Bo Liu

Dr. Bo Liu

Kempner Institute, Center for Brain Science, MCB, Harvard University, Cambridge, Maryland, USA
orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 August 2020 in PLOS Computational Biology
Verified
Journal Article
Open access
PDF file
JLJuan Lee
Juan Lee
BLBo Liu
Bo Liu
SLSophia Hsin-Jung Li
Sophia Hsin-Jung Li
CKChristopher G. King
Christopher G. King
ZGZemer Gitai
Zemer Gitai
NWNed S. Wingreen
Ned S. Wingreen
Microbes face intense competition in the natural world, and so need to wisely allocate their resources to multiple functions, in particular to metabolism. Understanding competition among metabolic strategies that are subject to trade-offs is therefore crucial for deeper insight into the competition, cooperation, and community assembly of microorganisms. In this work, we evaluate competing metabolic strategies within an ecological context by considering not only how the environment influences cell growth, but also how microbes shape their chemical environment. Utilizing chemostat-based resource-competition models, we exhibit a set of intuitive and general procedures for assessing metabolic strategies. Using this framework, we are able to relate and unify multiple metabolic models, and to demonstrate how the fitness landscape of strategies becomes intrinsically dynamic due to species-environment feedback. Such dynamic fitness landscapes produce rich behaviors, and prove to be crucial for ecological and evolutionarily stable coexistence in all the models we examined. Metabolism is the means by which organisms extract and process nutrients from the environment to live and grow. Various metabolic tasks cost cellular resources such as proteins and energy, while the total amount of resources within a cell are limited. In particular, microbial organisms–like companies with a tight budget–need to strategically allocate their limited resources. How can one assess the “effectiveness” of different microbial resource allocation strategies? Strategies are usually evaluated by their corresponding growth rates under a fixed environment, as if there were a static “fitness landscape". In our work, we demonstrate that even in a system as simple as a chemostat, assessing metabolic strategies is more complex than this static-landscape view: as microbes also shape their shared chemical environment, the fitness landscape becomes dynamic, and successful metabolic strategies are the ones that create a landscape that places themselves at the top. Focusing on how species create their own environment, we provide a geometric approach that unifies a variety of chemostat-type resource-competition models. Our approach yields an intuitive way of assessing the success of competing metabolic strategies, and shows how rich ecological dynamics can arise from species-environment feedback. Our work offers insights into the regulation and evolution of microbial metabolic strategies, and has implications for the biodiversity found in nature.

3 September 2020 in Quantitative Biology
Verified
Journal Article
Open access
PDF file
HRHuixia Ren
Huixia Ren
MZMengdi Zhao
Mengdi Zhao
BLBo Liu
Bo Liu
RYRuixiao Yao
Ruixiao Yao
QLQi Liu
Qi Liu
超唐超 唐
超 唐
Background: Time‐lapse live cell imaging of a growing cell population is routine in many biological investigations. A major challenge in imaging analysis is accurate segmentation, a process to define the boundaries of cells based on raw image data. Current segmentation methods relying on single boundary features have problems in robustness when dealing with inhomogeneous foci which invariably happens in cell population imaging.Methods: Combined with a multi‐layer training set strategy, we developed a neural‐network‐based algorithm — Cellbow.Results: Cellbow can achieve accurate and robust segmentation of cells in broad and general settings. It can also facilitate long‐term tracking of cell growth and division. To facilitate the application of Cellbow, we provide a website on which one can online test the software, as well as an ImageJ plugin for the user to visualize the performance before software installation.Conclusion: Cellbow is customizable and generalizable. It is broadly applicable to segmenting fluorescent images of diverse cell types with no further training needed. For bright‐field images, only a small set of sample images of the specific cell type from the user may be needed for training.

9 September 2024 in Research Square
Verified
Preprint
Open access
PDF file
BLBo Liu
Bo Liu
Sparse connectivity is ubiquitous in biological neural networks. In the machine learning field, it has a long history to train dense neural networks then prune them for sparse connectivity. Recently, this research direction was extended to pruning and sparsifying large language models (LLMs), for which many numerical methods have been invented. Previously, we adopted one of such methods to prune the Llama-2 7B model, in order to develop small-sized language models for on-device AI assistants. We figured out that its sparsity limit is 60%, at which the pruned Llama-2 7B model could generate fluent and helpful answers to daily queries despite occasional factual inaccuracies. In this article, we extended previous pruning and evaluation methods to the newly-released Llama-3 8B model. We found that this Llama 3 model has a sparsity limit of 53%, at which it produced fewer factual inaccuracies than Llama 2 at the sparsity limits but required more parameters. We hypothesized that the accuracy improvement is due to the longer context window and larger training dataset used in Llama 3, while the increased parameter requirement is due to its versatility in generating both text and code output. Based on these results and insights, we proposed future directions to boost the model performance after pruning, paving the way for lightweight LLMs as on-device AI assistants.

