Purpose Of Experiment - To understand how RNN Architecture helps in classification task of temporal data like Text.
Experiment done in Month - August
Models-
- Model0 - Basic DAN kind of structure
- Model1 - Modified Model0 architecture to use RNN
- Model2 - Use packed padded sequence to avoid giving padded word vectors as output.
Dataset-
- Data0 -
i. Train.neg - Contain Negative sentiments
ii. Train.pos - Contain Positive Sentiments
iii. TestData - Contain Test File having first half positive and second half negative sentiments - Data1 -
Imdb Dataset divided into training and testing dataset.
Commands-
usage:
python main.py
[--batchsize BATCHSIZE]
[--seqlen SEQLEN]
[--glovepath GLOVEPATH]
[--dnum DNUM]
[--cuda_num CUDA_NUM]
[--epochs EPOCHS]
[--model MODEL]
[--patience PATIENCE]
arguments: -h, --help show this help message and exit --batchsize Batch Size --seqlen Seq len of each word vector after padding and truncating --glovepath Path to Glove Embedding file --dnum Dataset are numbered. This take which dataset I want to use --cuda_num device number --epochs Number of Epochs --model Models are numbered. Select the appropriate model number. --patience How much you want to wait before early stopping