OSCR

Compressive learning scaffolds higher-order network structure to enhance human knowledge acquisition.

Code ↔ Paper

11 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 11 matches · 3 of them tie a paragraph to a whole file, not to given lines: weak matches, whose lines are not tinted
  1. [1] § Methods › Time-resolved mouse trajectory analysis ↔ Code/Fig2andSI/Fig2andSI.m, lines 15–40 · score 0.78 · distractor angle, mouse trajectory, target angle, binary classifier, incorrect, position
  2. [2] § Methods › Time-resolved mouse trajectory analysis ↔ Code/Fig3andSI/Fig3andSI.m, lines 15–40 · score 0.78 · distractor angle, mouse trajectory, target angle, binary classifier, incorrect, position
  3. [3] § Methods › Transition network ↔ Code/Fig1d/HK_network.m, the whole file · a weak match · score 0.71 · Barab si Albert, Scale free network, probability, connected, edges, transition
  4. [4] § Methods › Time-resolved mouse trajectory analysis › Overall accuracy computation ↔ Code/Fig2andSI/Fig2andSI.m, lines 15–40 · score 0.68 · mouse click, hint onset, binary classification, reaction, position
  5. [5] § Methods › Time-resolved mouse trajectory analysis › Overall accuracy computation ↔ Code/Fig3andSI/Fig3andSI.m, lines 15–40 · score 0.68 · mouse click, hint onset, binary classification, reaction, position
  6. [6] § Methods › Network compressibility analysis ↔ Code/Fig1c/Fig1c.m, lines 24–44 · score 0.66 · reconstruction error, adjacency matrix, truncating, MSE, singular, SVD
  7. [7] § Results › Two-stage computational modeling explains compressive learning ↔ Code/Fig5bandSI/FitHypergraphModel.m, the whole file · a weak match · score 0.54 · RandomWalk, LeafToHub, HubToLeaf, hypergraph, fit, model
  8. [8] § Results › Two-stage computational modeling explains compressive learning ↔ Code/Fig5bandSI/Fig5b.m, lines 56–115 · score 0.54 · corrected Akaike Information, AICc, HG, OS, SR, Model
  9. [9] § Results › Two-stage computational modeling explains compressive learning ↔ Code/Fig5bandSI/FitHypergraphModel.m, the whole file · a weak match · score 0.54 · Hypergraph model, LeafToHub, random walks, HubToLeaf
  10. [10] § Results › Two-stage computational modeling explains compressive learning ↔ Code/Fig5bandSI/Modelfitting_main.m, lines 193–242 · score 0.52 · SR models, hypergraph model, random walk, HG, OS, entropy
  11. [11] § Results › Two-stage computational modeling explains compressive learning ↔ Code/Fig5bandSI/Fig5b.m, lines 216–268 · score 0.51 · SR models, hypergraph model, random walk, HG, OS, entropy

