Entrance Examination in Computer Science
for students to be
transferred from other universities to MIPT
(major 01.03.02 “Applied mathematics and Computer Science”)
Test Format
The entrance test
shall take the form of a combination of a practical and oral examination.
1. The
exam takes place in accordance with the Regulation on the readmission after
expulsion and transfer from other organizations and the Regulation on the
procedure for conducting entrance examinations at MIPT.
2. To complete the exam practical part a computer with installed
software is used: a text editor or development environment, a C++ compiler and
a Web browser with access to the resource
contest.yandex.ru
(hereinafter -
Yandex.Contest system).
3. In the practical part of the entrance exam 5 tasks are required to
be solved
by test takers to be
transferred to semester 2 and 6 tasks by test takers to be transferred to
semester 3 through semester 8.
Solutions
should be submitted in Yandex.Contest system. During the practical part usage
of any web-resources is not allowed. Access is allowed only to the
Yandex.Contest system, as well as reference materials on standard libraries for
valid programming languages.
4.
The total practical test time is 2 hours for the test takers to be
transferred to semester 2 and is 2 hours and 30 minutes for test takers to be
transferred to semester 3 through semester 8, with no breaks between them.
5.
The oral part of the entrance exam include
- the discussion of the tasks which were submitted in Yandex.Contest
system and were solved by test takers;
- an answer to an exam card containing theoretical part (the total preparation time is 1
hour).
6. The
oral part is conducted in the format of an interview. The preparation time is
not provided. The exam cards in theory consist of
- 2 questions from “Programming and algorithms” section when
transferring to semester 2 through semester 3;
- 2 questions: one from “Programming and algorithms” section and one
from “Formal languages and translations” when transferring to semester 4
through semester 6;
- 3 questions from various sections when transferring to semester 7
through semester 8.
7. The
total duration of the oral part: 30 minutes.
Technical features of the remote exam
During the
practical part, remote monitoring will be carried out through the MIPT
proctoring system located at: http://exams.mipt.ru.
An applicant must
register in the proctoring system before the entrance exam and check the
technical feasibility of connecting to the proctoring system. Failure to
connect to the proctoring system is equivalent to failure to appear on the
practical part.
The oral part is conducted using one of the
commonly used group communication tools (Hangouts, Google Meet, Zoom, etc.).
Entrance Examination in
Computer Science
“Programming and algorithms”. Theoretical questions.
semester 2
semester 3
semester 4
through semester 8
“Formal languages and translations”. Theoretical questions.
semester 4
through semester 8
“Machine
learning”. Theoretical questions.
semester 7
through semester 8
1. Basic concepts of machine learning. Standard problems
(classification, regression, clustering). Examples of quality metrics. Examples
of simple algorithms solving standard problems: kNN, K-Means, naive Bayesian
classifier.
2. Quality metrics in classification and regression problems
(accuracy, precision, recall, F-measure, ROC-AUC, logloss, MSE, MAE, quantile
loss, MAPE, SMAPE). Feature engineering: feature extraction, categorical
feature encoding.
3. Linear methods of classification and regression. Loss functions
and regularizers. Stochastic gradient descent method. Logistic regression
optimization problem and estimation of class membership probability.
4. Linear methods of classification and regression. Support Vector
Machine optimization problem.
5. Decision trees in the classification problem and in the regression
problem. Decision tree ensemble: random forest and gradient boosting above the
trees.
6. Decision trees in the classification problem and in the regression
problem. Bias-variation trade-off (without proof). Analysis of boosting and
bagging using bias-variation trade-off.
7. Neural networks, training (backprop), convolutional networks
layers (dance, conv, pooling, batchnorm, dropout), nonlinearity (relu vs
sigmoid, softmax), loss functions (logloss, l2, hinge).
8. Recurrent neural networks, training (backprop tt), the difference
between recurrent and convolutional networks, recurrent layers (RNN, LSTM,
GRU), examples of usage.
9. Problem of clustering. Agglomerative and statistical clustering
methods. Lance-Williams formula, K-Means algorithm.
10.The problem of dimensionality reduction (reducing the dimension of
the feature space). Principal component analysis (PCA) and tSNE (for both
methods: basic idea, without proof).