Implementation of deep learning models for time series in PyTorch.

Overview

List of Implementations:

Currently, the reimplementation of the DeepAR paper(DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks https://arxiv.org/abs/1704.04110) is available in PyTorch. More papers will be coming soon.

Authors:

  • Yunkai Zhang([email protected]) - University of California, Santa Barbara

  • Qiao Jiang - Brown University

  • Xueying Ma - Columbia University

  • Acknowledgement: Professor Xifeng Yan's group at UC Santa Barbara. Part of the work was done at WeWork.

To run:

  1. Install all dependencies listed in requirements.txt. Note that the model has only been tested in the versions shown in the text file.

  2. Download the dataset and preprocess the data:

    python preprocess_elect.py
  3. Start training:

    python train.py
    • If you want to perform ancestral sampling,

      python train.py --sampling
    • If you do not want to do normalization during evaluation,

      python train.py --relative-metrics
  4. Evaluate a set of saved model weights:

    python evaluate.py
  5. Perform hyperparameter search:

     python search_params.py

Results

​ The model is evaluated on the electricity dataset, which contains the electricity consumption of 370 households from 2011 to 2014. Under hourly frequency, we use the first week of September, 2014 as the test set and all time steps prior to that as the train set. Following the experiment design in DeepAR, the window size is chosen to be 192, where the last 24 is the forecasting horizon. History (number of time steps since the beginning of each household), month of the year, day of the week, and hour of the day are used as time covariates. Notice that some households started at different times, so we only use windows that contain non-missing values.

​ Under Gaussian likelihood, we use the Adam optimizer with early stopping to train the model for 20 epoches. The same set of hyperparameters is used as outlined in the paper. Weights with the best ND value is selected, where ND = 0.06349, RMSE = 0.452, rou90 = 0.034 and rou50 = 0.063.

​ Sample results on electricity. The top 10 plots are sampled from the test set with the highest 10% ND values, whereas the bottom 10 plots are sampled from the rest of the test set.

Sample results on electricity. The top 10 plots are sampled from the test set with the highest 10% ND values, whereas the bottom 10 plots are sampled from the rest of the test set.

Owner
Yunkai Zhang
IEOR PhD @ UC Berkeley, math/computing @ UCSB CCS
Yunkai Zhang
Machine Learning University: Accelerated Natural Language Processing Class

Machine Learning University: Accelerated Natural Language Processing Class This repository contains slides, notebooks and datasets for the Machine Lea

AWS Samples 2k Jan 01, 2023
Warren - Stock Price Predictor

Web app to predict closing stock prices in real time using Facebook's Prophet time series algorithm with a multi-variate, single-step time series forecasting strategy.

Kumar Nityan Suman 153 Jan 03, 2023
dirty_cat is a Python module for machine-learning on dirty categorical variables.

dirty_cat dirty_cat is a Python module for machine-learning on dirty categorical variables.

637 Dec 29, 2022
A single Python file with some tools for visualizing machine learning in the terminal.

Machine Learning Visualization Tools A single Python file with some tools for visualizing machine learning in the terminal. This demo is composed of t

Bram Wasti 35 Dec 29, 2022
Official code for HH-VAEM

HH-VAEM This repository contains the official Pytorch implementation of the Hierarchical Hamiltonian VAE for Mixed-type Data (HH-VAEM) model and the s

Ignacio Peis 8 Nov 30, 2022
50% faster, 50% less RAM Machine Learning. Numba rewritten Sklearn. SVD, NNMF, PCA, LinearReg, RidgeReg, Randomized, Truncated SVD/PCA, CSR Matrices all 50+% faster

[Due to the time taken @ uni, work + hell breaking loose in my life, since things have calmed down a bit, will continue commiting!!!] [By the way, I'm

Daniel Han-Chen 1.4k Jan 01, 2023
A simple python program which predicts the success of a movie based on it's type, actor, actress and director

Movie-Success-Prediction A simple python program which predicts the success of a movie based on it's type, actor, actress and director. The program us

Mahalinga Prasad R N 1 Dec 17, 2021
πŸŽ› Distributed machine learning made simple.

πŸŽ› lazycluster Distributed machine learning made simple. Use your preferred distributed ML framework like a lazy engineer. Getting Started β€’ Highlight

Machine Learning Tooling 44 Nov 27, 2022
Pytools is an open source library containing general machine learning and visualisation utilities for reuse

pytools is an open source library containing general machine learning and visualisation utilities for reuse, including: Basic tools for API developmen

BCG Gamma 26 Nov 06, 2022
Price Prediction model is used to develop an LSTM model to predict the future market price of Bitcoin and Ethereum.

Price Prediction model is used to develop an LSTM model to predict the future market price of Bitcoin and Ethereum.

2 Jun 14, 2022
BudouX is the successor to Budou, the machine learning powered line break organizer tool.

BudouX Standalone. Small. Language-neutral. BudouX is the successor to Budou, the machine learning powered line break organizer tool. It is standalone

Google 868 Jan 05, 2023
monolish: MONOlithic Liner equation Solvers for Highly-parallel architecture

monolish is a linear equation solver library that monolithically fuses variable data type, matrix structures, matrix data format, vendor specific data transfer APIs, and vendor specific numerical alg

RICOS Co. Ltd. 179 Dec 21, 2022
A concept I came up which ditches the idea of "layers" in a neural network.

Dynet A concept I came up which ditches the idea of "layers" in a neural network. Install Copy Dynet.py to your project. Run the example Install matpl

Anik Patel 4 Dec 05, 2021
To-Be is a machine learning challenge on CodaLab Platform about Mortality Prediction

To-Be is a machine learning challenge on CodaLab Platform about Mortality Prediction. The challenge aims to adress the problems of medical imbalanced data classification.

Marwan Mashra 1 Jan 31, 2022
Book Item Based Collaborative Filtering

Book-Item-Based-Collaborative-Filtering Collaborative filtering methods are used

Şebnem 3 Jan 06, 2022
Winning solution for the Galaxy Challenge on Kaggle

Winning solution for the Galaxy Challenge on Kaggle

Sander Dieleman 483 Jan 02, 2023
Data science, Data manipulation and Machine learning package.

duality Data science, Data manipulation and Machine learning package. Use permitted according to the terms of use and conditions set by the attached l

David Kundih 3 Oct 19, 2022
BigDL: Distributed Deep Learning Framework for Apache Spark

BigDL: Distributed Deep Learning on Apache Spark What is BigDL? BigDL is a distributed deep learning library for Apache Spark; with BigDL, users can w

4.1k Jan 09, 2023
A classification model capable of accurately predicting the price of secondhand cars

The purpose of this project is create a classification model capable of accurately predicting the price of secondhand cars. The data used for model building is open source and has been added to this

Akarsh Singh 2 Sep 13, 2022
Confidence intervals for scikit-learn forest algorithms

forest-confidence-interval: Confidence intervals for Forest algorithms Forest algorithms are powerful ensemble methods for classification and regressi

272 Dec 01, 2022