Creating Common Ground: Architecture For Tactical Learning and Creative Convergence

Certain environments can inhibit learning and stifle enthusiasm, while others enhance learning or stimulate curiosity. Furthermore, in a world where technological change is accelerating we could ask how might architecture connect resource abundant and resource scarce innovation environments? Innovation environments developed out of necessity within urban villages and those developed with high intention and expectation within more institutionalized settings share a framework of opportunity for addressing change through learning and education. This thesis investigates formal and informal learning environments and how architecture can stimulate curiosity, enrich learning, create common ground, and expand access to education. The reason for this thesis exploration is to better understand how architects might design inclusive environments that bring people together to build sustainable infrastructure encouraging innovation and adaptation to change for years to come. The context of this thesis is largely based on Colin McFarlane’s theory that the “city is an assemblage for learning” The socio-spatial perspective in urbanism, considers how built infrastructure and society interact. Through the urban realm, inhabitants learn to negotiate people, space, politics, and resources affecting their daily lives. The city is therefore a dynamic field of emergent possibility. This thesis uses the city as a lens through which the boundaries between informal and formal logics as well as the public and private might be blurred. Through analytical processes I have examined the environmental devices and assemblage of factors that consistently provide conditions through which learning may thrive. These parameters that make a creative space significant can help suggest the design of common ground environments through which innovation is catalyzed.

Dynamic Coculture of a Prevascularized Engineered Bone Construct

The generation of functional, vascularized tissues is a key challenge for the field of tissue engineering. Before clinical implantations of tissue engineered bone constructs can succeed, in vitro fabrication needs to address limitations in large-scale tissue development, including controlled osteogenesis and an inadequate vasculature network to prevent necrosis of large constructs. The tubular perfusion system (TPS) bioreactor is an effective culturing method to augment osteogenic differentiation and maintain viability of human mesenchymal stem cell (hMSC)-seeded scaffolds while they are developed in vitro. To further enhance this process, we developed a novel osteogenic growth factors delivery system for dynamically cultured hMSCs using microparticles encapsulated in three-dimensional alginate scaffolds. In light of this increased differentiation, we characterized the endogenous cytokine distribution throughout the TPS bioreactor. An advantageous effect in the ‘outlet’ portion of the uniaxial growth chamber was discovered due to the system’s downstream circulation and the unique modular aspect of the scaffolds. This unique trait allowed us to carefully tune the differentiation behavior of specific cell populations. We applied the knowledge gained from the growth profile of the TPS bioreactor to culture a high-volume bone composite in a 3D-printed femur mold. This resulted in a tissue engineered bone construct with a volume of 200cm3, a 20-fold increase over previously reported sizes. We demonstrated high viability of the cultured cells throughout the culture period as well as early signs of osteogenic differentiation. Taking one step closer toward a viable implant and minimize tissue necrosis after implantation, we designed a composite construct by coculturing endothelial cells (ECs) and differentiating hMSCs, encouraging prevascularization and anastomosis of the graft with the host vasculature. We discovered the necessity of cell to cell proximity between the two cell types as well as preference for the natural cell binding capabilities of hydrogels like collagen. Notably, the results suggested increased osteogenic and angiogenic potential of the encapsulated cells when dynamically cultured in the TPS bioreactor, suggesting a synergistic effect between coculture and applied shear stress. This work highlights the feasibility of fabricating a high-volume, prevascularized tissue engineered bone construct for the regeneration of a critical size defect.

Learning Binary Code Representations for Effective and Efficient Image Retrieval

The size of online image datasets is constantly increasing. Considering an image dataset with millions of images, image retrieval becomes a seemingly intractable problem for exhaustive similarity search algorithms. Hashing methods, which encodes high-dimensional descriptors into compact binary strings, have become very popular because of their high efficiency in search and storage capacity. In the first part, we propose a multimodal retrieval method based on latent feature models. The procedure consists of a nonparametric Bayesian framework for learning underlying semantically meaningful abstract features in a multimodal dataset, a probabilistic retrieval model that allows cross-modal queries and an extension model for relevance feedback. In the second part, we focus on supervised hashing with kernels. We describe a flexible hashing procedure that treats binary codes and pairwise semantic similarity as latent and observed variables, respectively, in a probabilistic model based on Gaussian processes for binary classification. We present a scalable inference algorithm with the sparse pseudo-input Gaussian process (SPGP) model and distributed computing. In the last part, we define an incremental hashing strategy for dynamic databases where new images are added to the databases frequently. The method is based on a two-stage classification framework using binary and multi-class SVMs. The proposed method also enforces balance in binary codes by an imbalance penalty to obtain higher quality binary codes. We learn hash functions by an efficient algorithm where the NP-hard problem of finding optimal binary codes is solved via cyclic coordinate descent and SVMs are trained in a parallelized incremental manner. For modifications like adding images from an unseen class, we propose an incremental procedure for effective and efficient updates to the previous hash functions. Experiments on three large-scale image datasets demonstrate that the incremental strategy is capable of efficiently updating hash functions to the same retrieval performance as hashing from scratch.