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

MATLAB · 465 lines · 22 KB · no license · 2 matches

  1. % CompressiveLearningPaper_Figure2.m
  2. % Re-organized by XR @ May 31 2026
  3. % Script for plots in Figure 2
  5. %%
  6. clear
  7. clc
  9. %% add path
  10. % !!!!!!!!!! Please replace the 'folder' with your own path !!!!!!!!!!
  11. folder = '/Volumes/My Passport/HierarchicalCluster';
  12. addpath(genpath(fullfile(folder, 'CompressiveLearningPaper')));
  14. %% parameters
  15. %%% Columns index for the all_data.mat
  16. % columnNames = {'network', 'subjID', 'trialNo', 'blockNo', 'cue', 'target', 'distractor', ...
  17. % 'target-Xpos', 'target-Ypos', 'distractor-Xpos', 'distractor-Ypos', ...
  18. % 'mouse-Xpos', 'mouse-Ypos', 'mouse-Time', 'mouse-target-Angle', 'mouse-distractor-Angle', ...
  19. % 'choiceId', 'choice-Final', 'clickRT', 'clickChoice', 'clickConds'};
  20. col_network = 1; % 'lattice', 'random', 'smallWorld', 'scaleFree'
  21. col_subjID = 2; % subjID
  22. col_trialNo = 3; % trialNo: 1-1000
  23. col_blockNo = 4; % blockNo: 1-5
  24. col_objCue = 5; % cue node
  25. col_objTgt = 6; % target node
  26. col_objDtr = 7; % distractor node
  27. col_imgPosX_tgt = 8; % x-coordinate of the target during the choice period: (units) 'pixel' for in-lab exp and 'height' for online exp
  28. col_imgPosY_tgt = 9; % y-coordinate of the target during the choice period
  29. col_imgPosX_dtr = 10; % x-coordinate of the distractor
  30. col_imgPosY_dtr = 11; % y-coordinate of the distractor
  31. col_mouseXpos = 12; % mouse trajectory: x-coordinate
  32. col_mouseYpos = 13; % mouse trajectory: y-coordinate
  33. col_mouseTime = 14; % mouse trajectory: sampling time
  34. col_angle_tgt = 15; % mouse-to-target angle
  35. col_angle_dtr = 16; % mouse-to-distractor angle
  36. col_choice_t = 17; % choice (binary classification) at each time point: 1 = correct (choosing target), 0 = incorrect
  37. col_choice_final = 18; % choice per trial before hint onset: 1 = correct (choosing target), 0 = incorrect
  38. col_respRT = 19; % response reaction time (time when the stimulus was clicked)
  39. col_choice_click = 20; % choice based on the mouse click: 0 = correct, 1 = error/timeout (reversed compared to the choice from mouse trajectory)
  40. col_choice_label = 21; % response condition: (1) correct resp before cue: 0; (2) correct resp after cue: 1; (3) incorrect resp before cue: 2; (4) incorrect resp after cue: 3;
  42. %% network
  43. nodeNum = 16; % 16 nodes in each network
  44. measTime = 2; % 2 seconds
  45. nBlock_pred = 5; % prediction task: 5 blocks
  46. trialsBlc_pred = 200;
  47. grayoffT = 0.8;
  48. network_list = {'lattice', 'random', 'smallWorld', 'scaleFree'};
  50. circle_list = 0 : 1/60 : 1.5;
  51. tInBlc = 200 : 200 : 1000;
  52. nBlock = length(tInBlc); %% seperate the total trials into 10 blocks
  53. trialsInBlc = zeros(nBlock, 2);
  54. trialsInBlc(:, 1) = [1, tInBlc(1 : end - 1) + 1]';
  55. trialsInBlc(:, 2) = tInBlc;
  57. %% read the data
  58. codePath = fullfile(folder, 'CompressiveLearningPaper', 'Code', 'Fig2');
  59. load(fullfile(codePath, 'data_Exp1_Figure2.mat'), 'all_data');
  61. %% extract the relevant information for subsequent plot
  62. expMode_subj = {'offline', 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', ...
  63. 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', 'offline', ...
  64. 'online', 'online', 'online', 'online', 'online', 'online', 'online', 'online', 'online', 'online', ...
  65. 'online', 'online', 'online', 'online', 'online', 'online', 'online', 'online', 'online', 'online'};
  67. %%% binary classified choice from the mouse trajectory
  68. nExp = length(network_list);
  69. subLen = length(expMode_subj);
  70. angAcc_exp = nan(subLen, length(circle_list), nExp); %% accuracy in each time point
  71. angAcc_oneHot_exp = nan(subLen, nExp); %% accuracy per trial based on click behaviour: before cue or choice at 0.8 s
  72. angAcc_dgr_oneHot_exp = nan(subLen, 3, nExp); %% 3: (1) degree of fromNode = 4; (2) degree of toNode = 4; (3) both from & to = 4;
  73. dgr_oneHot_N_exp = nan(subLen, 3, nExp); %% trial numbers in each of the matched condition
  74. dgr_oneHot_trlIdx_exp = cell(subLen, 3, nExp);
  75. choiceId_trials_exp = cell(subLen, nExp); %% save the choiceId_trials.mat for each participant in order to do the re-sampling
  76. nBin = 100;
  77. BinL = 1000 / nBin;
  78. angAcc_oneHot_ln_exp = nan(subLen, BinL, nExp); %%% learning effect
  80. %%% click accuracy (overall, before-hint, after-hint)
  81. clickAcc_exp = nan(subLen, 3, nExp); % 3: 1) overall accuracy; 2) before-hint accuracy; 3) after-hint accuracy
  83. for iExp = 1 : nExp
  84. netWord = network_list{iExp};
