adjective + noun
applications that are used to evaluate the quality of features learned by self-supervised learning
Example: Another downstream task, which could benefit from disentangled representation, is learning fair predictions.
employ
verb
Example: A line of recent works employs explicit human-provided supervision to identify interpretable directions in the latent space.
ru: использовать
manifold
noun
topological space that locally resembles Euclidean space near each point
Example: Powerful generative models are an appealing tool for this task since the generated images lie on the image manifold by construction.
model-agnostic
adjective
does not depend on a model
Example: By a simple model-agnostic procedure, we find directions corresponding to sensible semantic manipulations without any form of (self-)supervision.
ru: не зависящий от модели
post hoc
adjective, adverb
occurring or done after the event
Example: This paradigm of post hoc disentanglement employs little or no hyper-parameters when learning representationsю
ru: постфактум
qualitative
adjective
relating to, measuring, or measured by the quality of something rather than its quantity
Example: Qualitative comparison of the latent semantics found by the supervised method.
ru: качественный
quantitatively
adverb
with regard to the quantity of something rather than its quality
Example: We also quantitatively compare these two approaches with FID, re-scoring analysis, and user study.
ru: количественно
rigorous
adjective
extremely thorough, accurate(analytical)
Example: The variational Bayesian (VB) approximation is known to be a promising approach to Bayesian estimation, when the rigorous calculation of the Bayes posterior is intractable.
ru: точное (аналитическое)
saliency
noun
meaningfulness (in computer vision)
Example: As an immediate practical benefit of our work, we show how to exploit this finding to achieve competitive performance for weakly-supervised saliency detection.
ru: заметность, значимость
seminal
adjective
strongly influencing later developments
Example: Since the seminal papers on disentanglement learning, such as InfoGAN and β-VAE, a large number of models were proposed, and this problem continues to attract much research attention.
ru: плодотворный, конструктивный
state-of-the-art
adjective
best among competitors
Example: The state-of-the-art GANs are currently able to produce good-looking high-resolution images often indistinguishable from real ones.
ru: лучшая (модель)
trial distribution
adjective + noun
distribution observed during the experiment
Example: The first term in Equation 5 is the Kullback-Leibler (KL) distance from the trial distribution to the Bayes posterior, and the second term is a constant.
ru: опытное(наблюдаемое) распределение
twofold
adverb
containing two parts
Example: The major progress is twofold.
ru: двухфазный, состоящий из двух этапов
ubiquitous
adjective
present, appearing, or found everywhere
Example: The exceptional generation quality paves the road to ubiquitous usage of GANs in applications, e.g., image editing, super-resolution, video generation and many others.