% Compare the speeds of various inference engines on the BAT DBN[bnet, names] = mk_bat_dbn;T = 3; % fixed length sequence - we make it short just for speedUSEC = exist('@jtree_C_inf_engine/collect_evidence','file');disp('constructing engines for BAT');engine = {}; % time in seconds for inferenceengine{end+1} = jtree_unrolled_dbn_inf_engine(bnet, T, 'useC', USEC);  % 0.39engine{end+1} = smoother_engine(jtree_2TBN_inf_engine(bnet)); % 4.89engine{end+1} = jtree_dbn_inf_engine(bnet); % 4.45if 0engine{end+1} = jtree_ndx_dbn_inf_engine(bnet, 'ndx_type', 'SD'); % 2.98engine{end+1} = jtree_ndx_dbn_inf_engine(bnet, 'ndx_type', 'D'); % 3.52engine{end+1} = jtree_ndx_dbn_inf_engine(bnet, 'ndx_type', 'B'); % 2.40if USEC, engine{end+1} = jtree_C_dbn_inf_engine(bnet); end % 3.54%engine{end+1} = hmm_inf_engine(bnet, onodes); % too bigend%tic; engine{end+1} = frontier_inf_engine(bnet); toc % very slow% The frontier engine thrashes badly on the BAT network%tic; engine{end+1} = bk_inf_engine(bnet, 'exact', onodes); toc  % SLOW!%tic; engine{end+1} = bk_inf_engine(bnet, 'ff', onodes); toc%clusters{1} = [stringmatch({'LeftClr', 'RightClr', 'LatAct', 'Xdot', 'InLane'}, names)];%clusters{2} = [stringmatch({'FwdAct', 'Ydot', 'Stopped', 'EngStatus', 'FBStatus'}, names)];      %tic; engine{end+1} = bk_inf_engine(bnet, clusters, onodes); tocdisp('inference')time = cmp_inference_dbn(bnet, engine, T)disp('learning')time = cmp_learning_dbn(bnet, engine, T)