Deepleanring using tensorflow

Image Processing and Classification using CNN & Transfer Leanring

• Dropout
• maxpooling
• Image Augmentation
• Convolutions

Times series Forecasting using

• Moving Average & Naive Appraoch
• Linear model using Neural nets, dense layers
• RNN stateless(Sequence-vector & sequence-sequence for imporved speed))
• Stateful RNN
• RNNs with LSTM cells
• CNNs (using conv1D) - will create a model using weave-net model

Some more details

Overview and purpose

Udacity - Intro to TensorFlow for Deep Learning : Lession 9 : NLP (Tweaking the model) Tweaking the model for detecting sentiments

• Vocab size (consider over size of corpus) (while tokenizing)
• Padding (before or after)
• More or less embedding
• Input length ( where to truncate)
• Number of embedding dimensions
• Flattend to GlobalAveragePooling1D (using the latter)

For data follow: [https://gist.github.com/Kumaava] For code: [https://colab.research.google.com/github/tensorflow/examples/blob/master/courses/udacity_intro_to_tensorflow_for_deep_learning/l09c05_nlp_tweaking_the_model.ipynb#scrollTo=SZzXE-pT8K57]

Use the link below to see how my network sees the sentiment related to each word from amazon and yelp reviews, visualizing embeddings: using embedding vector and meta data

http://projector.tensorflow.org/?config=https://gist.githubusercontent.com/Kumaava/8cd9aabaf3d56ac09fa6d4ac6e39d6f7/raw/999bcdf1dadb1bbda43f49a41133bcc596303dad/nlp:embedding_link