Find file History
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
data
model_params
prediction
result
results
swarm_description
.gitignore
README.md
__init__.py
errorMetrics.py
exampleTaxiPrediction_CLAClassifier.pdf
exampleTaxiPrediction_SDRClassifier.pdf
experiments.cfg
nupic_output.py
plot.py
plotPerformance.py
plotPerturbResult.py
plot_noisy_taxi_experiment.py
run_adaptive_filter.py
run_arima.R
run_elm.py
run_esn_model.m
run_esn_model_online.m
run_knn.py
run_lstm.py
run_lstm_scalarEncoder.py
run_lstm_suite.py
run_noisy_taxi_experiment.py
run_tdnn.py
run_tm_model.py
run_tm_model_compare_sdr_cla_classifier.py
swarm_runner.py

README.md

Description

Run HTM and other competing algorithms on continuous time series prediction problem

This experiment is used in

Cui Y, Ahmad S, Hawkins J. Continuous online sequence learning with an unsupervised neural network model. (2016) Neural Computation, 28(11) 2474-2504. DOI: 10.1162/NECO_a_00893

Example Usage

run HTM sequence memory on continuous time series prediction problems

python run_tm_model.py -d DATASET

Existing dataset includes nyc_taxi, nyc_taxi_perturb, sine, rec-center-hourly,

run LSTM on dataset (requires pyBrain)

python run_lstm_suite.py -e EXPERIMENT_NAME -d

EXPREIMENT_NAME is defined in experiments.cfg (e.g., nyc_taxi_experiment_continuous) The -d flag removes existing experiment logs and starts a new experiments

run ARIMA on dataset (requires R)

Rscript run_arima.R

run Extreme Learning Machine model (requires hpelm)

python run_elm.py -d DATASET

run time-delayed neural network (TDNN) (requires pyBrain)

python run_tdnn.py -d DATASET

run adaptive filter (requires adaptfilt

python run_adaptive_filter.py -d DATASET

run Echo State Network model (requires Matlab and ESNToolbox)

run run_esn_model.m in Matlab

compare HTM performance with other algorithms

python plotPerformance.py.py