Authentication of Ancient Pictures

Problem Description

You have been studying Artificial Intelligence for some time. An old friend of your father's, a famous archaeologist and art critic, heard about this and asked for your help. You need to design an algorithm that can classify ancient paintings as either authentic pieces or replicas.

Since professional authentication is expensive, the research team has obtained authenticity labels for only a small portion of the paintings. For most samples, authenticity remains unknown. It is known that the digital characteristics of paintings exhibit strong structural patterns. You are tasked with using all available samples - including those with unknown labels - to train a classification model for ancient painting authenticity.

Dataset

The dataset consists of a training set, a validation set, and a test set, each containing 500 independent samples.

Training set (training_set.csv)

• The first column represents the painting ID (SampleID).
• The next five columns represent the digital characteristics of each ancient painting (feature1, feature2, feature3, feature4, feature5, feature6).
• The seventh column (label) contains the label: 1 for authentic, -1 for replica, and 0 for unknown.

The training set is used for training models and can be accessed and downloaded directly during the competition.

Validation set (validation_set.csv)

• These are similar to the training set format, but do not contain the label column.

⚠️ The validation set is used for calculating the Leaderboard A (public) score and was not directly accessible during the competition.

Test set (test_set.csv)

• These are similar to the training set format, but do not contain the label column.

⚠️ The test set is used for calculating the Leaderboard B (private) score and was not directly accessible during the competition.

Task

Your task is to train a suitable model capable of predicting the authenticity of paintings from the test sets, despite the large number of unlabeled samples.

Submission

You need to generate a submission.csv file that will have the following format:

1SampleID, ClasamentType, label2VALIDATION_001, A, 13TEST_001, B, 1

where:

• SampleID represents the painting id (provided in validation_set.csv or test_set.csv).
• ClasamentType will have the value A for paintings from validation_set.csv and will have the value B for paintings from test_set.csv.
• label is the label predicted by the model.

Score

The evaluation metric will be classification accuracy, defined as the proportion of correctly predicted samples out of the total number of evaluated samples.

The final score is calculated based on the accuracy obtained using the following rules:

• accuracy ≥ 0.9 → 100 points
• accuracy ≤ 0.5 → 0 points
• For intermediate values, proportional scoring is awarded between 0 and 100.