lstm

问题 I am learning tensorflow2.0 and follow the tutorial. In the rnn example, I found the code: def build_model(vocab_size, embedding_dim, rnn_units, batch_size): model = tf.keras.Sequential([ tf.keras.layers.Embedding(vocab_size, embedding_dim, batch_input_shape=[batch_size, None]), tf.keras.layers.LSTM(rnn_units, return_sequences=True, stateful=True, recurrent_initializer='glorot_uniform'), tf.keras.layers.Dense(vocab_size) ]) return model My question is: why the code set the argument return

问题 I am learning tensorflow2.0 and follow the tutorial. In the rnn example, I found the code: def build_model(vocab_size, embedding_dim, rnn_units, batch_size): model = tf.keras.Sequential([ tf.keras.layers.Embedding(vocab_size, embedding_dim, batch_input_shape=[batch_size, None]), tf.keras.layers.LSTM(rnn_units, return_sequences=True, stateful=True, recurrent_initializer='glorot_uniform'), tf.keras.layers.Dense(vocab_size) ]) return model My question is: why the code set the argument return

问题 I've been working on reproducing a CNN-LSTM model for PV power forecasting from literature for the past four weeks for my Master Thesis in Energy Science (http://www.mdpi.com/2076-3417/8/8/1286). However I've been stuck on a seemingly simple issue: Any configuration of LSTM model that I've tried yields one of two things: Rediculous output, makes no sense whatsoever (flat line, complete stochasticity, negative values, you name it) Exactly the same (very believable) PV power forecast. I've done

问题 I have a similar DataFrame : df = pd.DataFrame([ {'date':'2021-01-15', 'value':145, 'label':'negative'}, {'date':'2021-01-16', 'value':144, 'label':'positive'}, {'date':'2021-01-17', 'value':147, 'label':'positive'}, {'date':'2021-01-18', 'value':146, 'label':'negative'}, {'date':'2021-01-19', 'value':155, 'label':'negative'}, {'date':'2021-01-20', 'value':157, 'label':'positive'}, {'date':'2021-01-21', 'value':158, 'label':'positive'}, {'date':'2021-01-22', 'value':157, 'label':'negative'},

问题 TLDR : How to implement this model using tf.lite.experimental.nn.TFLiteLSTMCell, tf.lite.experimental.nn.dynamic_rnn instead keras.layers.LSTM ? I have this network in keras: inputs = keras.Input(shape=(1, 52)) state_1_h = keras.Input(shape=(200,)) state_1_c = keras.Input(shape=(200,)) x1, state_1_h_out, state_1_c_out = layers.LSTM(200, return_sequences=True, input_shape=(sequence_length, 52), return_state=True)(inputs, initial_state=[state_1_h, state_1_c]) output = layers.Dense(13)(x1) model

问题 Currently I can train a LSTM network using one csv file based on this tutorial: https://machinelearningmastery.com/how-to-develop-lstm-models-for-time-series-forecasting/ This code generate sliding windows where the last n_steps of the features are saved to predict the actual target (similar to this: Keras LSTM - feed sequence data with Tensorflow dataset API from the generator): #%% Import import pandas as pd import tensorflow as tf from tensorflow.python.keras.models import Sequential,

问题 I am now trying to prepare the input data for LSTM-based NN. I have some big number of text documents and what i want is to make sequence vectors for each document so i am able to feed them as train data to LSTM RNN. My poor approach: import re import numpy as np #raw data train_docs = ['this is text number one', 'another text that i have'] #put all docs together train_data = '' for val in train_docs: train_data += ' ' + val tokens = np.unique(re.findall('[a-zа-я0-9]+', train_data.lower()))

问题 I want to build a model, that predicts next character based on the previous characters. I have spliced text into sequences of integers with length = 100(using dataset and dataloader). Dimensions of my input and target variables are: inputs dimension: (batch_size,sequence length). In my case (128,100) targets dimension: (batch_size,sequence length). In my case (128,100) After forward pass I get dimension of my predictions: (batch_size, sequence_length, vocabulary_size) which is in my case (128

问题 I want to build a model, that predicts next character based on the previous characters. I have spliced text into sequences of integers with length = 100(using dataset and dataloader). Dimensions of my input and target variables are: inputs dimension: (batch_size,sequence length). In my case (128,100) targets dimension: (batch_size,sequence length). In my case (128,100) After forward pass I get dimension of my predictions: (batch_size, sequence_length, vocabulary_size) which is in my case (128

问题 I am trying to augment a network. For the convolution part, I am using convlstm2d from keras. Is there a process to perform deconvolution ( i.e. lstmdeconv2d ? ) 回答1: There is Conv3D for that, checkout this example used to predict the next frame 回答2: It should be possible to combine any model with the TimeDistributed wrapper. So you can create a deconv-model, and apply it on the output (which is a sequence of vectors) of the LSTM using the TimeDistributed wrapper. An example. First create a