notesum.ai

Published at November 7

Series-to-Series Diffusion Bridge Model

cs.LG

cs.AI

Released Date: November 7, 2024

Authors: Hao Yang¹, Zhanbo Feng², Feng Zhou³, Robert C Qiu⁴, Zenan Ling⁴

Aff.: ¹School of AIA, Huazhong University of Science and Technology; ²Department of Computer Science and Engineering, Shanghai Jiao Tong University; ³Center for Applied Statistics and School of Statistics, Renmin University of China; ⁴School of EIC, Huazhong University of Science and Technology

Arxiv: http://arxiv.org/abs/2411.04491v1

		Diffusion-based Methods						Transformer-based Methods						Linear Model
Methods		Ours		CSDI		TMDM		Autoformer		Informer		iTransformer		NLinear		DLinear		RLinear
Metric		MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE
ETTh1	96	0.366	0.383	0.744	0.623	0.711	0.605	0.429	0.444	0.925	0.761	0.387	0.405	0.374	0.394	0.384	0.405	0.366	0.391
	192	0.405	0.407	0.952	0.715	0.922	0.720	0.440	0.451	0.995	0.778	0.441	0.436	0.408	0.415	0.443	0.450	0.403	0.412
	336	0.442	0.430	1.192	0.837	0.990	0.737	0.511	0.488	1.036	0.782	0.491	0.462	0.429	0.428	0.447	0.448	0.420	0.423
	720	0.469	0.478	1.822	1.005	1.152	0.836	0.499	0.501	1.175	0.858	0.509	0.494	0.441	0.454	0.504	0.515	0.442	0.456
ETTh2	96	0.274	0.331	1.017	0.729	0.496	0.510	0.418	0.445	3.017	1.369	0.301	0.350	0.283	0.343	0.290	0.353	0.262	0.331
	192	0.354	0.388	3.417	1.356	0.578	0.535	0.435	0.439	6.348	2.105	0.380	0.399	0.356	0.385	0.388	0.422	0.320	0.374
	336	0.433	0.454	2.642	1.216	0.715	0.598	0.480	0.481	5.628	1.998	0.424	0.432	0.362	0.403	0.463	0.473	0.326	0.388
	720	0.592	0.568	3.396	1.431	0.758	0.658	0.478	0.487	4.110	1.692	0.430	0.447	0.398	0.437	0.733	0.606	0.425	0.449
ETTm1	96	0.293	0.333	0.556	0.509	0.547	0.512	0.471	0.463	0.621	0.557	0.342	0.377	0.306	0.348	0.301	0.345	0.301	0.343
	192	0.333	0.355	0.608	0.532	0.689	0.592	0.592	0.521	0.723	0.618	0.383	0.396	0.349	0.375	0.336	0.366	0.341	0.367
	336	0.367	0.377	0.764	0.622	0.722	0.602	0.503	0.486	1.001	0.746	0.418	0.418	0.375	0.388	0.372	0.389	0.374	0.386
	720	0.442	0.422	1.071	0.792	1.072	0.785	0.751	0.582	0.980	0.747	0.487	0.457	0.433	0.422	0.427	0.423	0.430	0.418
ETTm2	96	0.164	0.249	0.859	0.587	0.328	0.400	0.233	0.313	0.407	0.482	0.186	0.272	0.167	0.255	0.172	0.267	0.164	0.253
	192	0.219	0.292	0.907	0.614	0.415	0.423	0.278	0.336	0.807	0.706	0.254	0.314	0.221	0.293	0.237	0.314	0.219	0.290
	336	0.274	0.328	1.584	0.862	0.871	0.611	0.379	0.394	1.453	0.926	0.316	0.351	0.274	0.327	0.295	0.359	0.273	0.326
	720	0.361	0.389	2.692	1.202	1.101	0.739	0.584	0.473	3.930	1.469	0.414	0.407	0.369	0.385	0.427	0.439	0.366	0.385
ILI	24	2.241	0.983	3.942	1.293	4.005	1.183	3.405	1.290	5.104	1.544	2.405	0.987	2.022	0.925	2.280	1.061	2.036	0.969
	36	2.811	1.060	4.982	1.497	3.456	1.300	3.522	1.291	5.158	1.571	2.328	0.984	1.974	0.932	2.235	1.059	1.928	0.940
	48	3.024	1.084	4.164	1.331	3.059	1.124	3.478	1.294	5.101	1.565	2.330	0.990	1.979	0.955	2.298	1.079	1.880	0.931
	60	3.758	1.229	5.725	1.651	2.771	1.163	2.880	1.154	5.319	1.596	2.413	1.015	1.954	0.949	2.573	1.157	2.016	0.976
Weather	96	0.172	0.210	0.251	0.235	1.048	0.300	0.269	0.339	0.335	0.406	0.176	0.216	0.181	0.232	0.174	0.233	0.175	0.225
	192	0.213	0.249	0.330	0.294	2.246	0.372	0.338	0.395	0.693	0.599	0.225	0.257	0.225	0.268	0.218	0.278	0.217	0.259
	336	0.257	0.287	0.420	0.357	3.636	0.470	0.339	0.381	0.564	0.527	0.281	0.299	0.271	0.301	0.263	0.314	0.265	0.294
	720	0.343	0.353	0.538	0.423	0.795	0.541	0.429	0.433	1.105	0.771	0.358	0.350	0.339	0.349	0.332	0.374	0.329	0.339
Exchange	96	0.096	0.229	0.902	0.647	0.202	0.334	0.143	0.274	0.943	0.772	0.086	0.206	0.089	0.208	0.085	0.209	0.089	0.209
	192	0.196	0.334	1.084	0.744	0.371	0.466	0.266	0.377	1.244	0.882	0.181	0.304	0.181	0.300	0.162	0.296	0.191	0.309
	336	0.886	0.733	0.775	0.678	1.122	0.852	0.465	0.509	1.790	1.070	0.338	0.422	0.330	0.415	0.333	0.441	0.363	0.434
	720	2.479	1.179	1.306	0.879	1.206	0.792	1.088	0.812	2.905	1.406	0.853	0.696	0.925	0.722	0.898	0.725	0.963	0.731
$1^{\text{st}}$ Count		10	11	0	0	0	0	0	0	0	0	1	2	4	6	3	1	12	9