Create a model that can deliver the following values for generated molecules:

• at least 95% validity
• at least 95% uniqueness
• at least 20% novelty

You can freely adjust training and model hyperparameters.

Always generate 200 compounds and leave temperature at 1.0 and maximum length at 100.

You can also try generating several times- the percentages will vary a bit and it might be enough to re-generate the samples to get a result that fulfills the requirements above.

List your choice of hyperparameters here and include a screenshot that shows the generated molecules along with the validity/uniqueness/novelty values.

Now we want to set up a random forest model. To find the optimal hyperparameters, do a grid search.

You get full points if you find a parameter combination that will get you more than 0.85 test AUC, half points if you get more than 0.825 test AUC.

Report the best set of hyperparameters that you found here. Also, include two screenshots/images:

• the grid search result
• the ROC curve of the model with your optimal hyperparameters

You can start with the default grid search settings but feel free to change/extend them. Just consider that if you add a lot of options for the various hyperparameters, the computational effort can grow quite significantly.

Run a k-Nearest Neighbor analysis on the data set, with default settings (k=5, euclidean distance).

Look at the ROC plot. What are your train/test accuracies and how do you interpret them? Also attach the ROC plot here.

Create the Morgan fingerprints with radius = 2 and fingerprint size = 1024, leave variance threshold at 0.0.

Take a look at the feature fraction plot. What is the average fraction of feature presence?

Create a downprojection by reducing the dimensionality to 16 components with PCA first, and then to 2 dimensions with t-SNE (perplexity: 10) and attach the plot here. How do you interpret the plot? Are there any visible clusters?

Attach the feature fraction plot here. How do you interpret the plot and the percentage of feature fraction?

Take a look at the molecules in cluster 10 and comment on their structure. Do you see any common substructures? If yes, what? Are there any outliers? Attach the image of the molecules here also.

Look at the cluster plot (also attach it here). Comment on the separability of the detected clusters with respect to the downprojection.

Go to the 'Transformers - Text to Speech' tab. Record yourself with the text:

Good morning. We will now check how you can understand me.

and try to get a flawless transcription. Include a screenshot of the transcribed text with the spectrogram. Also try this with your native language (if it is not English) - can the model handle it too? Also include a screenshot of that attempt.

In the 'Transformers - Attention' tab, enter the following text:

AI agents might soon overtake the world

Set d_model = 16, leave scaling on enabled and set a seed of 2026. Which word in the sentence does the word 'agents' most strongly attend to?