Information-Based Boundary Equilibrium Generative Adversarial Networks with Interpretable Representation Learning

<div>Comparison of our suggested model IBEGAN with previous model InfoGAN. Both models captured variation of hair style by changing latent code <span class="nowrap"><svg height="9.22069pt" id="M156" style="vertical-align:-3.18151pt" version="1.1" viewbox="-0.0498162 -6.03918 10.5526 9.22069" width="10.5526pt" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink"><g transform="matrix(.013,0,0,-0.013,0,0)"><path d="M399 120C363 89 334 79 292 79S154 123 154 256C154 363 199 401 238 401C266 401 298 385 339 349C348 341 354 338 361 338C378 338 404 364 404 393C405 407 401 419 386 429C369 443 336 454 300 454H299C261 454 196 440 130 392C67 345 29 274 29 201C29 92 109 -12 246 -12C307 -12 372 31 418 96L399 120Z"></path></g><g transform="matrix(.0091,0,0,-0.0091,5.486,3.132)"><path d="M389 0V32C297 38 291 46 291 118V635C234 613 175 595 109 583V556L161 554C203 552 207 547 207 497V118C207 46 201 38 110 32V0H389Z"></path></g></svg>.</span> (a) Generated images in InfoGAN. (b) Generated images in IBEGAN (Suggested model).</div>

Computational Intelligence and Neuroscience

fig10

Figure 10

Figure 10: Information-Based Boundary Equilibrium Generative Adversarial Networks with Interpretable Representation Learning 