NeuroIntelligence

Category Intelligent Software>Neural Network Systems/Tools

Abstract NeuroIntelligence is a neural network (NN) software application designed to assist experts in solving real-world problems. NeuroIntelligence features only proven algorithms and techniques; it is fast and easy-to-use. Product features/capabilities include:

All stages supported -- NeuroIntelligence supports all stages of neural net design and application. It can be used to:

1) analyze and preprocess datasets;

2) find the best neural network architecture;

3) train, test and optimize neural network;

4) apply the network to new data.

Advanced visualization – Offers visualized architecture search, network training, and testing. Architecture search: fitness bars, training graphs comparison. Training graphs: dataset error, network error, weights and errors distribution, input importance. Testing: actual vs. output graph, scatter plot, response graph, Receiver Operating Characteristic (ROC) curve, and confusion matrix.

Improve your productivity -- The interface of NeuroIntelligence is optimized to solve forecasting, classification and function approximation problems. You can create a better solution much faster using the application's easy-to-use graphical user interface (GUI) and unique time-saving capabilities.

Note: Many processes are automated but you can understand the underlying behavior with graphs, statistics and reports. Options and parameters are intelligible, comprehensive and accessible with one mouse click.

Product key features/capabilities include:

Data analysis and preprocessing -- Saves you time with automated column identification and encoding. Identify and encode date/time columns. Perform automatic and manual data partition. Identify and replace/remove data anomalies. Analyze your data with a response graph.

Automated network selection -- Select fitness criteria along with architecture search method and NeuroIntelligence does the rest. It automatically finds the best architecture offering you graphs of the search process and details for every tested neural network.

Proven training algorithms -- The most efficient algorithms, such as Conjugate Gradient Descent, Levenberg-Marquardt, Quick-Propagation, variations of Quasi-Newton and Back-Propagation, are available for neural network training.

Proven techniques -- Improve network performance with early-stopping, generalization loss control, outliers and missing values handling, dataset partition, jogging weights, different activation and error functions, weights initialization and more.

Ease of use -- NeuroIntelligence was developed with the highest attention to usability. It has an intelligible layout, options, reports and descriptions. It offers graphs and results visualization at all stages. You can save all graphs and tables, as well as other results with a single mouse click. Local toolbars and online help are always accessible.

Application in several clicks -- After a neural network is tested it can be easily applied to new data. Results are visualized with a response graph. You can apply the selected network to a single case, data file or records from your input dataset. The whole project or only selected neural networks can be saved for future use.

Additional product features/capabilities include:

Analyze and Pre-process Your Data --

1) Import Excel files;

2) Import popular American Standard Code for Information Interchange (ASCII) file formats [Comma Separated Values (CSV), TXT, PRN];

3) Custom date formats and file structure definition;

4) Input dataset size is limited only by the hardware of the computer;

5) Categorical values encoding;

6) Numeric values scaling;

7) Min/max values specification for numeric columns scaling;

8) Missing values handling for both numeric and categorical data;

9) Outliers handling for numeric data;

10) Automatic recognition of data entry errors (wrong type values); and more.

Design Neural Network --

1) Input feature selection [Genetic Algorithm (GA), stepwise, exhaustive];

2) Manual architecture specification (up to 5 hidden layers for multi-layer perceptron);

3) Heuristic architecture search with customizable range of search and sensitivity;

4) Exhaustive architecture search;

5) Customizable search range and search sensitivity;

6) Detailed statistics for each tested architecture;

7) Network fitness criteria: Akaike’s Information Criteria (AIC), Test set error, Correlation, R-squared;

8) Error functions: Sum-of-Squares, Cross-entropy;

9) Classification model: Winner-takes-all, Confidence-limits (Accept/Reject levels);

10) Automatic adjustment of learning rate and momentum for Back-Propagation algorithm; and more.

Control Network Training Process --

1) Real-time training error graph;

2) Real-time control on training parameters:

3) Continue training with new parameters;

4) Jog weights;

5) Early-stopping on generalization loss;

6) Retain and restore best network;

7) Stopping conditions:

8) Automatic network retrains and selection of the best network among retrains;

9) Retrains statistics;

10)Weights initialization: manual randomization range; optimized for Uniform or Gaussian distribution; and more.

Apply Network --

1) Enter new cases manually or insert from the Clipboard;

2) Load new cases from a new data file;

3) Apply to selected records from your original dataset;

4) Graphical network output representation;

5) Output representation with Results Table;

6) Confidence limits for network output;

7) Save results in a separate file or copy them to the Clipboard.

General --

1) Customizable interface;

2) Detailed reporting;

3) Online help system;

4) Free technical support;

5) Project files to keep all related information in one place;

6) Sample financial, marketing, real estate and scientific problems included.

Note: See G6G Abstract Number 20111 for additional product info from this manufacturer.

System Requirements

Manufacturer

Manufacturer Web Site NeuroIntelligence

Price Single User $499.00; Unlimited Site $4,999.00

G6G Abstract Number 20110

G6G Manufacturer Number 100225