Here is a list of best free neural network software for Windows. Using these software, you can build, simulate, and study artificial neural networks. These software can be used in different fields like Business Intelligence, Health Care, Science and Engineering, etc. for the simulations of artificial neural networks. In one of these, you can simulate and learn Neocognitron neural networks specifically. You can learn about different problems like Approximation, Classification, Forecasting, Association, Actor-Critic, Cortical Circuit, etc. Many of these come with some samples of neural network projects. So, the beginners may use them to understand neural networks and software functionality.
You can build a neural network with neurons or a group of input, hidden, and output nodes and then analyze it. You can view real time simulation of the generated neural networks. For simulation purpose, you can customize some learning control parameters like learning rate, validating rules, slow learning options, target error stops, etc. Many of these provide bar chart, pie charts, histograms, time series, projection plot, error graphs, etc. visualizations of neural network simulation.
Each of these neural network software provide a different set of tools. So, just go through the list to find the one which suits your need.
My Favorite Neural Network Software For Windows
Neural Designer is my favorite neural network software because it can be used for a wide number of applications like Bank Marketing Campaign, Credit Risk Management, Breast Cancer Diagnosis, Activity Recognition, Nanoparticle Adhesive Strength, Banknote Authentication, etc. It provides samples of projects which you can use to simulate neural networks. Plus, it has a clean and intuitive GUI which makes the entire simulation process quite smooth and easy.
You may also like some best free Digital Forensic Tools, Bioinformatics Software, and Phylogenetic Tree Viewer Software for Windows.
Neural Designer is a free and cross-platform neural network software. It can be used for simulating neural networks in different applications including Business Intelligence, Health Care, and Science and Engineering. Some preloaded examples of projects in each application are provided in it. For example, in Business Intelligence, Bank Marketing Campaign, Credit Risk Management, Telecommunications Churns, etc. sample projects are given. For Health Care, Breast Cancer Diagnosis, Activity Recognition, Nanoparticle Adhesive Strength, etc. projects can be simulated. As for Science and Engineering applications, you can learn projects like Banknote Authentication, Concrete Properties Assessment, Tree Wilt Detection, etc.
To start with a neural network from the scratch, you can choose a template to simulate a particular problem, including Approximation, Classification, Forecasting, and Association. You can import datasets by adding data files in formats like TXT, DAT, CSV, XLSX, etc. It divides various tasks into different categories such as Data Set (report data set, calculate data statistics, calculate box plots, calculate targets distribution, calculate correlation matrix, etc.), Neural Networks (report neural network, calculate parameters norm, calculate parameters statistics, calculate parameters histogram, calculate outputs histogram), Training Strategy (report training strategy, perform training), and Model Selection (report model selection, calculate input importance, perform inputs selection, perform order selection). For testing analysis purpose, you can calculate errors, confusion, binary classification tests, ROC curve, cumulative gain, lift chart, conversion rate, calibration plot, and misclassified instances. As for model deployment, you can calculate outputs, plot directional output, calculate Jacobian, and write mathematical expressions represented by the neural network. You can also export output data as TXT, CSV, XLSX, etc. files, export a script in R or python files with the expression of the model, and export the model as a PMML file.
The good part of this software is that its interface is very clean and intuitive. Separate sections named Task Manager, Output, Neural Editor, Neural Viewer, Status Bas, etc. provided for easy utilization and navigation. It also explains each task in the Neural Viewer with the output. So, it will be easier to understand the functionality of this neural network software.
In my opinion, it is one of the best neural network software in this list.
Note: You need to register a free account on its website in order to use this software.
Home Page Download Page
Simbrain is a free, portable neural network software for Windows. This software helps you create and analyze artificial neural networks. It comes with a wide number of sample neural networks which can directly be imported and studied. To start from the scratch, you can build a network by adding new neurons, setting source neurons, connecting them with all to all or one to one connection, inserting network (Backprop, Competitive Network, Echo State Network, Feed Forward Network, LMS, SRN, etc.), adding neuron groups (SOM, WTA, etc.), etc. It lets you configure network preferences including network time step, synapse visibility threshold, connections setting, etc.
There are various kinds of simulation to simulate created neural networks. These include simulations of Actor-Critic, Agent Trails, and Cortical Circuit problems. You can visualize network simulation with bar charts, pie charts, histograms, time series, projection plot, and raster plot. It also lets you run scripts to perform custom simulations. It provides Coupling Manager and Coupling List tools too. While simulation goes on, the time and iteration statistics are displayed on the main interface. A document viewer (New Doc Viewer) is also provided to add instructions to be included in a simulation.
