Optimizing Neural Networks that Generate Iimages

By Tijmen Tieleman. 2014

41 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Generative Models

Citations in the corpus

Spatial Transformer Networks (2015, background): “The notion of a composition of transformed parts is taken further by Tieleman [36], where learnt parts are explicitly affine-transformed, with the transform predicted by the network.”
DRAW: A Recurrent Neural Network For Image Generation (2015, background): “This configuration, which we refer to simply as “DRAW”, somewhat resembles the affine transformations used in computer graphics-based autoencoders (Tieleman, 2014).”
Deep Convolutional Inverse Graphics Network (2015, methodology): “r nonprobabilistic model with latent parameters, followed by an inference or optimization algorithms to ﬁnd the most appropriate set of latent parameters given observations. Recently, Tieleman et al. [27] moved beyond this two stage pipeline by using a generic encoder network with a domainspeciﬁc de”
Stacked Capsule Autoencoders (2019, methodology): “In order to make the values of presence probabilities (ak, ak,m and dm) closer to binary we inject uniform noise ∈ [−2, 2] into logits, similar to Tieleman, 2014.”
Weakly-supervised Disentangling with Recurrent Transformations for 3D View Synthesis (2015, background): “xplore the space of rotation about an object image by setting the control signal at each time step of our recurrent network. The problem of training neural networks that generate images is studied in [27]. Dosovitskiy et al. [8] proposed a convolutional network mapping shape, pose and transformation”
Picture: A probabilistic programming language for scene perception (2015, background): “We are also interested in extending Picture by taking insights from learning based “analysis-by-synthesis” approaches such as transforming auto-encoders [18], capsule networks [42] and deep convolutional inverse graphics network [25].”
Disentangled Representations in Neural Models (2016, background): “The exception to this is work by Hinton et al. (2011) and Tieleman (2014) on transforming autoencoders which use a domain-specific decoder to reconstruct input images.”
Sparse Unsupervised Capsules Generalize Better (2018, background): “duced from (Sabour et al., 2017). 1.2. Unsupervised Learning Sparse unsupervised learning was certainly considered during the early development of capsules networks, for example in Tieleman’s thesis (Tieleman, 2014), but is not a feature of recent works. In (Sabour et al., 2017; Hinton et al., 2018)”
Small Data Challenges in Big Data Era: A Survey of Recent Progress on Unsupervised and Semi-Supervised Methods (2019, background): “…networks [88] that assume a Lambertian reﬂectance model and implicitly construct the 3D representation, and transforming auto-encoders [60], [89] that use a domain-speciﬁc decoder to reconstruct images, as well as [90] with an approximate differentiable renderer to explicitly capture the…”