  85. idx = strcmp(all_data.network, netWord);
  86. clusterResult_iExp = all_data(idx, :);
  87. subjCol = clusterResult_iExp.subjID;
  89. subLen = max(subjCol);
  90. [transMat, nodeDegree] = network_class(netWord);
  91. uniqDgr = unique(nodeDegree);
  93. %% loop across subjects
  94. angAcc = zeros(subLen, length(circle_list)); %% accuracy in each time point
  95. lenData = zeros(subLen, length(circle_list)); %% data length in each time point
  96. angAcc_oneHot = zeros(subLen, 1); %% overall accuracy: before cue or choice at 0.8 s
  97. angAcc_dgr_oneHot = zeros(subLen, 3); %% 3: (1) degree of fromNode = 4; (2) degree of toNode = 4; (3) both from & to = 4;
  98. angAcc_oneHot_ln = nan(subLen, BinL);
  99. for iSub = 1 : subLen
  100. disp([netWord, '-subj', num2str(iSub)]);
  101. expMode = expMode_subj{iSub};
  103. % data from the current participant
  104. clusterResult = clusterResult_iExp(clusterResult_iExp.subjID == iSub, :);
  106. %% data from the clusterResult
  107. nTrials = size(clusterResult, 1);
  108. trials_Col = clusterResult.trialNo;
  109. blockNo_col = clusterResult.blockNo;
  110. respRT_Col = clusterResult.clickRT;
  111. respSig_Col = clusterResult.clickConds;
  112. from_nodes = clusterResult.cue;
  113. to_nodes = clusterResult.target;
  114. dtr_nodes = clusterResult.distractor;
  115. choiceId = cell2mat(clusterResult.choiceId); % nan(length(stim), length(circle_list));
  116. choiceId_trials = clusterResult.('choice-Final'); % nan(length(stim), 1);
  117. choiceId_trials_exp{iSub, iExp} = choiceId_trials;
  119. %% ------ Clicked-based measures ------
  120. %%% total accuracy
  121. clickAcc_exp(iSub, 1, iExp) = length(find(respSig_Col == 0 | respSig_Col == 1)) / length(respSig_Col); %% accuracy across all trials
  122. clickAcc_exp(iSub, 2, iExp) = length(find(respSig_Col == 0)) / length(find(respSig_Col == 0 | respSig_Col == 2)); %% accuracy in response before cue trials
  123. clickAcc_exp(iSub, 3, iExp) = length(find(respSig_Col == 1)) / length(find(respSig_Col == 1 | respSig_Col == 3)); %% accuracy in response after cue trials
  125. %% ------ Mouse-trajectory based choice ------
  126. %% merge all trials
  127. for iTp = 1 : length(circle_list)
  128. choiceId_i = choiceId(:, iTp);
  129. choiceId_i(isnan(choiceId_i)) = [];
  130. angAcc(iSub, iTp) = length(find(choiceId_i == 1)) / length(choiceId_i);
  131. lenData(iSub, iTp) = length(choiceId_i);
  132. end
  134. %% accuracy in different node degrees when subjects made a response or at t=0.8s
  135. dgr_i = 4;
  136. nodeFind = find(nodeDegree == dgr_i);
  137. %%% degree of from nodes = 4
  138. trlIdx = arrayfun(@(x) ismember(x, nodeFind), from_nodes); %% including correct and incorrect responses
  139. choiceId_from = choiceId_trials(trlIdx);
  140. choiceId_from(isnan(choiceId_from)) = [];
  141. angAcc_dgr_oneHot(iSub, 1) = length(find(choiceId_from == 1)) / length(choiceId_from);
  142. dgr_oneHot_N_exp(iSub, 1, iExp) = length(find(trlIdx == 1));
  143. dgr_oneHot_trlIdx_exp{iSub, 1, iExp} = find(trlIdx == 1);
  145. %%% degree of to nodes = 4
  146. trlIdx = arrayfun(@(x) ismember(x, nodeFind), to_nodes);
  147. choiceId_to = choiceId_trials(trlIdx);
  148. choiceId_to(isnan(choiceId_to)) = [];
  149. angAcc_dgr_oneHot(iSub, 2) = length(find(choiceId_to == 1)) / length(choiceId_to);
  150. dgr_oneHot_N_exp(iSub, 2, iExp) = length(find(trlIdx == 1));
  151. dgr_oneHot_trlIdx_exp{iSub, 2, iExp} = find(trlIdx == 1);
  153. %%% both the degree of fromNode & toNode equal 4
  154. trlIdx_from = arrayfun(@(x) ismember(x, nodeFind), from_nodes);
  155. trlIdx_to = arrayfun(@(x) ismember(x, nodeFind), to_nodes);
  156. trlIdx = find(trlIdx_from == 1 & trlIdx_to == 1);
  157. choiceId_both = choiceId_trials(trlIdx);
  158. choiceId_both(isnan(choiceId_both)) = [];
  159. angAcc_dgr_oneHot(iSub, 3) = length(find(choiceId_both == 1)) / length(choiceId_both);
  160. dgr_oneHot_N_exp(iSub, 3, iExp) = length(trlIdx);
  161. dgr_oneHot_trlIdx_exp{iSub, 3, iExp} = trlIdx;
  163. %% learning effect: accuracy in sliding bins at decision time
  164. for iB = 1 : BinL
  165. trlIdx = (iB - 1) * nBin + 1 : iB * nBin;
  166. choiceId_trl = choiceId_trials(trlIdx);
  167. choiceId_trl(isnan(choiceId_trl)) = [];
  168. angAcc_oneHot_ln(iSub, iB) = length(find(choiceId_trl == 1)) / length(choiceId_trl);
  169. end
  171. %% merge all trials: before-cue-response or at 0.8 s
  172. choiceId_trials(isnan(choiceId_trials)) = [];
  173. angAcc_oneHot(iSub) = length(find(choiceId_trials == 1)) / length(choiceId_trials);
  175. end
  176. angAcc_exp(:, :, iExp) = angAcc; %% angAcc = zeros(subLen, length(circle_list));
  177. angAcc_oneHot_exp(:, iExp) = angAcc_oneHot; %% angAcc_oneHot = zeros(subLen, 1);
