10 Matching Annotations
  1. Oct 2018
    1. Approximate Fisher Information Matrix to Characterise the Training of Deep Neural Networks

      深度神经网络训练(收敛/泛化性能)的近似Fisher信息矩阵表征,可自动优化mini-batch size/learning rate


      挺有趣的 paper,提出了从 Fisher 矩阵抽象出新的量用来衡量训练过程中的模型表现,来优化mini-batch sizes and learning rates | 另外 paper 中的figure画的很好看 | 作者认为逐步增加batch sizes的传统理解只是partially true,存在逐步递减该 size 来提高 model 收敛和泛化能力的可能。

    1. It was the schoolteacher and writer Anne Fisher whose English primer of 1745 began the notion that it's somehow bad to use they in the plural and that he stands for both men and women.
  2. Nov 2017
    1. pairwise overlaps using Fisher’s test and mutual exclusion (Leiserson et al., 2016xA weighted exact test for mutually exclusive mutations in cancer. Leiserson, M.D.M., Reyna, M.A., and Raphael, B.J. Bioinformatics. 2016; 32: i736–i745Crossref | PubMed | Scopus (4)See all ReferencesLeiserson et al., 2016)
  3. Jul 2017
  4. Apr 2017
    1. I want toconsider the mechanics of one of the processes of authentication, before turning to theobstacles that digital medieval projects face in bridging the divide that can separate cred-ible scholarship and the new technologies used to facilitate its creation. Consider thehybrid edition ofCædmon’s Hymn, edited by D.P. O’Donnell and published as a bookwith an accompanying CD in 2005.15Several years after publication, O’Donnell noted ofhis own edition that, despite the inclusion of substantial textual variants made possible bythe mixed digital⁄physical publication, he has ‘yet to see a citation that does not use aform found in the print edition’ (O’Donnell, 114). There are three important points tobe made here. One, in 7 years since publication, CDs have largely become obsolete, andthe future of optical media more generally is questionable. Two, textual variants are notfrequently cited – only very particular types of scholarship are concerned with exploringand discussing textual variants, whether as philological evidence or as literary texts ontheir own right. Scholars regularly discuss Langland’sPiers Plowman, a poem about whichcritics tend to be keenly aware of the state of the text in competing critical editions,without tracking back to the constituent readings made available through the editorialapparatus. Three, O’Donnell’s comments point to a practical difficulty, though one heattempted to resolve by encouraging citation of numbered paragraphs rather than print-bound page numbers: it is difficult to cite things in the hybrid digital⁄analog world thatdominates the present moment.

      Discussion of my edition of Caedmon's Hymn

    2. The digital social networks that have quickly become ubiquitous have made visiblemany of the patterns underlying existing academic personal and professional relationships,and the ways in which reputation and reliability circulate in these structures. Social andintellectual networks have long constituted the professional contexts of scholars, but digi-tal networks representing some subset of those contexts have exposed more of what takesplace at the margins of those networks.

      Digital Social Networks, particularly Facebook and Twitter.

      Makes an interesting point about homogenisation in Facebook and Twitter (i.e. people are a binary of friend or not friend, categories that collapse all different categories.

      Interestingly, both Facebook and Twitter have taken steps to address this recently.

    3. The venerableLISTSERV email lists such as mediev-l (founded in 1992) and other medieval-focusedlistservs are early instances of the digital democratization of scholarship, conducted as anasynchronous and geographically dispersed conversation.

      Listservs and their relationship to Notes and Queries

    4. From July 2008 to April 2012, Googleoffered a service called Google Knol, where a ‘knol’ is a basic ‘unit of knowledge’ asopposed, presumably, to a unit of information.4Users wrote ‘knols’ predicated upon theirown interests and expertise.

      The "KNOL" a unit of knowledge: a Google experiment in crowd-sourcing knowledge

    5. firstwant to consider the ways in which our increased online presence has exposed manyof the existing networks that ground the sources of academic and intellectual authority(reputation, credibility, reliability).

      How online communities have changed the way humanists work.

    6. Fisher, Matthew. 2012. “Authority, Interoperability, and Digital Medieval Scholarship.” Literature Compass 9 (12): 955–64. doi:10.1111/lic3.12018.

      /home/dan/.mozilla/firefox/rwihx4ee.default/zotero/storage/PHS4P7D6/Fisher - 2012 - Authority, Interoperability, and Digital Medieval .pdf