scholarly journals SpaceSheets: Interactive Latent Space Exploration through a Spreadsheet Interface

2020 ◽  
Author(s):  
Bryan Loh ◽  
Tom White

Generative models capture properties and relationships of images in a generic vector space representation called a latent space. Latent spaces can be sampled to create novel images and perform semantic operations consistent with the principles inferred from the training set. Designers can use representations learned by generative models to express design intent enabling more effective design experimentation. We present the SpaceSheet, a general-purpose spreadsheet interface designed to support the experimentation and exploration of latent spaces.

2020 ◽  
Author(s):  
Bryan Loh ◽  
Tom White

Generative models capture properties and relationships of images in a generic vector space representation called a latent space. Latent spaces can be sampled to create novel images and perform semantic operations consistent with the principles inferred from the training set. Designers can use representations learned by generative models to express design intent enabling more effective design experimentation. We present the SpaceSheet, a general-purpose spreadsheet interface designed to support the experimentation and exploration of latent spaces.


2016 ◽  
Author(s):  
Nasrin Mostafazadeh ◽  
Lucy Vanderwende ◽  
Wen-tau Yih ◽  
Pushmeet Kohli ◽  
James Allen

2004 ◽  
Vol 22 (3) ◽  
pp. 704-715 ◽  
Author(s):  
Yasuhiro Kitazoe ◽  
Hirohisa Kishino ◽  
Takahisa Okabayashi ◽  
Teruaki Watabe ◽  
Noriaki Nakajima ◽  
...  

Author(s):  
Thomas Hunziker

Many common adaptive beamforming methods are based on a sample matrix inversion (SMI). The schemes can be applied in two ways. The sample covariance matrices are either computed over preambles, or the sample basis for the SMI and the target of the beamforming are identical. A vector space representation provides insight into the classic SMI-based beamforming variants, and enables elegant derivations of the well-known second-order statistical properties of the output signals. Moreover, the vector space representation is helpful in the definition of appropriate interfaces between beamfoming and soft-decision signal decoding in receivers aiming at adaptive cochannel interference mitigation. It turns out that the performance of standard receivers incorporating SMI-based beamforming on short signal intervals and decoding of BICM (bit-interleaved coded modulation) signals can be significantly improved by proper interface design.


Sign in / Sign up

Export Citation Format

Share Document