  178. angAcc_dgr_oneHot_exp(:, :, iExp) = angAcc_dgr_oneHot; %% angAcc_dgr_oneHot = zeros(subLen, 3);
  179. angAcc_oneHot_ln_exp(:, :, iExp) = angAcc_oneHot_ln; %% angAcc_oneHot_ln = nan(subLen, BinL);
  180. end
  182. %% color settings
  183. colorSets = [249, 183, 176; ...
  184. 138, 170, 51; ...
  185. 84, 185, 211; ...
  186. 248, 218, 172; ...
  187. 184, 204, 225; ...
  188. 210, 234, 200; ...
  189. 198, 127, 192; ...
  190. 219, 204, 226] ./ [255, 255, 255];
  192. %% ------ Figure 2d: Time-resolved prediction accuracy of four networks ------
  193. % & %% ------ Figure S3A: Decision preference within a trial ------
  194. datFlg = 1;
  195. if datFlg == 1 % all 40 subjects
  196. angAcc_tc = angAcc_exp;
  197. elseif datFlg == 2 % 20 in-lab subjects
  198. angAcc_tc = angAcc_exp(1 : 20, :, :);
  199. elseif datFlg == 3 % 20 online subjects
  200. angAcc_tc = angAcc_exp(21 : end, :, :);
  201. end
  202. stat_mat = zeros(length(circle_list), 2, nExp); % 4 networks
  203. for iExp = 1 : nExp
  204. for iTime = 1 : length(circle_list)
  205. angAcc_exp_i = angAcc_tc(:, iTime, iExp);
  206. [h, p, ci, stats] = ttest(angAcc_exp_i, 1/2, 'Tail', 'both');
  207. stat_mat(iTime, 1, iExp) = p;
  208. stat_mat(iTime, 2, iExp) = stats.tstat;
  209. end
  210. end
  212. LineSty = '-';
  213. for iExp = 1 : nExp
  214. figure('Position', [100 100 240 160]), clf;
  215. angAcc_exp_i = angAcc_tc(:, :, iExp);
  216. for iSub = 1 : size(angAcc_exp_i, 1)
  217. plot(circle_list, angAcc_exp_i(iSub, :), 'Color', [0.6, 0.6, 0.6], 'LineStyle', '-', 'LineWidth', 0.5); hold on;
  218. end
  219. [acc_avg, acc_sem] = Mean_and_Se(angAcc_exp_i, 1);
  220. shadedErrorBar(circle_list, acc_avg, acc_sem, {'Color', colorSets(iExp, :), 'MarkerFaceColor', colorSets(iExp, :), 'LineStyle', LineSty, 'LineWidth', 3}, 0.5); hold on;
  221. ylim([0.4, 1]);
  222. ylimit = ylim;
  223. xLoc = 0.99;
  224. % mark the significance
  225. [~, ~, ~, adj_p] = fdr_bh(squeeze(stat_mat(:, 1, iExp)), 0.05, 'pdep');
  226. for iCir = 1 : length(circle_list)
  227. pval_j = adj_p(iCir);
  228. if pval_j < 0.05
  229. plot(circle_list(iCir), xLoc * ylimit(end), 'Marker', '.', 'MarkerSize', 4, 'Color', colorSets(iExp, :), 'MarkerFaceColor', colorSets(iExp, :), 'LineStyle', 'none'); hold on;
  230. end
  231. end
  232. plot(xlim, [1/2, 1/2], 'k--', 'LineWidth', 1); hold on;
  233. plot([0.8, 0.8], ylim, 'k--', 'LineWidth', 1); hold on;
  234. ylim([0.4, 1]);
  236. set(gca, 'LineWidth', 2);
  237. set(gca, 'FontSize', 14, 'FontWeight', 'bold', 'FontName', 'Arial');
  238. set(gca, 'XTick', 0 : 0.4 : 1.5, 'XTickLabel', [0, 0.4, 0.8, 1.2]);
  239. set(gca, 'YTick', 0.4 : 0.1 : 1, 'YTickLabel', {'', '0.5', '', '', '', '', '1'});
  241. box off;
  242. end
  244. %% ------ Figure 2e: Overall prediction accuracy of four networks ------
  245. % & %% ------ Figure S3B: Overall accuracy comparison of the four networks (in-lab and online separately) ------
  246. datFlg = 1;
  247. if datFlg == 1 % all 40 subjects
  248. angAcc_oneHot_tc = angAcc_oneHot_exp;
  249. elseif datFlg == 2 % 20 in-lab subjects
  250. angAcc_oneHot_tc = angAcc_oneHot_exp(1 : 20, :);
  251. elseif datFlg == 3 % 20 online subjects
  252. angAcc_oneHot_tc = angAcc_oneHot_exp(21 : end, :);
  253. end
  255. stat_mat = zeros(nExp, 2); % 4 networks
  256. for iExp = 1 : nExp
  257. [h, p, ci, stats] = ttest(angAcc_oneHot_tc(:, iExp), 1/2, 'Tail', 'both');
  258. stat_mat(iExp, 1) = p;
  259. stat_mat(iExp, 2) = stats.tstat;
  260. end
  261. [~, ~, ~, adj_p] = fdr_bh(stat_mat(:, 1), 0.05, 'pdep');
  263. figure('Position', [100 100 420 220]), clf;
  264. vs = violinplot(angAcc_oneHot_tc, [1,2,3,4], 'Width', 0.3, 'ViolinColor', colorSets, 'ViolinAlpha', 0.1, 'MarkerSize', 30,...
  265. 'MedianMarkerSize', 100, 'EdgeColor', [0,0,0], 'BoxColor', [0,0,0]);
  266. xlim([0.5, 4.5]);
  267. ylim([0.45, 0.7]);
  268. set(gca, 'FontName', 'Arial', 'FontSize', 14, 'FontWeight', 'Bold', 'LineWidth', 2);
  269. set(gca, 'XTick', '', 'XTickLabel', '');
  270. set(gca, 'YTick', 0.5 : 0.1 : 0.7, 'YTickLabel', 0.5 : 0.1 : 0.7);
  271. plot(xlim, [0.5, 0.5], 'k--', 'LineWidth', 1); hold on;
  272. box off;
  274. %% ------ Figure S4: Learning curves for the 4 transition networks in Experiment 1 ------
  275. datFlg = 1;
  276. if datFlg == 1 % all 40 subjects
  277. angAcc_oneHot_ln_tc = angAcc_oneHot_ln_exp;
  278. elseif datFlg == 2 % 20 in-lab subjects
  279. angAcc_oneHot_ln_tc = angAcc_oneHot_ln_exp(1 : 20, :, :);
  280. elseif datFlg == 3 % 20 online subjects
  281. angAcc_oneHot_ln_tc = angAcc_oneHot_ln_exp(21 : end, :, :);
  282. end
  284. [data_avg, data_sem] = Mean_and_Se(angAcc_oneHot_ln_tc, 1);
