Haokai Lu

Stochastic projective methods for simulating stiff chemical reacting systems

Abstract In this paper, stochastic projective methods are proposed to improve the stability and efficiency in simulating stiff chemical reacting systems. The efficiency of existing explicit tau-leaping methods can often severely be limited by the stiffness in the system, forcing the use of small time steps to maintain stability. The methods presented in this paper, namely stochastic projective (SP) and telescopic stochastic projective (TSP) method, can be considered as more general stochastic versions of the recently developed stable projective numerical integration methods for deterministic ordinary differential equations. SP and TSP method are developed by fully re-interpreting and extending the key projective integration steps in the deterministic regime under a stochastic context. These new stochastic methods not only automatically reduce to the original deterministic stable methods when applied to simulating ordinary differential equations, but also carry the enhanced stability property over to the stochastic regime. In some sense, the proposed methods are stochastic generalizations to their deterministic counterparts. As such, SP and TSP method can adopt a much larger effective time step than is allowed for explicit tau-leaping, leading to noticeable runtime speedup. The explicit nature of the proposed stochastic simulation methods relaxes the need for solving any coupled nonlinear systems of equations at each leaping step, making them more efficient than the implicit tau-leaping method with similar stability characteristics. The efficiency benefits of SP and TSP method over the implicit tau-leaping is expected to grow even more significantly for large complex stiff chemical systems involving hundreds of active species and beyond.

Discovering What You're Known For: A Contextual Poisson Factorization Approach

Discovering what people are known for is valuable to many important applications such as recommender systems. Unlike an individuals personal interests, what a user is known for is reflected by the views of others, and is often not easily discerned for a long-tail of the vast majority of users. In this paper, we tackle the problem of discovering what users are known for through a probabilistic model called Bayesian Contextual Poisson Factorization. Moving beyond just modeling users content, it naturally models and integrates additional contextual factors, concretely, users geo-spatial footprints and social influence, to overcome noisy online activities and social relations. Through GPS-tagged social media datasets, we find that the proposed method can improve known-for prediction performance by 17.5% in precision and 20.9% in recall on average, and that it can capture the implicit relationships between a users known-for profile and her content, geo-spatial and social influence.

Linking brain behavior to underlying cellular mechanisms via large-scale brain modeling and simulation

To understand brain behaviors, it is important to directly associate the network level activities to the underlying biophysical mechanisms, which require large-scale simulations with biophysically realistic neural models like Hodgkin-Huxley models. However, when simulations are conducted on models with sufficient biophysical details, great challenges arise from limited computer power, thereby restricting most existing computational works with biophysical models only to small-scale networks. On the other hand, with the emergence of powerful computing platforms, many recent works are geared to performing large-scale simulations with simple spiking models. However, the applicability of those works is limited by the nature of the underlying phenomenological model. To bridge the gap, an intermediate step is taken to construct a scalable brain model with sufficient biophysical details. In this work, great efforts are devoted to taking into account not only local cortical microcircuits but also the global brain architecture, and efficient techniques are proposed and adopted to address the associated computational challenges in simulation of networks of such complexity. With the customized simulator developed, we are able to simulate the brain model to generate not only sleep spindle and delta waves but also the spike-and-wave pattern of absence seizures, and directly link those behaviors to underlying biophysical mechanism. Those initial results are interesting because they show the possibility to determine underlying causes of diseases by simulating the biologically realistic brain model. With further development, the work is geared to assisting the clinicians in selecting the optimal treatment on an individual basis in the future.

Neural Personalized Ranking for Image Recommendation

We propose a new model toward improving the quality of image recommendations in social sharing communities like Pinterest, Flickr, and Instagram. Concretely, we propose Neural Personalized Ranking (NPR) -- a personalized pairwise ranking model over implicit feedback datasets -- that is inspired by Bayesian Personalized Ranking (BPR) and recent advances in neural networks. We further build an enhanced model by augmenting the basic NPR model with multiple contextual preference clues including user tags, geographic features, and visual factors. In our experiments over the Flickr YFCC100M dataset, we demonstrate the proposed NPR model is more effective than multiple baselines. Moreover, the contextual enhanced NPR model significantly outperforms the base model by 16.6% and a contextual enhanced BPR model by 4.5% in precision and recall.

Learning Geo-Social User Topical Profiles with Bayesian Hierarchical User Factorization

Understanding user interests and expertise is a vital component toward creating rich user models for information personalization in social media, recommender systems and web search. To capture the pair-wise interactions between geo-location and users topical profile in social-spatial systems, we propose the modeling of fine-grained and multi-dimensional user geo-topic profiles. We then propose a two-layered Bayesian hierarchical user factorization generative framework to overcome user heterogeneity and another enhanced model integrated with users contextual information to alleviate multi-dimensional sparsity. Through extensive experiments, we find the proposed model leads to a 5\textasciitilde13% improvement in precision and recall over the alternative baselines and an additional 6\textasciitilde11% improvement with the integration of users contexts.

Quality-aware neural complementary item recommendation.

Complementary item recommendation finds products that go well with one another (e.g., a camera and a specific lens). While complementary items are ubiquitous, the dimensions by which items go together can vary by both product and category, making it difficult to detect complementary items at scale. Moreover, in practice, user preferences for complementary items can be complex combinations of item quality and evidence of complementarity. Hence, we propose a new neural complementary recommender Encore that can jointly learn complementary item relationships and user preferences. Specifically, Encore (i) effectively combines and balances both stylistic and functional evidence of complementary items across item categories; (ii) naturally models item latent quality for complementary items through Bayesian inference of customer ratings; and (iii) builds a novel neural network model to learn the complex (non-linear) relationships between items for flexible and scalable complementary product recommendations. Through experiments over large Amazon datasets, we find that Encore effectively learns complementary item relationships, leading to an improvement in accuracy of 15.5% on average versus the next-best alternative.

Community-based geospatial tag estimation

This paper tackles the geospatial tag estimation problem, which is of critical importance for location-based search, retrieval, and mining applications. However, tag estimation is challenging due to massive sparsity, uncertainty in the tags actually used, as well as diversity across locations and times. Toward overcoming these challenges, we propose a community-based smoothing approach that seeks to uncover hidden conceptual communities which link multiple related locations by their common interests in addition to their proximity. Through extensive experiments over a sample of millions of geotagged Twitter posts, we demonstrate the effectiveness of the smoothing approach and validate the intuition that geo-locations have the tendency to share similar “ideas” in the formation of conceptual communities.