ENG4200 Introduction to Artificial Intelligence

Coursework 2: Neural networks

ENG4200 Introduction to Artificial Intelligence and Machine Learning 4

1. Key Information

• Worth 30% of overall grade
• Submission 1 (/2): Report submission
• Deadline uploaded on Moodle
• Submission 2 (/2): Code submission to CodeGrade
• Deadline uploaded on Moodle (the same as for report)

2. Training dataThe training dataset has been generated by maximum flow analysis between nodes 12 and 2. The

eature dataset has 19 fields, which of each represents the maximum flow capacity of each of the

19 edges, taking the values of 0, 1, and 2. The output dataset has 20 fields, where the first 19fields refer to the actual flow taking place on each of the 19 edges, and the lastone refers to themaximum flow possible between nodes 12 and 2.Figure 1 The network used to generate training dataset. This information is just to help you understand the training requirement/note:Only the provided training dataset should be used, i.e. furthur traning dataset must NOT becreated by performing maximum flow analysis over the network in Figure 1.

• The accuracy on a hidden test dataset will be evaluated by a customised function asfollows, where the accuracy on the maximum flow field is weighted by 50%, and other 19fields share the rest 50% (you may design your loss function accordingly):You should prepare two submissions, code submission and report submission. In blue colour areassessment criteria.

• Code submission should include two files (example code uploaded on Moodle):

o A .py file with two functions

▪ demo_train demonstrates the training process for a few epochs. It has threeinputs: (1) the file name of taining feature data (.csv), (2) the file name oftraining output data (.csv), and (3) the number of epochs. It needs to do twothings: (1) it needs to print out a graph with two curves of training andvalidation accuracy,respectively; and (2) save the model as .keras file.

▪ predict_in_df makes predictions on a provided feature data. It has two

inputs: (1) the file name of a trained NN model 代写ENG4200 Introduction to Artificial Intelligence  (.keras) and (2) the file nameof the feature data (.csv). It needs to return the predictions by the NN modelas a dataframe thathas the same format as ‘train_Y’.o A .keras file of your trained model ▪ This will be used to test the hidden test dataset on CodeGrade.o You cantest your files on CodeGrade. There is no limit in the number ofsubmissions on CodeGrade until the deadline.o Assessment criteria▪ 5% for the code running properlyaddressing all requirements.▪ 10% for a third of the highest accuracy, 7% (out of 10%) for a third of thesecond highest accuracy, and 5% (out of 10%) for the rest.

• Report submission should be at maximum 2 pages on the following three questions andone optional question:o Parametric studies of hyperparameters (e.g. structure, activators, optimiser, learningrate, etc.): how did you test different values, what insights have you obtained, andhowdid you decide the final setting of your model?

o How did you address overfitting and imbalanced datasets?o How did you decide your loss function?

o [Optional] Any other aspects you’d like to highlight (e.g. using advanced methodssuch as graphical neural network and/or transformer)?

o [Formatting] Neither cover page nor content list is required. Use a plain worddocument with your name and student ID in the first line.o Assessment criteria▪ 5% for each of the questions, evaluated by technical quality ANDwriting/presentation▪ Any brave attempts of methods (e.g. Graphical Neural Network, Transformer,

 Physics-Informed Neural Network using physical relationships e.g. thatthe flows going in and out of a node should be balanced) that go beyondwhat we learned in classroom will earn not only the top marks for report, but (unless the accuracy is terribly off) will earn a full 10% mark foraccuracy in the code submission part. If you have made such attempts, don’tforget to highlight your efforts on the report.