  285. data_avg = squeeze(data_avg); % BinL * nExp
  286. data_sem = squeeze(data_sem);
  287. figure('Position', [100 100 300 200]), clf;
  288. for iExp = 1 : length(network_list)
  289. errorbar(1 : 1 : BinL, data_avg(:, iExp), data_sem(:, iExp), 'Color', colorSets(iExp, :), 'LineStyle', '-', 'LineWidth', 3); hold on;
  290. plot(1 : 1 : BinL, data_avg(:, iExp), 'Marker', '.', 'MarkerSize', 15, 'Color', colorSets(iExp, :), 'LineStyle', 'none'); hold on;
  291. end
  292. xlim([1-0.5, BinL+0.5]);
  293. ylim([0.4, 0.7]);
  294. plot(xlim, [1/2, 1/2], 'k--', 'LineWidth', 1); hold on;
  295. set(gca, 'FontSize', 14, 'FontWeight', 'Bold', 'LineWidth', 2);
  296. set(gca, 'XTick', 1 : 1 : BinL, 'XTickLabel', 1 : 1 : BinL);
  297. set(gca, 'YTick', 0.4 : 0.1 : 0.7, 'YTickLabel', 0.4 : 0.1 : 0.7);
  298. box off;
  300. %% ------ Figure S5A: The numbers of trials under each node-degree matched condition ------
  301. figure('Position', [100 100 120 220]), clf;
  302. for iExp = 1 : nExp
  303. dgr_matched_N = dgr_oneHot_N_exp(:, :, iExp); % 3 columns: (1) from-node matched; (2) to-node matched; (3) both from- and to-node matched
  304. [Nmean, Nsem] = Mean_and_Se(dgr_matched_N, 1);
  305. plot(1 : 1 : 3, Nmean, 'Marker', '.', 'MarkerSize', 35, 'Color', colorSets(iExp, :), 'lineStyle', '-', 'LineWidth', 3); hold on;
  306. for i = 1 : 3
  307. errorbar(i, Nmean(i), Nsem(i), 'Color', colorSets(iExp, :), 'LineStyle', '-', 'LineWidth', 3); hold on;
  308. end
  309. end
  310. xlim([1-0.5, 3+0.5]);
  311. ylim([0, 1000]);
  312. set(gca, 'FontSize', 14, 'FontWeight', 'Bold', 'LineWidth', 2);
  313. set(gca, 'XTick', 1 : 1 : 3, 'XTickLabel', {'Cue', 'Target', 'Cue & target'});
  314. set(gca, 'YTick', 0 : 500 : 1000, 'YTickLabel', 0 : 500 : 1000);
  315. box off;
  317. %% ------ Figure 5B: Overall prediction accuracy of four networks with three types of node-degree matched conditions ------
  318. plotFlg = 1; % 1: degree of fromNode = 4; 2: degree of toNode = 4; 3: both from & to = 4;
  319. dataPlot = squeeze(angAcc_dgr_oneHot_exp(:, plotFlg, :));
  320. figure('Position', [100 100 420 220]), clf;
  321. vs = violinplot(dataPlot, [1,2,3,4], 'Width', 0.3, 'ViolinColor', colorSets, 'ViolinAlpha', 0.1, 'MarkerSize', 30,...
  322. 'MedianMarkerSize', 100, 'EdgeColor', [0,0,0], 'BoxColor', [0,0,0]);
  323. xlim([0.5, 4.5]);
  324. set(gca, 'FontName', 'Arial', 'FontSize', 14, 'FontWeight', 'Bold', 'LineWidth', 2);
  325. set(gca, 'XTick', '', 'XTickLabel', '');
  326. if plotFlg == 1 || plotFlg == 2
  327. ylim([0.4, 0.7]);
  328. set(gca, 'YTick', 0.4 : 0.1 : 0.7, 'YTickLabel', 0.4 : 0.1 : 0.7);
  329. elseif plotFlg == 3
  330. ylim([0.3, 0.8]);
  331. set(gca, 'YTick', 0.3 : 0.1 : 0.8, 'YTickLabel', 0.3 : 0.1 : 0.8);
  332. end
  333. plot(xlim, [0.5, 0.5], 'k--', 'LineWidth', 1); hold on;
  334. box off;
  336. %% ------ Figure S6A: Proportion of above-chance subjects ------
  337. % ------ convert the accuracy per time point across participants into
  338. % proportation of above-chance participants ------
  339. propSubj_aboveChance_time = zeros(nExp, length(circle_list));
  340. for iExp = 1 : nExp
  341. angAcc_exp_i = angAcc_exp(:, :, iExp);
  342. propSubj_count = (~isnan(angAcc_exp_i) & angAcc_exp_i > 0.5);
  343. propSubj_aboveChance_time(iExp, :) = sum(propSubj_count, 1) / size(propSubj_count, 1);
  344. end
  345. LineSty = '-';
  346. figure('Position', [100 100 240 160]), clf;
  347. for iExp = 1 : nExp
  348. plot(circle_list(2 : end), propSubj_aboveChance_time(iExp, 2 : end), 'Color', colorSets(iExp, :), 'LineStyle', '-', 'LineWidth', 2); hold on;
  349. ylim([0, 1]);
  350. ylimit = ylim;
  352. plot(xlim, [1/2, 1/2], 'k--', 'LineWidth', 1); hold on;
  353. plot([0.8, 0.8], ylim, 'k--', 'LineWidth', 1); hold on;
  354. ylim([0, 1]);
  356. set(gca, 'LineWidth', 2);
  357. set(gca, 'FontSize', 14, 'FontWeight', 'bold', 'FontName', 'Arial');
  358. set(gca, 'XTick', 0 : 0.4 : 1.5, 'XTickLabel', [0, 0.4, 0.8, 1.2]);
  359. set(gca, 'YTick', 0 : 0.5 : 1, 'YTickLabel', [0, 0.5, 1]);
  360. box off;
  361. end
  363. %% ------ Figure S6B: Time-resolved prediction accuracy for high- and low-performance subjects of four networks ------
  364. % For each group, splitting participants into high and low performance and
  365. % redraw the accuracy time curve
  366. HighLow_idx_exp = nan(20, 2, nExp); % 2: high vs. low performance participants
  367. angAcc_HighLow_time = nan(20, length(circle_list), 2, nExp);
  368. angAcc_oneHot_HighLow = nan(20, 2, nExp);
  369. stat_mat_HighLow = nan(length(circle_list), 2, 2, nExp);
  370. for iExp = 1 : nExp
  371. % ------ Sort based on one-hot metric ------
  372. angAcc_iExp = angAcc_oneHot_exp(:, iExp);
