Supervisor's review for a bachelor's thesis a fourth-year student of the Department of Informatics of the Mathematics and Mechanics Department Kutukova S.Е.   Bachelor's work Kutukova S.Е. is devoted to the problem of detecting anomalies on chest x-ray images. The task is actual due to the shortage of qualified radiologists for the primary description of the pictures in a massive regular population survey, which is the main method of early detection of tuberculosis, other chronic infectious diseases and oncological diseases of the chest.  The purpose of the work was to test the hypothesis of the possibility of detecting anomalies using an autoconcoder trained only on the class of norms for open image bases when comparing the original and decoded images.  The following results are achieved in the work: • Competitive auto-encoder trained; • A conventional convolutional auto-encoder is trained; • The hypothesis of the possibility of comparing images was tested, the hypothesis of confirmation was not received; • Implemented and trained a hierarchical classifier consisting of a convolutional auto-encoder and SVM; • Implemented and trained a hierarchical classifier consisting of a convolutional autoencoder and a random forest of decision trees; • Comparison of the results with the published results, better metric AUC-ROC = 0.61 for a random forest; best published result 0.84  The results are of great practical importance for the choice of classifier architecture in solving the problem of finding anomalies.  The work has also disadvantages: • the exact cause of the low accuracy of the proposed method is not accurate; • the text of the thesis note has some stylistic shortcomings.  Despite the shortcomings, considering the valuable practical results of the work, I believe that the bachelor's work of S.Kutukov. deserves an evaluation - "EXCELLENT".  f. Mr., Art. prep. Department of Informatics Faculty of Mathematics and Mechanics of St. Petersburg State University Salischev SI 05/15/2018