Artificial Intelligence #6 : LSTM Neural Networks with Keras

Explore recurrent neural networks and long short-term memory units, including gated memory, forget/input/output gates, and backpropagation through time, to model sequences and time series such as speech and handwriting.

Develop an lstm-based forecast of Google stock price using Keras, loading a Google stock dataset in Python, and preparing the data with Pandas and Excel inputs.

Select high, low, and close columns from Google stock data and use high and low as inputs to forecast the price with an LSTM in Keras, supported by visual plots.

Combine high and low stock data into input X, define y as the close price, apply transposes and reshapes, and scale data for a 3D LSTM model in Keras.

Build an lstm neural network in keras to forecast google stock prices, using 100 units in first layer with tanh and sigmoid activations, trained on mean squared error and mae.

Forecast Google stock price with LSTM neural networks in Keras, train the model, and visualize predicted versus real data to demonstrate time series forecasting power.

Forecast the Nasdaq index close price as a time series using an lstm neural network with keras, including data loading, preprocessing, and plotting the results.

Define three lagged inputs of the Nasdaq index and combine them into X to forecast the next value with LSTM using Keras; standardize data and prepare Y via transposition.

Forecast the NASDAQ index using LSTMs with Keras, preparing data with scalar normalization, reshaping to 3D inputs, and splitting into train and test sets before building a sequential model.

Forecast the Nasdaq index with a Keras LSTM model using three input delays and a 10-unit tanh layer, with a single output and mean squared error training.

Forecast the Nasdaq index using lstm neural networks with keras, compare test and predict results, and improve accuracy with adjusted units.

Import libraries and load the New York City temperature dataset from 2017, then forecast the minimum, maximum, and average temperatures using LSTM neural networks with Keras.

Explore predicting New York annual temperature with LSTM using Keras, by defining min, max, and average temperatures, building 3D visualizations, and creating input–output datasets for the recurrent model.

This lecture demonstrates using keras lstm to predict New York temperature by preparing one-dimensional data: plotting average temperature, creating x and y datasets, normalizing, transposing, and reshaping for 3d inputs.

Use a Keras sequential LSTM model to predict New York annual temperature with a 1x2 input, trained via a train-test split and mean squared error loss.

Forecast New York City's annual temperature with a Keras LSTM, evaluate on test data using mean absolute error, and compare predictions to real values.

Forecast New York wind speed with LSTM neural networks using Keras. Part 1 covers importing libraries, scaling data with MinMaxScaler, and loading the course dataset for preprocessing.

Forecast New York wind speed with LSTM using Keras, preparing data with three time delays, plotting last year's wind data, and shaping input for neural networks.

Transpose y, apply delays, and scale X and y with a min-max approach; reshape into a three-dimensional LSTM input for Keras, then set up train-test data with a defined size.

Forecast New York wind speed with LSTM networks in Keras by building a model, using a five-unit input layer with tanh activation, and training on three-day data.

Train an LSTM model with Keras to forecast New York wind speed, evaluate absolute error and accuracy against real data, and discuss tuning units and time steps for noisy data.