  373. [~, angAcc_idx] = sort(angAcc_iExp, 'descend'); % default: 'ascend'
  374. split_half = ceil(length(angAcc_idx) / 2);
  375. high_idx = angAcc_idx(1 : split_half);
  376. low_idx = angAcc_idx((split_half + 1) : end);
  377. % ------ assign the participants index for High- and Low-performance group ------
  378. HighLow_idx_exp(:, 1, iExp) = high_idx;
  379. HighLow_idx_exp(:, 2, iExp) = low_idx;
  381. % ------ Reassign participants into high- vs. low-performance group ------
  382. angAcc_HighLow_time(:, :, 1, iExp) = angAcc_exp(high_idx, :, iExp);
  383. angAcc_HighLow_time(:, :, 2, iExp) = angAcc_exp(low_idx, :, iExp);
  384. angAcc_oneHot_HighLow(:, 1, iExp) = angAcc_oneHot_exp(high_idx, iExp);
  385. angAcc_oneHot_HighLow(:, 2, iExp) = angAcc_oneHot_exp(low_idx, iExp);
  387. % Statistical test
  388. for ii = 1 : 2 % High and Low-performance participants
  389. for iTime = 1 : length(circle_list)
  390. angAcc_exp_i = angAcc_HighLow_time(:, iTime, ii, iExp);
  391. [h, p, ci, stats] = ttest(angAcc_exp_i, 1/2, 'Tail', 'both');
  392. stat_mat_HighLow(iTime, 1, ii, iExp) = p;
  393. stat_mat_HighLow(iTime, 2, ii, iExp) = stats.tstat;
  394. end
  395. end
  396. end
  397. %%
  398. LineSty_list = {'-', ':'};
  399. for iExp = 1 : nExp
  400. figure('Position', [100 100 240 160]), clf;
  401. angAcc_exp_i = angAcc_exp(:, :, iExp);
  402. for iSub = 1 : size(angAcc_exp_i, 1)
  403. plot(circle_list, angAcc_exp_i(iSub, :), 'Color', [0.6, 0.6, 0.6], 'LineStyle', '-', 'LineWidth', 0.5); hold on;
  404. end
  405. for ii = 1 : 2 % high- vs. low-performance participants
  406. [acc_avg, acc_sem] = Mean_and_Se(angAcc_HighLow_time(:, :, ii, iExp), 1);
  407. shadedErrorBar(circle_list(2 : end), acc_avg(2 : end), acc_sem(2 : end), {'Color', colorSets(iExp, :), 'MarkerFaceColor', colorSets(iExp, :), 'LineStyle', LineSty_list{ii}, 'LineWidth', 3}, 0.5); hold on;
  408. end
  410. ylim([0.4, 1]);
  411. ylimit = ylim;
  412. xLoc = 0.99;
  413. for ii = 1 : 2 % High and Low-performance participants
  414. % mark the significance
  415. [~, ~, ~, adj_p] = fdr_bh(squeeze(stat_mat_HighLow(:, 1, ii, iExp)), 0.05, 'pdep'); % method: 'dep', 'pdep'
  416. for iCir = 1 : length(circle_list)
  417. pval_j = adj_p(iCir);
  418. if pval_j < 0.05
  419. plot(circle_list(iCir), xLoc * ylimit(end), 'Marker', '.', 'MarkerSize', 4, 'Color', colorSets(iExp, :), 'MarkerFaceColor', colorSets(iExp, :), 'LineStyle', 'none'); hold on;
  420. end
  421. end
  422. xLoc = xLoc - 0.03;
  423. end
  424. plot(xlim, [1/2, 1/2], 'k--', 'LineWidth', 1); hold on;
  425. plot([0.8, 0.8], ylim, 'k--', 'LineWidth', 1); hold on;
  426. ylim([0.4, 1]); % left axis limits
  428. set(gca, 'LineWidth', 2);
  429. set(gca, 'FontSize', 14, 'FontWeight', 'bold', 'FontName', 'Arial');
  430. set(gca, 'XTick', 0 : 0.4 : 1.5, 'XTickLabel', [0, 0.4, 0.8, 1.2]);
  431. set(gca, 'YTick', 0.4 : 0.1 : 1, 'YTickLabel', {'', '0.5', '', '', '', '', '1'});
  433. box off;
  434. ax = gca;
  435. end
  437. %% ------ Figure S6C: Overall accuracy for high- and low-performance subjects across four networks ------
  438. for ii = 1 : 2 % High and Low-performance participants
  439. figure('Position', [100 100 420 220]), clf;
  440. vs = violinplot(squeeze(angAcc_oneHot_HighLow(:, ii, :)), [1,2,3,4], 'Width', 0.3, 'ViolinColor', colorSets, 'ViolinAlpha', 0.1, 'MarkerSize', 30,...
  441. 'MedianMarkerSize', 100, 'EdgeColor', [0,0,0], 'BoxColor', [0,0,0]);
  442. xlim([0.5, 4.5]);
  443. ylim([0.4, 0.7]);
  444. set(gca, 'FontName', 'Arial', 'FontSize', 14, 'FontWeight', 'Bold', 'LineWidth', 2);
  445. set(gca, 'XTick', '', 'XTickLabel', '');
  446. set(gca, 'YTick', 0.4 : 0.1 : 0.7, 'YTickLabel', 0.4 : 0.1 : 0.7);
  447. plot(xlim, [0.5, 0.5], 'k--', 'LineWidth', 1); hold on;
  448. box off;
  449. end
  451. %% ------ Figure S8A: Click-accuracy from before-hint-response trials in Experiment 1 ------
  452. % clickAcc_exp = zeros(subLen, 3, nExp); % 3: 1) overall accuracy; 2) before-hint accuracy; 3) after-hint accuracy
  453. metric_plot_category = squeeze(clickAcc_exp(:, 2, :)); % before-hint click accuracy
  455. figure('Position', [100 100 420 220]), clf;
  456. vs = violinplot(metric_plot_category, [1,2,3,4], 'Width', 0.3, 'ViolinColor', colorSets, 'ViolinAlpha', 0.1, 'MarkerSize', 30,...
  457. 'MedianMarkerSize', 100, 'EdgeColor', [0,0,0], 'BoxColor', [0,0,0]);
  458. xlim([0.5, 4.5]);
  459. ylim([0, 1]);
  460. set(gca, 'FontName', 'Arial', 'FontSize', 14, 'FontWeight', 'Bold', 'LineWidth', 2);
  461. set(gca, 'XTick', '', 'XTickLabel', '');
  462. set(gca, 'YTick', 0 : 0.2 : 1, 'YTickLabel', 0 : 0.2 : 1);
  463. plot(xlim, [0.5, 0.5], 'k--', 'LineWidth', 1); hold on;
  464. box off;
  465. ax = gca;

