Michael Swift
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Michael Swift has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Stephen R. Quake
Stanford University
3 shared publications Felix Horns
Division of Biology and Bioengineering, California Institute of Technology, Pasadena, CA, USA
2 shared publications- Hayden Coffey
Department of Computer Sciences, University of Wisconsin-Madison, Madison, WI, USA
1 shared publication - Ivana Cvijović
Department of Applied Physics, Stanford University
1 shared publication - Konstantinos Kanellis
University of Wisconsin-Madison, Madison, WI, USA
1 shared publication - Shivaram Venkataraman
University of Wisconsin-Madison, Madison, WI, USA
1 shared publication
Latest Publications
- 21 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 75 (4) , 1378-1390
From Good to Great: Parameter Tuning in Memory Tiering Systems
To meet the memory demands of modern data-intensive applications such as graph processing and machine learning, datacenters are adopting tiered memory architectures that include fast tiers such as locally attached DRAM and slower tiers built using CXL-attached memory. While these newer tiers expand capacity, they have higher access latencies and lower bandwidth, which can lead to application performance degradation. To minimize performance loss, frequently accessed (hot) data must be placed close to the CPU in faster tiers and infrequently accessed (cold) data can be placed in farther, slower memory tiers. Existing software-managed tiering solutions employ heuristics and pre-configured thresholds to determine hot and cold data, and make placement and migration decisions. Unfortunately, due to their use of static thresholds, these systems fail to adapt to different workloads and the underlying hardware, and hence perform poorly. In this paper, we study the relationship between thresholds and parameters used by tiering systems and application performance. We leverage Bayesian Optimization to tune the parameters of three tiering systems: HeMem, Memtis, and HMSDK. We find that applications are sensitive to these parameter values, and that best parameter values vary widely across workloads, inputs, thread counts, memory ratios, and hardware characteristics. Our experiments reveal that the Bayesian optimizer can learn workload behaviors and set parameter values that result in best performance. With proper parameter tuning, performance can improve by up to 2x compared to default configurations, highlighting the importance of parameter selection for memory tiering systems.
- 21 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computers
- Vol. 75 (4) , 1378-1390
From Good to Great: Parameter Tuning in Memory Tiering Systems
To meet the memory demands of modern data-intensive applications such as graph processing and machine learning, datacenters are adopting tiered memory architectures that include fast tiers such as locally attached DRAM and slower tiers built using CXL-attached memory. While these newer tiers expand capacity, they have higher access latencies and lower bandwidth, which can lead to application performance degradation. To minimize performance loss, frequently accessed (hot) data must be placed close to the CPU in faster tiers and infrequently accessed (cold) data can be placed in farther, slower memory tiers. Existing software-managed tiering solutions employ heuristics and pre-configured thresholds to determine hot and cold data, and make placement and migration decisions. Unfortunately, due to their use of static thresholds, these systems fail to adapt to different workloads and the underlying hardware, and hence perform poorly. In this paper, we study the relationship between thresholds and parameters used by tiering systems and application performance. We leverage Bayesian Optimization to tune the parameters of three tiering systems: HeMem, Memtis, and HMSDK. We find that applications are sensitive to these parameter values, and that best parameter values vary widely across workloads, inputs, thread counts, memory ratios, and hardware characteristics. Our experiments reveal that the Bayesian optimizer can learn workload behaviors and set parameter values that result in best performance. With proper parameter tuning, performance can improve by up to 2x compared to default configurations, highlighting the importance of parameter selection for memory tiering systems.
- Open Access
Long-term B cell memory emerges at uniform relative rates in the human immune response
B cells generate pathogen-specific antibodies and play an essential role in providing adaptive protection against infection. Antibody genes are modified in evolutionary processes acting on the B cell populations within an individual. These populations proliferate, differentiate, and migrate to long-term niches in the body. However, the dynamics of these processes in the human immune system are primarily inferred from mouse studies. We addressed this gap by sequencing the antibody repertoire and transcriptomes from single B cells in four immune-rich tissues from six individuals. We find that B cells descended from the same pre-B cell (“lineages”) often co-localize within the same tissue, with the bone marrow harboring the largest excess of lineages without representation in other tissues. Within lineages, cells with different levels of somatic hypermutation are uniformly distributed among tissues and functional states. This suggests that the relative probabilities of localization and differentiation outcomes change negligibly during affinity maturation, and quantitatively agrees with a simple dynamical model of B cell differentiation. While lineages strongly co-localize, we find individual B cells nevertheless make independent differentiation decisions. Proliferative antibody secreting cells, however, deviate from these global patterns. These cells are often clonally expanded, their clones appear universally distributed among all sampled organs, and form lineages with an excess of cells of the same type. Collectively, our findings show the limits of peripheral blood monitoring of the immune repertoire, and provide a probabilistic model of the dynamics of antibody memory formation in humans.
Open AccessLong-term B cell memory emerges at uniform relative rates in the human immune response
B cells generate pathogen-specific antibodies and play an essential role in providing adaptive protection against infection. Antibody genes are modified in evolutionary processes acting on the B cell populations within an individual. These populations proliferate, differentiate, and migrate to long-term niches in the body. However, the dynamics of these processes in the human immune system are primarily inferred from mouse studies. We addressed this gap by sequencing the antibody repertoire and transcriptomes from single B cells in four immune-rich tissues from six individuals. We find that B cells descended from the same pre-B cell (“lineages”) often co-localize within the same tissue, with the bone marrow harboring the largest excess of lineages without representation in other tissues. Within lineages, cells with different levels of somatic hypermutation are uniformly distributed among tissues and functional states. This suggests that the relative probabilities of localization and differentiation outcomes change negligibly during affinity maturation, and quantitatively agrees with a simple dynamical model of B cell differentiation. While lineages strongly co-localize, we find individual B cells nevertheless make independent differentiation decisions. Proliferative antibody secreting cells, however, deviate from these global patterns. These cells are often clonally expanded, their clones appear universally distributed among all sampled organs, and form lineages with an excess of cells of the same type. Collectively, our findings show the limits of peripheral blood monitoring of the immune repertoire, and provide a probabilistic model of the dynamics of antibody memory formation in humans.
- Open Access
- 13 January 2023
- Published by Life Science Alliance, LLC in Life Science Alliance
- Vol. 6 (3) , e202201792
Lineage tracing reveals fate bias and transcriptional memory in human B cells
We combined single-cell transcriptomics and lineage tracing to understand fate choice in human B cells. Using the antibody sequences of B cells, we tracked clones during in vitro differentiation. Clonal analysis revealed a subset of IgM+ B cells which were more proliferative than other B-cell types. Whereas the population of B cells adopted diverse states during differentiation, clones had a restricted set of fates available to them; there were two times more single-fate clones than expected given population-level cell-type diversity. This implicated a molecular memory of initial cell states that was propagated through differentiation. We then identified the genes which had strongest coherence within clones. These genes significantly overlapped known B-cell fate determination programs, suggesting the genes which determine cell identity are most robustly controlled on a clonal level. Persistent clonal identities were also observed in human plasma cells from bone marrow, indicating that these transcriptional programs maintain long-term cell identities in vivo. Thus, we show how cell-intrinsic fate bias influences differentiation outcomes in vitro and in vivo.
- 13 January 2023
- Published by Life Science Alliance, LLC in Life Science Alliance
- Vol. 6 (3) , e202201792
Open AccessLineage tracing reveals fate bias and transcriptional memory in human B cells
We combined single-cell transcriptomics and lineage tracing to understand fate choice in human B cells. Using the antibody sequences of B cells, we tracked clones during in vitro differentiation. Clonal analysis revealed a subset of IgM+ B cells which were more proliferative than other B-cell types. Whereas the population of B cells adopted diverse states during differentiation, clones had a restricted set of fates available to them; there were two times more single-fate clones than expected given population-level cell-type diversity. This implicated a molecular memory of initial cell states that was propagated through differentiation. We then identified the genes which had strongest coherence within clones. These genes significantly overlapped known B-cell fate determination programs, suggesting the genes which determine cell identity are most robustly controlled on a clonal level. Persistent clonal identities were also observed in human plasma cells from bone marrow, indicating that these transcriptional programs maintain long-term cell identities in vivo. Thus, we show how cell-intrinsic fate bias influences differentiation outcomes in vitro and in vivo.
- Open Access
Fate Bias and Transcriptional Memory of human B cells
Lineage tracking offers a direct approach to study cell fate determination. In this work we combined single cell transcriptomics and lineage tracing to better understand fate-choice in human B cells. Using the antibody sequence to trace cell lineage during in vitro differentiation, we identified intrinsic proliferative and cell fate biases of B cell subtypes. Clonal analysis revealed that IgM memory B cells were more proliferative than any other B cell subtype, and that cells from the same clone had highly concordant fates. We found that transcriptional memory within clones varies across genes, with strongest persistence in genes related to cell fate determination. Similar persistent transcriptional programs were observed in human plasma cells from bone marrow, suggesting that these programs maintain long-term cell fate in vivo. These results show that cell-intrinsic fate bias influences human B cell differentiation and reveal molecular programs underpinning cell fate determination in B cells.
Open AccessFate Bias and Transcriptional Memory of human B cells
Lineage tracking offers a direct approach to study cell fate determination. In this work we combined single cell transcriptomics and lineage tracing to better understand fate-choice in human B cells. Using the antibody sequence to trace cell lineage during in vitro differentiation, we identified intrinsic proliferative and cell fate biases of B cell subtypes. Clonal analysis revealed that IgM memory B cells were more proliferative than any other B cell subtype, and that cells from the same clone had highly concordant fates. We found that transcriptional memory within clones varies across genes, with strongest persistence in genes related to cell fate determination. Similar persistent transcriptional programs were observed in human plasma cells from bone marrow, suggesting that these programs maintain long-term cell fate in vivo. These results show that cell-intrinsic fate bias influences human B cell differentiation and reveal molecular programs underpinning cell fate determination in B cells.