Hiroki Kojima and Takashi Ikegami
URL: https://www.sciencedirect.com/science/article/abs/pii/S0893608022001447
arXiv: https://arxiv.org/abs/2102.01852
Original Abstract: We present a novel artificial cognitive mapping system using generative deep neural networks, called variational autoencoder/generative adversarial network (VAE/GAN), which can map input images to latent vectors and generate temporal sequences internally. The results show that the distance of the predicted image is reflected in the distance of the corresponding latent vector after training. This indicates that the latent space is self-organized to reflect the proximity structure of the dataset and may provide a mechanism through which many aspects of cognition are spatially represented. The present study allows the network to internally generate temporal sequences that are analogous to the hippocampal replay/pre-play ability, where VAE produces only near-accurate replays of past experiences, but by introducing GANs, the generated sequences are coupled with instability and novelty.
Summary (EN): Proposes an artificial cognitive-mapping system using a VAE/GAN that maps input images to latent vectors and internally generates temporal sequences. The latent space self-organizes to reflect the proximity structure of the dataset; adding a GAN endows the hippocampal-replay-like internal sequences with instability and novelty.
概要（日本語）: VAE/GAN（変分オートエンコーダ＋敵対的生成ネットワーク）を用いた人工的な認知地図システムを提案。入力画像を潜在ベクトルに写像し、内部で時系列を生成する。学習後、予測画像間の距離が対応する潜在ベクトル間の距離に反映され、潜在空間がデータの近接構造を反映して自己組織化されることを示す。海馬のリプレイ／プリプレイに類似した内部生成において、GANの導入が不安定性と新奇性をもたらす。