Fig2andSI.m, no license · at the source

Overview

Authors: Xiangjuan Ren1,2,3,4,5, Muzhi Wang1,2,3,6, Tingting Qin7, Fang Fang1,2,3,8, Aming Li7,9,10, Huan Luo1,2,3
ORCID iDs: Xiangjuan Ren, Muzhi Wang, Fang Fang, Huan Luo
  1. School of Psychological and Cognitive Sciences, Peking University, Beijing, China
  2. PKU-IDG/McGovern Institute for Brain Research, Peking University, Beijing, China
  3. Key Laboratory of Machine Perception (Ministry of Education), Peking University, Beijing, China
  4. Latent Learning Lab, Institute of Psychology, Universität Hamburg, Hamburg, Germany
  5. Max Planck Institute for Human Development, Berlin, Germany
  6. Applied Computational Psychiatry Lab, Max Planck UCL Centre for Computational Psychiatry and Ageing Research, Queen Square Institute of Neurology, UCL, London, UK
  7. Center for Systems and Control, School of Advanced Manufacturing and Robotics, Peking University, Beijing, China
  8. Peking-Tsinghua Center for Life Sciences, Peking University, Beijing, China
  9. Research Center for Robotics, Peking University, Beijing, China
  10. Center for Multi-Agent Research, Institute for Articial Intelligence, Peking University, Beijing, China
Institutions: Universität Hamburg (Germany); Peking University (China); Max Planck Institute for Human Development (Germany); UCL Queen Square Institute of Neurology (United Kingdom); University College London (United Kingdom); Center for Life Sciences (China)
Journal: Nature communications, volume 17, issue 1, article 8771
Dates: received 12 September 2025; accepted 14 July 2026; published online 17 July 2026
Type: Research article · Language: English
License: CC BY-NC-ND
Identifiers: DOI 10.1038/s41467-026-75843-7 · PMID 42469233 · PMCID PMC13494043 · OpenAlex W7169508696
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: MEG (modality), human (organism), cognitive (subfield)
Methods: Connectivity, Spectral & time-frequency, Statistics, Smoothing, state filtering, decompositions, Graphs, Machine learning
Keywords: Human behaviour, Cognitive neuroscience
MeSH: Knowledge*, Learning*, Neural Networks, Computer*, Computer Simulation, Gyrus Cinguli, Humans, Magnetoencephalography, Nerve Net (* major topic)
Topic: Neural Networks and Reservoir Computing (Artificial Intelligence, Computer Science), according to OpenAlex
Funding: National Natural Science Foundation of China (National Science Foundation of China) (32541013, T2525017, 62533002, 62173004, T2421004); European Research Council (852669)
Citations: cited by 1 paper (Europe PMC); 108 references in the paper

Abstract

The abstract is not reproduced here: the paper's license (CC BY-NC-ND) does not allow it. Read it in the paper, at the publisher or on Europe PMC.

Repository

Its files are read in the Code ↔ Paper reader above, with 11 matches between paragraphs and lines of code.

OSF dm3z9

License: none: the authors keep all their rights
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Languages: MATLAB (34)
Size: 45 files, 34 scripts
Software Heritage: not checked
Found in: “Code availability”
Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation
Tools: Statistics and Machine Learning Toolbox (7 files), fdr_bh (Benjamini-Hochberg FDR) (2 files), shadedErrorBar (2 files), Violinplot-Matlab (2 files)
Availability: 1 check, the latest on 27 September 2026: the link answers (HTTP 200)
  • 27 September 2026: the link answers (HTTP 200)
34 files
At the source:

Code availability statement

The paper has a code availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:

  • it points to the authors' code: OSF dm3z9

Read it in the paper: doi.org/10.1038/s41467-026-75843-7.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 34 scripts, each with its path and the digest of its content;
  • 11 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability statement

The paper has a data availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:

  • no repository, dataset or request procedure was recognized in it

Read it in the paper: doi.org/10.1038/s41467-026-75843-7.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 6 authors, 2 keywords, 8 MeSH terms, 2 funders, 89 references.

