| Author | Message | ||
     David Russell (russell) Senior Member Username: russell Post Number: 41 Registered: 2-2001 |
I'd tend to agree with Dr Davies on this one. I too applied ANN technology to a NIR measurement many years ago when NIR instruments were still using DOS. Subsequently elmininated the need for a NN by upgrading the analyzer to a newer model. The same problem was then described by a very few PLS factors | ||
| Tony Davies (td) Moderator Username: td Post Number: 181 Registered: 1-2001 |
Hello Ami, The first question to ask is "Why do you want to use NNs?" Do you have a lot of NIR data? The experience with NIR and NNs is that you need thousands of spectra before it becomes useful. Then it is often good but with little data it is too easy to discover very good calibrations which have very poor predictive performance. With NIR calibration it is important to remember that the aim is NOT to get a fantasticly good calibration! The aim is to get a calibration which gives good predictions. Best wishes, Tony | ||
| Gustavo Figueira de Paula (gustavo) Junior Member Username: gustavo Post Number: 6 Registered: 6-2008 |
Hi Ami, Before spending your money, try Weka - it�s free, easy and powerful. Although limited to Multilayer Perceptron, has lots of powerful classification algorithms and filters. Anyway, I believe that MLP is the best choice for a basic work. Download it from http://www.cs.waikato.ac.nz/ml/weka/ Hugs, Gustavo. | ||
| Amina Isma�l (ami) New member Username: ami Post Number: 1 Registered: 11-2008 |
Hello everybody, I would like to buy a software to do neural network. I want a software which is not too sophisticated for basic work. To you, which one is the best for me? Thanks |