TRENDY method code files by Yue Wang (20581715)

“…<p dir="ltr">code files for TRENDY method, used for inferring gene regulatory networks (GRN) from single-cell gene expression data</p><p dir="ltr">TRENDY algorithm itself uses the following packages: numpy=1.24.3, scipy=1.13.1, sklearn=1.4.2, torch=2.2.2,</p><p dir="ltr">other algorithms and comparison files need extra packages: pingouin=0.5.3, pandas=2.2.2, xgboost=2.0.3</p><p dir="ltr">warning: the NonlinearODEs algorithm (xgbgrn.py) runs well on Windows, but might crash on MacOS</p><p dir="ltr">manuscript for TRENDY method: GRN_transformer.pdf</p><p dir="ltr">major code files:</p><p dir="ltr">TRENDY_tutorial.py: a tutorial for using TRENDY method. to apply the TRENDY method, it also needs the following files: models.py, wendy_solver.py in previous_methods folder, trendy_1.pth and trendy_2.pth in weights folder</p><p dir="ltr">models.py: contains all three versions of the TE(k) model</p><p dir="ltr">train_trendy_first_half.py: train the TE(k=1) model for the first half of TRENDY method. the trained weights trendy_1.pth is in the folder weights</p><p dir="ltr">train_trendy_second_half.py: train the TE(k=3) model for the second half of TRENDY method. the trained weights trendy_2.pth is in the folder weights</p><p dir="ltr">code files for training other models:</p><p dir="ltr">train_GENIE3_rev.py: train the GENIE3_rev method. the trained weights genie_rev.pth is in the folder weights</p><p dir="ltr">train_NonlinearODEs_rev.py: train the NonlinearODEs_rev method. the trained weights nlode_rev.pth is in the folder weights</p><p dir="ltr">train_SINCERITIES_rev.py: train the SINCERITIES_rev method. the trained weights sinc_rev.pth is in the folder weights</p><p dir="ltr">code files for comparing different methods:</p><p dir="ltr">test_SINC_new.py: used to compare different methods on SINC data</p><p dir="ltr">test_DREAM4_new.py: used to compare different methods on DREAM4 data</p><p dir="ltr">test_THP1_new.py: used to compare different methods on THP-1 data</p><p dir="ltr">test_hESC_new.py: used to compare different methods on hESC data</p><p dir="ltr">code for previously known methods, all in the folder previous_methods:</p><p dir="ltr">sincerities.py: code for SINCERITIES method</p><p dir="ltr">xgbgrn.py: code for NonlinearODEs method</p><p dir="ltr">GENIE3.py: code for GENIE3 method</p><p dir="ltr">wendy_solver.py: code for WENDY method</p><p dir="ltr">auxiliary code files:</p><p dir="ltr">methods.py: contains functions for different methods</p><p dir="ltr">evaluation.py: compare the inferred GRN with the ground truth GRN and calculate AUROC and AUPRC</p><p dir="ltr">plots_new.py: draw plots</p><p dir="ltr">nd_alg.py: code for network deconvolution method for enhancing inferred GRNs</p><p dir="ltr">brane_alg.py: code for BRANE Cut method for enhancing inferred GRNs</p><p dir="ltr">generate training data, all in the folder Data_generation:</p><p dir="ltr">A_data_generation.py generates a random GRN, generate.py generates all files in the following folder total_data_10, the other four files are previously known methods</p><p dir="ltr">data sets:</p><p dir="ltr">folder total_data_10 (<a href="https://zenodo.org/records/13929908" rel="nofollow" target="_blank">https://zenodo.org/records/13929908</a>, not uploaded here): saves the generated data for training and validation. for different endings: A is the ground truth GRN, cov is the covariance matrix, data is the original data, genie is the inferred GRN by GENIE3, nlode is the inferred GRN by NonlinearODEs, revcov is the Ktstar matrix, sinc is the inferred GRN by SINCERITIES, wendy is the inferred GRN by WENDY. these data files are generated by the files in folder Data_generation. here we only upload the files for testing. for full files of this folder, see <a href="https://zenodo.org/records/13929908" rel="nofollow" target="_blank">https://zenodo.org/records/13929908</a></p><p dir="ltr">folder rev_wendy_all_10 (not uploaded here. see <a href="https://zenodo.org/records/13929908" rel="nofollow" target="_blank">https://zenodo.org/records/13929908</a>): saves the inferred Kt' matrix (xxx_ktstar files) and A_1 matrix (xxx_revwendy files) in TRENDY</p><p dir="ltr">folder SINC: ground truth GRNs and inferred GRNs in SINC data set. …”

The code for sample size calculation. by Ying Zhou (25031)

Research data for paper: Efficient Event-based Delay Learning in Spiking Neural Networks by Balázs Mészáros (16890225)

“…</li></ol><p dir="ltr">The data was generated and analysed with the code available on GitHub at https://github.com/mbalazs98/deventprop/</p><p dir="ltr">results.py contains all test accuracies shown in figures 4-7. …”

Prime Number Source Code with Dataset by Ayman Mostafa (19839867)

The principle of Partial Convolution. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Ablation experiments results of YOLOv5s. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Overall network architecture of FCMI-YOLO. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

The principle of MLCA mechanism. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Parameters of the dataset. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Comparison of mAP@0.5 for different ratios. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Primary training parameters for the model. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Distribution of the dataset. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Parameters of the FasterNext and C3. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

System diagram. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Schematic diagram of Inner-IoU. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Model train environment. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

The structure of FasterNext. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

The structure of MLCA mechanism. by Junjie Lu (160350)

“…To address this issue, this paper proposes FCMI-YOLO, a real-time fire detection algorithm optimized for edge devices. Firstly, the FasterNext module is proposed to reduce computational cost and enhance detection precision through lightweight design. …”

Reflection matrix microscopy data and CLASS algorithms by Sungsam Kang (18931031)

“…<h3><b>Reflection matrix microscopy data and CLASS algorithms</b></h3><p dir="ltr">This software package includes essential algorithms for reflection matrix microscope (RMM) and the CLASS algorithm. …”

MAT-PointPillars network structure diagram. by Xinpeng Yao (18882573)