This creates a vast shared memory pool across systems, allowing users to load the largest and most demanding frontier-class open-weight models available today.
用 Thunderbolt 5 加 RDMA 把多机拼成共享内存池，是在没有 NVLink 的前提下硬凑容量。注意苹果只敢报四台三倍，已见明显衰减。这条路线适合权重长期驻留的本地推理，不适合训练，也撑不起多租户服务。
Giving a cluster a name can impose formality prematurely, adding friction to the process.
Naming clusters can be incorporated into sensemaking efforts though, when not used as result but as intermediate step. As in [[2 step archetype extraction 20121130152904]]
(20) ReconfigBehSci on Twitter: ‘RT @martikagv: New book “Urban Informatics”, #OpenAccess https://t.co/wp45uWU9Mi Edited by @jmichaelbatty @CUHKofficial Michael Goodchild,…’ / Twitter. (n.d.). Retrieved 21 April 2021, from https://twitter.com/SciBeh/status/1384411081456582662
Mann, P., Smith, V. A., Mitchell, J. B. O., & Dobson, S. (2020). Two-pathogen model with competition on clustered networks. ArXiv:2007.03287 [Physics, q-Bio]. http://arxiv.org/abs/2007.03287
Fontanet, A., Grant, R., Tondeur, L., Madec, Y., Grzelak, L., Cailleau, I., Ungeheuer, M.-N., Renaudat, C., Pellerin, S. F., Kuhmel, L., Staropoli, I., Anna, F., Charneau, P., Demeret, C., Bruel, T., Schwartz, O., & Hoen, B. (2020). SARS-CoV-2 infection in primary schools in northern France: A retrospective cohort study in an area of high transmission. MedRxiv, 2020.06.25.20140178. https://doi.org/10.1101/2020.06.25.20140178
Candelieri, A., Giordani, I., Ponti, A., & Archetti, F. (2020). Resilience in urban networked infrastructure: The case of Water Distribution Systems. ArXiv:2006.14622 [Physics]. http://arxiv.org/abs/2006.14622
Ben-David, S. (2018). Clustering—What Both Theoreticians and Practitioners are Doing Wrong. ArXiv:1805.08838 [Cs, Stat]. http://arxiv.org/abs/1805.08838
Riolo, M. A., & Newman, M. E. J. (2020). Consistency of community structure in complex networks. Physical Review E, 101(5), 052306. https://doi.org/10.1103/PhysRevE.101.052306
in clustering analyses, standardization may be especially crucial in order to compare similarities between features based on certain distance measures. Another prominent example is the Principal Component Analysis, where we usually prefer standardization over Min-Max scaling, since we are interested in the components that maximize the variance
Use standardization, not min-max scaling, for clustering and PCA.
missing values are managed, without any pre-processing, by the model used to cluster with the assumption that values are missing completely at random.
VarSelLCM package
Threshold values of 0.8-0.9 are recommended for well separated clusters; to allow for overlapping clusters, we chose a threshold of 0.6.
ClusterGAN : Latent Space Clustering in Generative Adversarial Networks
貌似结果效果很惊艳<del>~</del>~！
用 GAN 做聚类的！
AutoSOME: a clustering method for identifying gene expression modules without prior knowledge of cluster number
main characters' names