Skip to main content
Fig. 5 | Visual Computing for Industry, Biomedicine, and Art

Fig. 5

From: Visual analytics tool for the interpretation of hidden states in recurrent neural networks

Fig. 5

Example of a wrongly classified sequence from the IMDB dataset. For the first part of the sequence, the classification is rather negative (pink). Only in the second half of the sequence does the classification become positive (blue). The last words in particular contribute to the incorrect classification, where the name of an actor appears, and afterward, the words refer to another film that was rated better compared to the current one. Underlying data source: IMDB as available in Keras [13]

Back to article page