Download as docx, pdf, or txt
Download as docx, pdf, or txt
You are on page 1of 2

LSTM

Classification processes require relevant input instances and outputs, and methods such as
logistic regression, nearest neighbor, naive Bayes, support vector machines, decision trees, and
artificial neural networks are most used. These methods require each instance to be a single shot.
This means that the internal state is not updated in a timely manner. Descriptive statistics are
used to represent sequence data, but they cannot represent sequence data such as Data
dependencies are fully represented. Recurrent neural network (RNN) structures allow training
without preprocessing and reduce the shape of sequence features to single-shot vectors. The
internal state of the RNN model is updated during training to identify relationships between each
sequence. RNNs are used in natural language processing and other sequence data applications to
achieve excellent results. However, the problem of vanishing or exploding gradients in RNNs
makes RNNs effective only for short-term dependencies. Long Short-Term Memory (LSTM)
was developed to solve this problem and remember longer dependencies. LSTM has three gates.
a is forget, input and output, which determines what information to keep in memory and what
information to discard. The sigmoid layer plays an important role in the decision-making
process.

Figure #: LSTM cell


ANN

An artificial neural network (ANN) is a computer program that simulates the human brain's
information processing. They learn through experience and consist of hundreds of artificial
neurons or processing elements (PEs) connected to weights. The power of neural computation
depends on the connectivity of neurons in a network, each with a weighted input, transfer
function, and output. The behavior of a neural network is determined by the transfer functions of
the neurons, the learning rules, and the architecture itself. The basic architecture consists of his
three types of neuron layers: input layers, hidden layers, and output layers. A feedforward
network has a signal flow from an input unit to an output unit, whereas a recurrent network
includes a feedback connection. Network dynamics are important and neural networks must be
configured to produce desired results. There are many ways to determine the strength of a
connection, including using a priori knowledge to explicitly set weights.

Figure #: artificial neural network

You might also like