Data and models for : Cross-individual translation of spontaneous zebrafish brain activity through a shared latent representation

This repository contains data and model files used in the article Cross-individual translation of spontaneous zebrafish brain activity through a shared latent representation. You can find the code to create those files and use them to produce the article figures here. Whole‑brain Single‑Cell Datasets WBSC.zip Archive of recorded whole‑brain single‑neuron binarized spike trains and 3D neuron coordinates. Each file is the form DATA_<FishName>.h5. Voxelized Datasets Voxelized.zip Voxelized brain‑activity datasets (various resolutions). Voxelgrids.zip Corresponding voxel‑grid geometry files used to map activity onto brain volumes. Pretrained Voxelized RBMs vRBMr_multivoxelized_6fish_20.0vox_M40_l2l10.1.h5 Main pretrained RBM for the multi‑fish voxelized dataset (20μm voxels). Dropped.zip RBMs trained using 5 fish and leaving-out 1 fish (20μm voxels). CrossValidation_VOX.zip RBMs trained for cross‑validation experiments on voxelized data from 1 fish. Repeats_VOX.zip Multiple training repeats of voxelized RBMs to assess variability. Pretrained Single‑Cell RBMs bRBMs.zip Trained RBMs for individual fish using whole‑brain single‑cell data (teachers). biRBMs.zip Trained LaRBMs (students). biRBMs_before_training.zip Initialized but untrained student LaRBMs. Repeats_WBSC.zip Repeated training runs of single‑cell RBMs. Precomputed Analysis WeightDist_6fish_WBSC_M100_l10.02_l2l10_sigma4_epsilon1.0e-5.h5 Spatial weight‑distribution metrics for whole-brain single-cell RBMs. DeepFakeFreeEnergy_6fish_WBSC_M100_l10.02_l2l10.h5 Free‑energy evaluations of translated neuronal configurations generated from the teacher and student RBMs. DeepFakeTransferMethods_6fish_WBSC_M100_l10.02_l2l10.h5 Analysis of different compute methods to translate neuronal configurations between brains. DeepFakeStats_6fish_WBSC_M100_l10.02_l2l10.h5 Summary statistics for translation analyses. DeepFakeActivityDistance_6fish_WBSC_M100_l10.02_l2l10.h5 Spatial distance metrics between translated neuronal activities.本仓库包含论文《通过共享隐表示实现自发斑马鱼全脑活动的跨个体翻译》中所使用的数据与模型文件。 您可在此处获取生成上述文件并复现论文配图的代码。 全脑单细胞数据集 WBSC.zip：记录的全脑单个神经元二值化尖峰序列与三维神经元坐标存档。所有文件均遵循命名格式：DATA_<鱼名>.h5。 体素化数据集 Voxelized.zip：不同分辨率的脑活动体素化数据集。 Voxelgrids.zip：用于将活动映射至脑体积的对应体素网格几何文件。 预训练体素化受限玻尔兹曼机（Restricted Boltzmann Machine，简称RBM） vRBMr_multivoxelized_6fish_20.0vox_M40_l2l10.1.h5：用于多鱼体素化数据集（20μm体素）的主预训练RBM。 Dropped.zip：使用5条鱼的数据训练、留1条鱼作为测试集的RBM（20μm体素）。 CrossValidation_VOX.zip：针对单条鱼的体素化数据训练的、用于交叉验证实验的RBM。 Repeats_VOX.zip：为评估模型变异性而多次训练的体素化RBM。 预训练单细胞RBM bRBMs.zip：使用全脑单细胞数据（教师模型）训练的、针对单条鱼的RBM。 biRBMs.zip：训练得到的学生模型LaRBM。 biRBMs_before_training.zip：已初始化但尚未训练的学生模型LaRBM。 Repeats_WBSC.zip：多次训练运行的单细胞RBM。 预计算分析结果 WeightDist_6fish_WBSC_M100_l10.02_l2l10_sigma4_epsilon1.0e-5.h5：全脑单细胞RBM的空间权重分布指标。 DeepFakeFreeEnergy_6fish_WBSC_M100_l10.02_l2l10.h5：教师与学生RBM生成的翻译后神经元配置的自由能评估结果。 DeepFakeTransferMethods_6fish_WBSC_M100_l10.02_l2l10.h5：不同脑间神经元配置翻译计算方法的分析结果。 DeepFakeStats_6fish_WBSC_M100_l10.02_l2l10.h5：翻译分析的汇总统计量。 DeepFakeActivityDistance_6fish_WBSC_M100_l10.02_l2l10.h5：翻译后神经元活动间的空间距离指标。