Shape and Time Distortion Loss for Training Deep Time Series Forecasting Models (2019-09-19T00:00:00.000000Z)

TL;DR

DILATE (DIstortion Loss including shApe and TimE), a new objective function for training deep neural networks that aims at accurately predicting sudden changes, is introduced, and explicitly incorporates two terms supporting precise shape and temporal change detection.

Abstract

This paper addresses the problem of time series forecasting for non-stationary signals and multiple future steps prediction. To handle this challenging task, we introduce DILATE (DIstortion Loss including shApe and TimE), a new objective function for training deep neural networks. DILATE aims at accurately predicting sudden changes, and explicitly incorporates two terms supporting precise shape and temporal change detection. We introduce a differentiable loss function suitable for training deep neural nets, and provide a custom back-prop implementation for speeding up optimization. We also introduce a variant of DILATE, which provides a smooth generalization of temporally-constrained Dynamic TimeWarping (DTW). Experiments carried out on various non-stationary datasets reveal the very good behaviour of DILATE compared to models trained with the standard Mean Squared Error (MSE) loss function, and also to DTW and variants. DILATE is also agnostic to the choice of the model, and we highlight its benefit for training fully connected networks as well as specialized recurrent architectures, showing its capacity to improve over state-of-the-art trajectory forecasting approaches.

Authors

V. Guen

2 papers

Nicolas Thomé

3 papers

TL;DR

Abstract

Authors

References64 items

A New Timing Error Cost Function for Binary Time Series Prediction

Enhancing the Locality and Breaking the Memory Bottleneck of Transformer on Time Series Forecasting

Deep Factors for Forecasting

Think Globally, Act Locally: A Deep Neural Network Approach to High-Dimensional Time Series Forecasting

Supervised Kernel Change Point Detection with Partial Annotations

Exploiting Negative Evidence for Deep Latent Structured Models

Autowarp: Learning a Warping Distance from Unlabeled Time Series Using Sequence Autoencoders

Kernel Change-point Detection with Auxiliary Deep Generative Models

HybridNet: Classification and Reconstruction Cooperation for Semi-Supervised Learning

Deep Multi-Output Forecasting: Learning to Accurately Predict Blood Glucose Trajectories

Multi-step Time Series Forecasting of Electric Load Using Machine Learning Models

Hierarchical Attention-Based Recurrent Highway Networks for Time Series Prediction

Foundations of Sequence-to-Sequence Modeling for Time Series

Differentiable Dynamic Programming for Structured Prediction and Attention

A Multi-Horizon Quantile Recurrent Forecaster

Long-term Forecasting using Tensor-Train RNNs

Long-term Forecasting using Higher Order Tensor RNNs

Deep Forecast: Deep Learning-based Spatio-Temporal Forecasting

Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting

Towards a standardized procedure to assess solar forecast accuracy: A new ramp and time alignment metric

Co-evolutionary multi-task learning with predictive recurrence for multi-step chaotic time series prediction

Assessing energy forecasting inaccuracy by simultaneously considering temporal and absolute errors

Attention is All you Need

DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks

A Dual-Stage Attention-Based Recurrent Neural Network for Time Series Prediction

Modeling Long- and Short-Term Temporal Patterns with Deep Neural Networks

Conditional Time Series Forecasting with Convolutional Neural Networks

Soft-DTW: a Differentiable Loss Function for Time-Series

Consistent change-point detection with kernels

Bayesian Intermittent Demand Forecasting for Large Inventories

Temporal Regularized Matrix Factorization for High-dimensional Time Series Prediction

WaveNet: A Generative Model for Raw Audio

RETAIN: An Interpretable Predictive Model for Healthcare using Reverse Time Attention Mechanism

Multi-step-ahead chaotic time series prediction using coevolutionary recurrent neural networks

The Lovász Hinge: A Novel Convex Surrogate for Submodular Losses

Anomaly detection in ECG time signals via deep long short-term memory networks

MANTRA: Minimum Maximum Latent Structural SVM for Image Classification and Ranking

Comparison of Strategies for Multi-step-Ahead Prediction of Time Series Using Neural Network

Deep Learning for Event-Driven Stock Prediction

M-Statistic for Kernel Change-Point Detection

Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning

Traffic Flow Prediction With Big Data: A Deep Learning Approach

Improving Multi-Step Prediction of Learned Time Series Models

Sequence to Sequence Learning with Neural Networks

Optimal Decisions from Probabilistic Models: The Intersection-over-Union Case

Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

Multi-step-ahead time series prediction using multiple-output support vector regression

Identifying Wind and Solar Ramping Events

Weighted dynamic time warping for time series classification

A review and comparison of strategies for multi-step ahead time series forecasting based on the NN5 forecasting competition

Forecasting with Exponential Smoothing: The State Space Approach

A support vector method for optimizing average precision

Multiple-step ahead prediction for non linear dynamic systems: A Gaussian Process treatment with propagation of the uncertainty

Time Series Analysis by State Space Methods

Long Short-Term Memory

Dynamic programming algorithm optimization for spoken word recognition

Time series analysis, forecasting and control

On the Theory of Dynamic Programming.

Deep State Space Models for Time Series Forecasting

Learning Chaotic Dynamics using Tensor Recurrent Neural Networks

Time-series Extreme Event Forecasting with Neural Networks at Uber

Electric Load Forecasting in Smart Grid Using Long-Short-Term-Memory based Recurrent Neural Network

A Bias and Variance Analysis for Multistep-Ahead Time Series Forecasting

The UCR time series classification archive

Field of Study

Journal Information

Name

Volume

Venue Information

Name

Type

URL

Alternate Names