In order to view video tutorials of Simbrain, you can check their official YouTube channel.
Home Page Download Page
JustNN is another free neural network software for Windows. Using this free software, you can train, validate, and query neural networks. It lets you build neural networks by importing data from files like text, CSV, binary, XLS, etc. It provides some sample data files to start building a neural network. You can start with some exercises to get familiar with the software, such as simulation of XOR, Color Circle, Horse Races, etc.
To start with neural networks, you can create a grid with input columns, output columns, training example row, validating example row, and querying example row. You can add grid cell values as integer, real, boolean, and text. You can check the created grid to find problems in it and fix them accordingly. From the formed grid, a neural network can be created with input nodes, hidden nodes, output nodes, and connection weights. You can then start the learning process using Action > Start Learning option. And for this, you can set up some control options like learning rate, validating rules, slow learning options, target error stops, etc. The display mode can be set to Grid, Network, Input Importance, or Learning Progress Graph. You can view the related information of a created network including learning cycles, training error, etc.
Home Page Download Page
Sharky Neural Network
Sharky Neural Network is another free neural network software to study neural network classifications. As you learn a neural network, it displays error value (wrongly classified points), ni (learning speed parameter), age (number of epochs), learning speed (K/s), and time (s) statistics on its interface. It displays points graph in the middle of interface and real-time error graph at the bottom of screen.
The main features of this neural network software:
- You can select a network architecture from some predefined ones and see respective structure details including maximum/minimum of absolute weights/bias, Absolute Arithmetic Mean (AAM), and Root Mean Square (RMS) values.
- It lets you generate learning points by selecting amount and arrangement (square cartesian, circle cartesian, square radial, circle radial). You can save generated points as a points or text file.
- You can select a shape to display learning points on the graph like XOR, circle, square, diamond, ring, face, etc. Also, you can select what to show on the points graph including points, points and network answer, points and areas, or network answers.
- From Learn tab, you can configure some settings like order to sort learning points (fixed, random, swapping permutation, etc.), enable/disable premphase error and verify options, etc.
- To choose 2D graph display options, move to Draw tab. From here, you can select animation speed and error plots (Mean Squared Error, Verify Error, Error, ni) to draw.
- You can learn neural networks in different modes including Hard, Normal, and Soft, and you can select number of epochs for the same.
To learn more about it, you can refer to its online help webpage.
Home Page Download Page
MemBrain is one more neural network software for Windows. This freeware is widely used in industrial manufacturing and technical control applications.
You can create a neural network by inserting neurons as input, hidden, and output nodes, input links, and output links. You can set neuron properties which include normalization settings, input function, activation function, output fire level, output recovery time, etc. It provides tools like Analyze Network, Randomize Network, Shotgun randomization, Set Simulation Speed, etc. A Lesson Editor is provided to manage input/output data. It calculates the activations and output signals of the generated neural network. You can analyze net error and pattern error graphs. It lets you generate C-Source Code from the created network.
This is another nice software to simulate artificial neural networks. In it, you can save and encrypt network files.
Note: It is free for non-commercial or educational purpose only. Home Page Download Page
Neocognitron is another free neural network software for Windows. This software is specially designed to simulate and study Neocognitron neural networks. It comes with some preloaded project samples that you can use to start with.
Here are the main options of this software and their functionalities:
- Set Input U0: Using this button, you can enter an input layer. You can also save the drawn input in a text file.
- Structure: It lets you set input panel size as per the number of layers.
- Patterns: You can draw, import, or save training patterns using this option.
- Train: After creating training patterns and Neocognitron’s structure, you can start the learning process using this button. It displays total learning time as well.
You can view output routing as well as the winner output value and pattern. You can click on a panel to view cell values and export the whole table in a text file. It lets you save a project as a binary file (.bin). Home Page Download Page
Spice-Neuro is the next neural network software for Windows. It provides a Spice MLP application to study neural networks. Spice MLP is a Multi-Layer Neural Network application. In it, you can first load training data including number of neurons and data sets, data file (CSV, TXT), data normalize method (Linear, Ln, Log10, Sqrt, ArcTan, etc.), etc. You can then select training parameters like inputs (random or in turn), number of hidden neurons, activated functions for hidden and output layer, splitting data, stop conditions, etc. In the Training and Testing tab, you can view training graph, weight and average input graph, and learning rate. You can save modeled data as CSV file.
It provides a separate Spice-SOM application which basically displays distances among neurons on output map.
Note: It is free for study purpose only. To commercially use it, you need to first contact the author. Home Page Download Page