Case based reasoning CBR Neural networks NN

After the database is ready, we can query the rules using an inference engine IE, which is at the heart of the ES. We present the IE with observed signs, and it will search the rules, looking for those which fit the observations. Using those rules, the IE will try to find diseases whose rules have ‘fired’. Going the other way, using a list of possible diseases, the IE will enable us to see other rules and define other observations needed to make the differential diagnosis. The advantages of this technique are that it is mature, there are many systems to choose from, the system can ‘explain’ it’s results by showing those rules used to reach an answer and rules can be edited so that the IE is tuned to obtain better results. Another strength is that in most cases the IE uses an existing ES-shell so the development will not entail extensive programming as such, only the entry of rules Tu´nez et al., 1996. The main deficiency of rule-based systems is that they require some ‘deep’ knowledge in order to be truly effective. This means we need to know the causal relationships between signs and disease. Another problem is that the rule database grows very rapidly as the problem space expands. As a result, the computation time grows and for the problem space we needed, it became totally impractical. In Fish-Vet we use a rule based system with a very limited set of rules, whose only purpose is to cut down the problem space to a manageable size quickly. This entails rules for species membership i.e. some diseases are species-specific so that for any diagnosis we can ‘throw out’ all species-specific diseases not belonging to the species in question. A similar approach is used for water type i.e. seawater specific diseases can be disregarded when looking at a freshwater fish, etc.

4. Case based reasoning CBR

This method uses the storage of a large number of previously solved cases in some normalized form. When the user presents a new case to the system, it searches the database, locates ‘similar’ cases and presents them to the user. By looking at those similarities, the user is able to formulate a presumptive diagnosis and is guided toward further data collection in order to narrow the list of possible problems Althoff et al., 1998. The main advantages of this method are that systems become better as more cases are entered into the database, the program has the ability to explain it’s decisions and the fact that there is no need for programming Evans and Winter, 1995. The main problem with CBR in our case is the need for a large number of documented cases for each condition we would like to diagnose. These are lacking in the field of fish disease even though there are existing databases for other species, the largest being the veterinary medical data base VMDB at Purdue more than 5 million cases. Another problem inherent in CBR is that as the problem space grows, there is an exponential growth in the number of cases needed such that the system quickly becomes unwieldy Evans, 1995. For this reason, we did not utilize this method in our program.

5. Neural networks NN

Neural networks NN are a computer ‘simulation’ of the interconnected neurons in our brains. This method simulates a network of interconnected neurons usually arranged in three layers, where the lowest one receives inputs and the top signals the outputs. Work with a NN starts with a ‘clean’ network and the developer has to ‘train’ it to recognize a specific set of inputs. Training consists of presenting the NN with a series of cases and providing feedback as to the desired output. During the training the NN adjusts the ‘weights’ given to each input neuron till the NN will give the desired output i.e. until the NN learns to recognize this set of inputs; Kappen, 1996. The advantages of this method lie in the fact that the network ‘learns’ by itself so that there is no need for a priori knowledge. Therefore, to obtain to the right output a NN can often find relationships of which we were not aware Babic et al., 1995; Smith et al., 1996. The problems with NN are that there is a need for a large number of test cases for each disease of interest and that the NN cannot ‘explain’ it’s results. As discussed above, there is a dearth of well documented test cases to use in training the NN, and the ability to explain results is critical in the context of disease diagnosis. Therefore, we could not use NN in Fish-Vet.

6. Fuzzy logic systems FL