Cite

This paper

Ren, X., Wang, M., Qin, T., Fang, F., Li, A., & Luo, H. (2026). Compressive learning scaffolds higher-order network structure to enhance human knowledge acquisition. Nature communications, 17(1), 8771. https://doi.org/10.1038/s41467-026-75843-7

BibTeX

@article{ren2026compressive,
author = {Ren, Xiangjuan and Wang, Muzhi and Qin, Tingting and Fang, Fang and Li, Aming and Luo, Huan},
title = {{Compressive learning scaffolds higher-order network structure to enhance human knowledge acquisition}},
journal = {Nature communications},
year = {2026},
month = jul,
volume = {17},
number = {1},
pages = {8771},
publisher = {Nature Publishing Group},
issn = {2041-1723},
doi = {10.1038/s41467-026-75843-7},
url = {https://doi.org/10.1038/s41467-026-75843-7},
pmid = {42469233},
pmcid = {PMC13494043}
}

RIS

TY - JOUR
AU - Ren, Xiangjuan
AU - Wang, Muzhi
AU - Qin, Tingting
AU - Fang, Fang
AU - Li, Aming
AU - Luo, Huan
TI - Compressive learning scaffolds higher-order network structure to enhance human knowledge acquisition
T2 - Nature communications
J2 - Nat Commun
PY - 2026
DA - 2026/07/17
VL - 17
IS - 1
SP - 8771
SN - 2041-1723
PB - Nature Publishing Group
DO - 10.1038/s41467-026-75843-7
UR - https://doi.org/10.1038/s41467-026-75843-7
LA - en
ER -

CSL-JSON

{
"id": "10.1038/s41467-026-75843-7",
"type": "article-journal",
"title": "Compressive learning scaffolds higher-order network structure to enhance human knowledge acquisition",
"container-title": "Nature communications",
"author": [
{
"family": "Ren",
"given": "Xiangjuan"
},
{
"family": "Wang",
"given": "Muzhi"
},
{
"family": "Qin",
"given": "Tingting"
},
{
"family": "Fang",
"given": "Fang"
},
{
"family": "Li",
"given": "Aming"
},
{
"family": "Luo",
"given": "Huan"
}
],
"container-title-short": "Nat Commun",
"volume": "17",
"issue": "1",
"page": "8771",
"DOI": "10.1038/s41467-026-75843-7",
"PMID": "42469233",
"PMCID": "PMC13494043",
"ISSN": "2041-1723",
"publisher": "Nature Publishing Group",
"URL": "https://doi.org/10.1038/s41467-026-75843-7",
"language": "en",
"issued": {
"date-parts": [
[
2026,
7,
17
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1371/journal.pbio.3003756
Human brains construct individualized global rankings from identical few-shot learning input.
Journal: PLoS biology
In common: MEG, cognitive, 3 references, 2 authors
[2] doi:10.1016/j.cub.2026.05.068 [code]
An abstract relational map emerges in the human medial prefrontal cortex with consolidation.
Journal: Current biology : CB
In common: Statistics and Machine Learning Toolbox, 9 references
[3] doi:10.1038/s41597-026-07350-9 [code]
An open multi-center MEG-EEG dataset for studying conscious visual perception.
Journal: Scientific data
In common: Statistics and Machine Learning Toolbox, MEG, 1 reference, author Huan Luo
[4] doi:10.1038/s41467-026-71428-6 [code]
Binding items to contexts through conjunctive neural representations with the method of loci.
Journal: Nature communications
In common: 6 references
[5] doi:10.7554/elife.108023 [code]
Challenges in replay detection by TDLM in post-encoding resting state.
Journal: eLife
In common: shadedErrorBar, Statistics and Machine Learning Toolbox, MEG, 3 references
[6] doi:10.1038/s41467-026-71604-8 [code]
Respiration as a dynamic modulator of sensory sampling.
Journal: Nature communications
In common: Violinplot-Matlab, fdr_bh (Benjamini-Hochberg FDR), Statistics and Machine Learning Toolbox, MEG, cognitive, 1 reference
[7] doi:10.1038/s41467-026-72510-9 [code]
Single-neuron network topology governs neural computation and learning in primate cortex.
Journal: Nature communications
In common: Statistics and Machine Learning Toolbox, cognitive, 1 reference, author Fang Fang
[8] doi:10.1371/journal.pbio.3003938 [code]
Theta oscillations tag episodic memories for sleep-dependent consolidation.
Journal: PLoS biology
In common: fdr_bh (Benjamini-Hochberg FDR), shadedErrorBar, Statistics and Machine Learning Toolbox, cognitive, 1 reference
[9] doi:10.7554/elife.103689 [code]
Dissociable dynamic effects of expectation during statistical learning.
Journal: eLife
In common: Violinplot-Matlab, shadedErrorBar, Statistics and Machine Learning Toolbox, 1 reference
[10] doi:10.1002/aur.70312 [code]
Aberrant Neural Entrainment to Word-Level Speech Patterns in Fragile X Syndrome: Evidence for a Statistical Learning Deficit.
Journal: Autism research : official journal of the International Society for Autism Research
In common: shadedErrorBar, Statistics and Machine Learning Toolbox, cognitive, 2 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.