Definitions: ''' Redundant feature map: A feature map is redundant if it doesnt provide discriminatory information across various classes. Important: the feature map which is not redundant. '''
This code focuses on identifying feature maps which are redundant/irrelavant as defined above in a well-trained Convolution neural network. Three statistical methods [1] and one geometrical method [2] are used to identify redundancy.
Flow chart of the codes is as follows:
(a) Extract intermediate representations form SoundNet using file .."Soundnet_layerwise_featuremap_extraction.py"
(b) Put files class-wise into a directory using code....."filecopy.py"
(c) Read class-wise training and testing dataset using "soundnet_layerwise_feature_read.py"
(d) Identification important feature maps using code.."ICASSP_PRL_filter_pruning.py"
References:
-
Singh, Arshdeep, Padmanabhan Rajan, and Arnav Bhavsar. "Deep hidden analysis: A statistical framework to prune feature maps." ICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2019.
-
Singh, Arshdeep, Padmanabhan Rajan, and Arnav Bhavsar. "SVD-based redundancy removal in 1-D CNNs for acoustic scene classification." Pattern Recognition Letters 131 (2020): 383-389.
-
Aytar, Yusuf, Carl Vondrick, and Antonio Torralba. "Soundnet: Learning sound representations from unlabeled video." Advances in neural information processing systems. 2016.