Neural network

As explained in Demand Forecasting Best Practices

A neural network is a computational model composed of interconnected parameterized units organized in layers that transform inputs into outputs.

A neural network is a layered computational model that transforms numerical inputs into numerical outputs. Its interconnected neurons assign weights to inputs, apply activation functions, and pass results forward toward a prediction. section 15.2.3

Each neuron acts like a small mathematical cell: it receives numerical information, gives different importance to each input, and produces an output through an activation function. The activation function shapes the neuron's behavior, so choosing it matters. section 15.2.3

A typical neural network organizes computation into three layer types:

• Input layer: passes input features to the first hidden layer. section 15.2.3
• Hidden layers: combine weights and activation functions; their number and size define a critical part of the network's design. section 15.2.3
• Output layer: produces the final predictions, with each neuron corresponding to an output. section 15.2.3

Data flows from one layer to the next in forward propagation. Training, also called fitting, uses a specific optimization algorithm to adjust the network so it can generate useful outputs from its inputs. section 15.2.3 section 15.2.3

Forecasting multiple periods

A network can receive four input features and produce forecasts for two periods: one output neuron represents Period 1 and another represents Period 2. The example makes the connection between network outputs and planning horizons explicit. section 15.2.3

Figure 15.7 Artificial neuron

Inspect the artificial-neuron figure to see how numerical inputs, weights, an activation function, and an output fit together. section 15.2.3

Figure 15.8 Artificial neural network

Inspect the network figure to follow the flow from input features through hidden layers to multiple predictions. section 15.2.3

How Demand Forecasting Best Practices covers it

How deep it goes

Explains neurons, activation functions, weights, layers, forward propagation, and training, but does not detail the optimization procedure.

Examples

Uses four input features, two forecast periods, artificial-neuron and network figures, and a timeline reaching back to 1957.

What it assumes

Assumes numerical inputs and teaches the neuron, activation function, layers, propagation, and training concepts from their basic definitions.

What you can do after

Evaluate/understand how layered neurons convert features into predictions and why network shape and activation functions matter.

How it connects

Connects neural networks to machine learning, activation functions, optimization algorithms, forward propagation, and forecasting.

liveBookmanning.com

Read Demand Forecasting Best Practices on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Bayesian Optimization in ActionA neural network is a machine-learning model composed of interconnected computational layers that transform inputs into outputs.
• 

  liveBookmanning.com
  Designing Deep Learning Systems: A guide for software engineersA computational model composed of interconnected parameterized units that learns mappings from inputs to outputs.
• 

  liveBookmanning.com
  Ensemble Methods for Machine LearningA neural network is a parameterized computational model composed of connected units that transform inputs through learned weights and nonlinear functions.

Compare how 4 books explain neural network →

Related concepts

• deep learningused by · Deep learning is a machine-learning approach that uses neural networks with multiple computational layers to learn representations from data.
• machine learningkind of · Machine learning is a field of computational methods in which systems learn patterns or relationships from data to make predictions, decisions, or other outputs.
• autoencoderincludes · An autoencoder is a neural network that encodes input data into an internal representation and decodes it to reconstruct the input or an approximation.
• recurrent neural networkincludes · A recurrent neural network is a neural network that processes sequences by updating an internal state across successive inputs, allowing earlier inputs to influence later computations.
• U-Netincludes · U-Net is an encoder-decoder neural-network architecture with downsampling and upsampling paths connected by skip connections for combining features at multiple spatial scales.
• gradient descentuses · Gradient descent is an iterative optimization method that updates parameters opposite the gradient of an objective function to reduce its value.

This explainer was generated by Manning's knowledge graph from the text of Demand Forecasting Best Practices. The section links point to the parts of the book it draws on.