Journal: Sensors (Basel, Switzerland)

Article Title: Out-of-Distribution Detection of Human Activity Recognition with Smartwatch Inertial Sensors

doi: 10.3390/s21051669

Figure Lengend Snippet: Effect on shoulder girdle stabilization exercise inclusion/exclusion from training data for OOD detection of the SPARS9x dataset on: ( a ) FCN Softmax activation; and ( b ) normalized KNN nearest neighbor distance. This figure highlights the resiliency to changes in training class inclusion/exclusion of the KNN algorithm compared to deep learning algorithms for detecting OOD data with this dataset. Note that, even though the mean Softmax activation decreases for the OOD data and increases for the in-distribution classes with shoulder girdle stabilization removed, there is still a slight decrease in AUROC (see Softmax Threshold of b).

Article Snippet: Conversely, in b, the KNN algorithm has no such issue with the shoulder girdle stabilization activity, so the advantage that the traditional algorithms have over Softmax cannot be explained by class confusion alone.

Techniques: Activation Assay

Journal: Nature Communications

Article Title: An amphioxus neurula stage cell atlas supports a complex scenario for the emergence of vertebrate head mesoderm

doi: 10.1038/s41467-024-48774-4

Figure Lengend Snippet: a drawings of Mediterranean amphioxus developmental stages from the egg to the larva (with one open gill slit) stage. The developmental time (hours post fertilization, hpf) is given for embryos raised at 19 °C, and we highlight the neurula stage presented in this study (21 hpf). b Two-dimensional projection of cell clusters (metacells) using a force-directed layout based on the co-clustering graphs for individual cells (see Methods ). Metacells are color-coded by cell type. c Normalized fold change expression of top variable genes (rows) per metacell (columns, grouped by cell type). For each metacell, we selected up to 30 markers with a minimum fold change ≥ 2. Selected gene names from known markers, used to annotate each cell type, are indicated to the right of the heatmap. Genes in bold case are shown in panel ( d ). d Pie charts depicting the fraction of cells mapped to each cell type among the cell transcriptomes and the cell counting experiment (top); and the 3D reconstruction with assignment of nuclei to each germ layer (bottom). A transverse section is shown on the left, and dorsal views with anterior to the top on the right (full, without epidermis nuclei, without epidermis and neural cells nuclei. e Expression profile of previously unknown marker genes for specific cell types (neural, endoderm, anterior epidermis, presumptive cerebral vesicle, notochord, and tailbud) analyzed by in situ hybridization (ISH, top, with anterior to the left and dorsal to the top in side views, n = 10 embryos) and corresponding two-dimensional expression maps (bottom, based on the same layout as panel b ). Gene expression is shown as density maps representing UMI counts (per 10,000 UMIs) in each cell.

Article Snippet: Two-dimensional projection of the metacells were created using a force-directed layout based on the metacell co-clustering graph ( mcell_mc2d_force_knn function).

Techniques: Expressing, Cell Counting, Marker, In Situ Hybridization, Gene Expression

Journal: Nature Communications

Article Title: An amphioxus neurula stage cell atlas supports a complex scenario for the emergence of vertebrate head mesoderm

doi: 10.1038/s41467-024-48774-4

Figure Lengend Snippet: a 2D projection of endodermal metacells on a side view scheme of an amphioxus neurula stage embryo with anterior to the left and dorsal to the top. b Gene expression as normalized fold changes of selected gene markers and corresponding ISH ( n = 10 embryos). Gene expression is shown as density maps representing UMI counts (per 10,000 UMIs) in each cell, with cells positioned in the vicinity of their corresponding metacells. c Multiple in situ hybridization for Dmbx , Fgfrl and Brachyury2 ( n = 5 embryos). Lateral views are shown on the left, with anterior to the left and dorsal to the top. Optical cross-sections at the level of the dotted line are also shown. The labeling intensity is stronger on one side due to the thickness of the embryo that was imaged on one side after mounting. The three genes are coexpressed in the anteriormost region of the dorsal mesendoderm (metacell 1). Normalized fold change expression of each marker is shown on the right. Scale bar: 25 µm.

Article Snippet: Two-dimensional projection of the metacells were created using a force-directed layout based on the metacell co-clustering graph ( mcell_mc2d_force_knn function).

Techniques: Gene Expression, In Situ Hybridization, Labeling, Expressing, Marker

Journal: Nature Communications

Article Title: An amphioxus neurula stage cell atlas supports a complex scenario for the emergence of vertebrate head mesoderm

doi: 10.1038/s41467-024-48774-4

Figure Lengend Snippet: a 2D projection of somitic metacells on a side view scheme of an amphioxus neurula stage embryo, with anterior to the left and dorsal to the top. b Gene expression as normalized fold changes of selected gene markers and corresponding ISH ( n = 10 embryos). c Multiple in situ hybridization for Ripply+Alx+Gata1/2/3 and Ripply+Tcf21/Msc+Alx ( n = 5 embryos). Lateral views are shown on the left, with anterior to the left and dorsal to the top. Optical cross-sections at the level of the dotted line are also shown. The labeling intensity is stronger on one side due to the thickness of the embryo. Ripply , Alx and Gata1/2/3 are coexpressed in the non-muscular part of formed somites (metacells 2 and 11) whereas Ripply , Tcf21/Msc and Alx are coexpressed in the first somite pair (metacells 1 and 3). Scale bar: 25 µm.

Article Snippet: Two-dimensional projection of the metacells were created using a force-directed layout based on the metacell co-clustering graph ( mcell_mc2d_force_knn function).

Techniques: Gene Expression, In Situ Hybridization, Labeling

Journal: Proceedings of machine learning research

Article Title: Temporally Multi-Scale Sparse Self-Attention for Physical Activity Data Imputation

doi:

Figure Lengend Snippet: Macro MAE ± 95% CI on completely held-out test participants.

Article Snippet: The remaining plots in present the ratio of the error obtained by the DW+HD Median, k NN-Softmax and MRNN approaches (the best other models in their groups) to that obtained by the proposed model.

Techniques: