Skip to content

Learning to Rank with Alpha-divergence and Entropy Regularization

Notifications You must be signed in to change notification settings

xiaohai2016/StatRank

Repository files navigation

Learning to Rank with Alpha-divergence and Entropy Regularization

Steps to reproduce results in the paper

  1. Create a folder named resources under the root of the cloned code tree;
  2. Download MQ2007 and MQ2008 datasets and decompress them under the resources folder.
  3. Prepare your execution environment with PyTorch 1.1.0 and Scikit.
  4. Run the following commands to generate the test results:
# dataset - MQ2007, Ranking function - Single layer, Divergence - listed below
nohup python -u main.py --listnet --listmle --alpha-divergence --weighted-kl-divergence > ../mq2007-onelayer-multi.txt &

# dataset - MQ2008, Ranking function - Single layer, Divergence - listed below
nohup python -u main.py --listnet --listmle --alpha-divergence --weighted-kl-divergence --data-set mq2008 > ../mq2008-onelayer-multi.txt &

# dataset - MQ2007, Ranking function - Single layer, Divergence - alpha divergence with entropy regularization
nohup python -u main.py --alpha-and-entropy > ../mq2007-onelayer-alpha-entropy.txt &

# dataset - MQ2008, Ranking function - Single layer, Divergence - alpha divergence with entropy regularization
nohup python -u main.py --alpha-and-entropy --data-set mq2008 > ../mq2008-onelayer-alpha-entropy.txt &

# dataset - MQ2007, Ranking function - Three layer, Divergence - listed below
nohup python -u main.py --listnet --listmle > ../mq2007-threelayer-multi.txt &

# dataset - MQ2008, Ranking function - Three layer, Divergence - listed below
nohup python -u main.py --listnet --listmle --alpha-divergence --weighted-kl-divergence --data-set mq2008 --model three_layer > ../mq2008-threelayer-multi.txt &

# dataset - MQ2007, Ranking function - Three layer, Divergence - alpha divergence with entropy regularization
#nohup python -u main.py --alpha-and-entropy --model three_layer > ../mq2007-threelayer-alpha-entropy.txt &
# I once did run the above command and later abandoned it. It takes many days to complete and the hyper parameters are not good.

# dataset - MQ2008, Ranking function - Three layer, Divergence - alpha divergence with entropy regularization
nohup python -u main.py --alpha-and-entropy --data-set mq2008 --model three_layer > ../mq2008-threelayer-alpha-entropy.txt &

# rerun for MQ2007 three_layer
nohup python -u main.py --alphas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-divergence --model three_layer > ../mq2007-threelayer-alpha.txt &
nohup python -u main.py --alphas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --weighted-kl-divergence --model three_layer > ../mq2007-threelayer-weighted-kl.txt &

nohup python -u main.py --alphas -0.9 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-neg0.9.txt &
nohup python -u main.py --alphas -0.6 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-neg0.6.txt &
nohup python -u main.py --alphas -0.4 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-neg0.4.txt &
nohup python -u main.py --alphas -0.2 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-neg0.2.txt &
nohup python -u main.py --alphas -0.1 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-neg0.1.txt &
nohup python -u main.py --alphas 0.1 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-0.1.txt &
nohup python -u main.py --alphas 0.2 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-0.2.txt &
nohup python -u main.py --alphas 0.3 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-0.3.txt &
nohup python -u main.py --alphas 0.4 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-0.4.txt &
nohup python -u main.py --alphas 0.5 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-0.5.txt &
nohup python -u main.py --alphas 0.6 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-0.6.txt &
nohup python -u main.py --alphas 0.7 --lambdas -0.9 -0.6 -0.4 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 --alpha-and-entropy --model three_layer > ../mq2007-alpha-and-entropy-0.7.txt &

About

Learning to Rank with Alpha-divergence and Entropy Regularization

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published