22 April 2020 in bioRxiv
Verified
Preprint
Open access
PDF file
HRHuixia Ren
Huixia Ren
MZMengdi Zhao
Mengdi Zhao
BLBo Liu
Bo Liu
RYRuixiao Yao
Ruixiao Yao
QLQi Liu
Qi Liu
超唐超 唐
超 唐
Time-lapse live cell imaging of a growing cell population is routine in many biological investigations. A major challenge in imaging analysis is accurate segmentation, a process to define the boundaries of cells based on raw image data. Current segmentation methods relying on single boundary features have problems in robustness when dealing with inhomogeneous foci which invariably happens in cell population imaging. Here, we demonstrated that combined with multi-layer training set strategy, a neural-network-based algorithm Cellbow can achieve accurate and robust segmentation of cells in broad and general settings. It can also facilitate long-term tracking of cell growth and division. Furthermore, Cellbow is customizable and generalizable. It is broadly applicable to segmenting fluorescent images of diverse cell types with no further training needed. For bright-field images, only a small set of sample images of the specific cell type from the user may be needed for training. To facilitate the application of Cellbow, we provide a website on which one can online test the software, as well as an ImageJ plugin for the user to visualize the performance before software installation.

8 August 2024 in Preprints.org
Verified
Preprint
Open access
PDF file
BLBo Liu
Bo Liu
YXYuru Xu
Yuru Xu
Large language models (LLMs) have shown impressive performance and versatility. However, their billions of parameters and high computational costs hinder the development of personalized and privacy-preserving AI assistants operating locally on user devices. In this work, we explored the potential of pruning LLMs to create lightweight models suitable for user devices, using the moderate-sized Llama-2 7B model as an example. By adopting a simple yet effective pruning method, we found that up to 60% of the weights in the Llama-2 7B model could be pruned without significantly impairing its language modeling capabilities. Furthermore, despite occasional factual inaccuracies, the pruned model at the sparsity limit generated fluent and helpful answers to daily queries, demonstrating its feasibility of on-device AI assistants. These inaccuracies might originate from forgetting or hallucination due to pruning. We proposed a simple protocol to distinguish between the two mechanisms, as well as future directions to improve the pruned models for local AI assistants.

2 December 2024 in PRX Life
Verified
Journal Article
Open access
BLBo Liu
Bo Liu
SQShanshan Qin
Shanshan Qin
VMVenkatesh N. Murthy
Venkatesh N. Murthy
YTYuhai Tu
Yuhai Tu
The integration of neural representations in the two brain hemispheres has been studied extensively in vision, audition, and somatosensation, but less so in olfaction. Recent experiments have revealed that odor responses in cortical neurons driven by separate stimulation of the two nostrils are highly correlated. This bilateral alignment points to structured, nonrandom interhemispheric connections, but how this structure leads to alignment is unclear. Here, we hypothesized that continuous exposure to environmental odors shapes these projections and modeled it as online learning with a local Hebbian rule. We found that Hebbian learning with sparse connections achieves bilateral alignment, exhibiting a linear tradeoff between speed and accuracy (lower learning rate leads to higher alignment level but slower convergence). Furthermore, we identified an inverse scaling relationship between the number of cortical neurons and the interhemispheric projection density required for desired alignment accuracy. Thus, more cortical neurons allow sparser interhemispheric projections, which was explained analytically. We next compared the alignment performance of local Hebbian rule and the global stochastic gradient descent (SGD) learning rule used in artificial neural networks. We found that although SGD leads to the same alignment accuracy with a slightly reduced sparsity, the same inverse scaling relation holds. Our analysis showed that their similar performance originates from the fact that the update vectors of the two learning rules align with each other throughout the entire learning process. Our work suggests that a biologically plausible mechanism with sparse connections suffices for correlated bilateral responses. The quantitative comparison between the local Hebbian rule and the global SGD rule may inspire efficient sparse local learning algorithms for more complex problems. Published by the American Physical Society 2024