Goto

Collaborating Authors

Neural Architecture Search in Embedding Space

arXiv.org Machine Learning

The neural architecture search (NAS) algorithm with reinforcement learning can be a powerful and novel framework for the automatic discovering process of neural architectures. However, its application is restricted by noncontinuous and high-dimensional search spaces, which result in difficulty in optimization. To resolve these problems, we proposed NAS in embedding space (NASES), which is a novel framework. Unlike other NAS with reinforcement learning approaches that search over a discrete and high-dimensional architecture space, this approach enables reinforcement learning to search in an embedding space by using architecture encoders and decoders. The current experiment demonstrated that the performance of the final architecture network using the NASES procedure is comparable with that of other popular NAS approaches for the image classification task on CIFAR-10. The beneficial-performance and effectiveness of NASES was impressive even when only the architecture-embedding searching and pre-training controller were applied without other NAS tricks such as parameter sharing. Specifically, considerable reduction in searches was achieved by reducing the average number of searching to 100 architectures to achieve a final architecture for the NASES procedure. Introduction Deep neural networks have enabled advances in image recognition, sequential pattern recognition, recommendation systems, and various tasks in the past decades.



7 of the Prettiest Architecture Instagrams You Ever Did See

WIRED

Matthias Heiderich's (@massju) keen eye can transform just about any building into a fascinating examination of shape, color and form. The self-taught photographer works in Berlin, but shoots internationally including South Korea, Norway and Canada. For Heiderich, the point is to create something visually compelling. "It gives me a feeling of satisfaction to put things in order and to look at well designed architecture and also well composed photos," he says. "I'm trying to get into that state of visual satisfaction as often as possible."


Amsterdam offers World Architecture Festival form and function

Al Jazeera

Top architects are competing for design glory at the World Architecture Festival under way in Amsterdam. The Netherlands may be well known for its industrial windmills, but many of its more modern structures also show an innovative blend of form and function. Complex engineering ingenuity was required to build the Dutch capital's new north-south metro line through the city's canals and thick layers of mud, while the University of Amsterdam is replacing some of its older buildings with new ones that were designed with social interaction and exchange in mind, both architectural feats accomplished with quiet aspiration to improve the way the city's citizens live their lives.


Japan's Arata Isozaki wins Pritzker Architecture Prize

Al Jazeera

Arata Isozaki, a prominent Japanese architect renowed for his versatility and transnational approach to design, has won his field's highest accolade, the Pritzker Architecture Prize. The 87-year-old will receive his coveted award, which consists of $100,000 and a bronze medallion, in May at the Palace of Versailles in France. The jury praised Isozaki for surpassing "the framework of architecture to raise questions that transcend eras and borders", and for his "profound knowledge of architectural history and theory, and embracing the avant-garde". It said in a statement: "He never merely replicated the status quo, but his search for meaningful architecture was reflected in his buildings that, to this day, defy stylistic categorisations." Watch 2019 Laureate, Arata Isozaki, speak about his journey - thinking about architecture, artistic expression, cities and universal phenomenons.https://t.co/uTOr88RjHR