Effect of parameter variations on accuracy of Convolutional Neural Network
Kaushik Gobindram Pasi, Sowmiyaraksha R. Naik · 2016
In this paper, we implement a Convolutional Neural Network especially designed for Natural Language processing. With the help of this CNN, we try to classify sentences for sentiment analysis for which the embeddings used were learned from scratch rather than using pre-trained word2vec vectors. Here we try to vary the different parameters and learn how they effect on the performance of the CNN. From the observations we try to demonstrate that a fairly less-complex CNN that has a small amount of parameter adjustments and fine-tuning can achieve a significant growth in performance.