The Age of Correlated Features in Supervised Learning based Forecasting

  title={The Age of Correlated Features in Supervised Learning based Forecasting},
  author={Md. Kamran Chowdhury Shisher and Heyang Qin and Lei Yang and Feng Yan and Yin-Bo Sun},
  journal={IEEE INFOCOM 2021 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS)},
In this paper, we analyze the impact of information freshness on supervised learning based forecasting. In these applications, a neural network is trained to predict a time-varying target (e.g., solar power), based on multiple correlated features (e.g., temperature, humidity, and cloud coverage). The features are collected from different data sources and are subject to heterogeneous and time-varying ages. By using an information-theoretic approach, we prove that the minimum training loss is a… 

Figures from this paper

Age of Information: An Introduction and Survey
The current state of the art in the design and optimization of low-latency cyberphysical systems and applications in which sources send time-stamped status updates to interested recipients is described and AoI timeliness metrics are described.
Age of Sensed Information in a Cognitive Radio Network
  • C. Kam, S. Kompella, A. Ephremides
  • Computer Science
    2021 19th International Symposium on Modeling and Optimization in Mobile, Ad hoc, and Wireless Networks (WiOpt)
  • 2021
This work studies a two-user, single-channel cognitive radio network, where the primary user’s transmit/idle dynamics are modeled as a binary Markov chain, and the secondary user decides to either sense or transmit.
Closed-form Characterization of the MGF of AoI in Energy Harvesting Status Update Systems
The generality of this analysis is demonstrated by recovering several existing results for the corresponding system with no energy constraints as special cases of the new results, and the necessity of incorporating the higher moments of AoI in the implementation/optimization of real-time status update systems rather than just relying on its average value is demonstrated.


Machine Learning Advances for Time Series Forecasting
The most recent advances in supervised machine learning and highdimensional models for time series forecasting are surveyed and ensemble and hybrid models by combining ingredients from different alternatives are considered.
Short-Term Solar Forecasting Performance of Popular Machine Learning Algorithms: Preprint
A framework for assessing the performance of short-term solar forecasting is presented in conjunction with a range of numerical results using global horizontal irradiation (GHI) from the open-source
An object-oriented neural network approach to short-term traffic forecasting
  • H. Dia
  • Computer Science
    Eur. J. Oper. Res.
  • 2001
The results obtained indicate that the TLRN is capable of predicting speed up to 5 minutes into the future with a high degree of accuracy, which represents substantial improvements on conventional model performance and clearly demonstrate the feasibility of using the object-oriented approach for short-term traffic prediction.
Applying long short term momory neural networks for predicting stock closing price
  • Tingwei Gao, Y. Chai, Y. Liu
  • Computer Science
    2017 8th IEEE International Conference on Software Engineering and Service Science (ICSESS)
  • 2017
The main goal of this paper is to assess the hypothesis that combining RNNs with informative input variables can provide a more effective method for predicting the next-day stock movement. Moreover,
An LSTM Based System for Prediction of Human Activities with Durations
Two distinct Long-Short Term Memory (LSTM) networks are developed that cater to different assumptions about the data and achieve different modeling complexities and prediction accuracies.
Spatial-Temporal Graph Attention Networks: A Deep Learning Approach for Traffic Forecasting
A novel deep learning framework, Spatial-Temporal Graph Attention Networks (ST-GAT), a graph attention mechanism is adopted to extract the spatial dependencies among road segments and a LSTM network is introduced to extract temporal domain features.
A novel hybrid (Mycielski-Markov) model for hourly solar radiation forecasting
In this study, short term predictions of solar radiation are reviewed and an alternative approach and model is proposed, a novel Mycielski based model that assumes that solar radiation data repeats itself in the history.
Probabilistic forecasting of high-resolution clear-sky index time-series using a Markov-chain mixture distribution model
The Markov-chain mixture distribution model is suggested as a candidate benchmark model in probabilistic forecasting, in particular for solar irradiance forecasting, and concluded to be a computationally inexpensive, accurate and parameter insensitive probabilism model.
A Minimax Approach to Supervised Learning
The maximum entropy machine minimizes the worst-case 0-1 loss over the structured set of distribution, and by the numerical experiments can outperform other well-known linear classifiers such as SVM.
On Universal Features for High-Dimensional Learning and Inference
This framework facilitates understanding and optimizing aspects of learning systems, including multinomial logistic (softmax) regression and the associated neural network architecture, matrix factorization methods for collaborative filtering and other applications, rank-constrained multivariate linear regression, and forms of semi-supervised learning.