you could also just use a Task-agnostic CNN as an encoder to get extract features like in (1) and then use the output of the last global pooling layer and then feed that as an input to the LSTM layer or any other downstream task. Add another small Neural Network (projection head) after the CNN. And then use contrastive loss on output of this projection head ...
I think what you need to use is 3D convolution operation. Your data is 3D, width, height, and num_channels. Your data is similar to color images with RGB channels. However, since you are trying to consider the correlation amongst channels 2D convolution will not work for you. You can use 3D convolution which is available to use with deep learning tools such ...
The approach that you don't train the whole net, but just the latter part of it (all starting with lstm in our case), can actually work. The idea is that the inception was already pretrained a very large dataset (imagenet for instance). And it's capable of extracting some useful information from it. Actually there are different domains of images in the ...
If you have a lexicon for each language, you could just take every word of the input, and see what lexicons it is in. If one language has (almost) all the words of the input, and others score a lot worse, that would be your best guess.
At that point, you can try parsing the input using a parser for the best few candidates, just to double-check.
As it is told in PIL documentation
It uses some filters to resize images.And those filters are explained here
uses mostly numerical methods as I see. So it is approximating the image data input. Which means you are right about data loss. But here might be the question would it change the data so much If it is done after augmentation or before?
Since in ...
I suggest you to have a look at this repo. It contains state-of-the art algorithms, papers, frameworks, courses and some implementations. You can also check "Deep Reinforcement Learning Hands On" book examples written by Max Lapan here. This repo contains many programming and reinforcement learning examples with PyTorch framework.
Arthur Juliani has some interesting Medium articles on reinforcement learning with TensorFlow backed up with code on GitHub.
Part 0 — Q-Learning Agents
Part 1 — Two-Armed Bandit
Part 1.5 — Contextual Bandits
Part 2 — Policy-Based Agents
Part 3 — Model-Based RL
Part 4 — Deep Q-Networks and Beyond
Part 5 — Visualizing an Agent’s Thoughts and Actions
Part 6 — ...
LSTM can be tricky, I'll give my $0.02.
LSTM input layer defines the shape so it would be something like this.
If I am understanding your question correctly, your data can be framed as 184 samples with 2 time steps and 70 features?
So the start of the code might look like this.
model = Sequential()
model.add(LSTM(184, input_shape=(50, 2)))