CMSC5707 Advanced Topics in Artificial Intelligence

Dept. of Computer Science and Engineering, MSC Programme in Computer Science, Time table2023-4

CMSC5707 Advanced Topics in Artificial Intelligence

This web page contains information about the course CMSC5707: Advanced Topics in Artificial Intelligence
This course will cover selected topics from: advanced pattern recognition, neural networks, expert systems and fuzzy systems, evolutionary computing, learning theory, constraint processing, logic programming, probabilistic reasoning, computer vision, speech processing, and natural language processing.
Edited:22 Nov. 2023

Web based support : https://blackboard.cuhk.edu.hk

CMSC5707 - Term 1 - Wed (6:30-9:30pm) from 6 Sept.2023, Teaching time table 2023-24
Venue: YIA LT4. See classroom_info and cuhk-campus-map
Dates: Every Wed. (12 lectures).

Data for assignments. Tools for assignments: MATLAB online , COLAB, Colab_tutorial, keras_examples , SKLEARN, anaconda(python system)
data57073_1.zip

Final Exam: Date: Wed. 6 Dec. , venue:

CYT_201 ,if your surname starts with A-S.
CYT_202 ,if your surname starts with T-Z.

The exam will start from 6:45pm (please come to the classroom at 6:30pm). See classroom_info and cuhk-campus-map.

The exam will last for 2 hours, it is a closed-book exam and no cheat sheet allowed. However, the CMSC5707_formulas.docx will be printed and attached to your exam question paper. The exam will cover and include all materials taught in the lectures. We will provide each student an answer book, supplementary sheets and scrap paper for calculation. A non-programmable calculator can be used. We accept calculators that are in (or similar to that) the list https://www.hkeaa.edu.hk/en/IPE/hkia/index.html.

Sample exam paper.

Exam_name_list

CTE

Professor : Dr. K.H. Wong ( khwong

)

See AVSU and cuhk-campus-map

No. of Enrollment: (quota 110)

Tentative timetable: (reference only):

Titles: Teaching materials will be updated regularly	Week/date	Topics	Additional information	wav sounds/ Matlab code	Assignments
Overview
Ch1: Introduction to audio signal processing	1(6/9)		octave_guide Guide to use CUHK Matlab licenses -- for CUHK students	tz1, tz2, sor1, violin3x A4_oboe , trumpet , sor1, A4_violin ,A5_flute num1, num5
Ch2: Pre-processing of audio signals	2 (13/9)	Time & frequency domain processing, Fourier transform, spectrogram		demo_spectrogram_release16.rar demo_time_frame1_fft.zip (2023) demo_dft_tutorial.rar
Ch3: Feature extraction from audio signals	2	Filtering, LPC , Cepstral coefficients		demo_time_frame1_fft_lpc.rar
Ch4: Feature representation and compression:Vector quantization	3(20/9)	VQ, K-means		demo041_kmeans_demo_3D video
Ch5: Speech Recognition	3	End point detection, per-emphasis, distortion measures, dynamic programming	5707_revision_Q&A_audio.doc		Assignment1( released at https://blackboard.cuhk.edu.hk )
Ch6: ensemble methods. Classification using Adaboost,	4(27/9)	Classifiers, boosting
Ch7: Face detection	5(4/10)	Face features, Attentional cascade,
Ch8: Neural Networks	6(11/10)	Feed-forward, feed-backward networks, training and recognition	https://sites.google.com/site/hongslinks/tensor_windows https://keras.io/examples/ Ch8b: Softmax (optional)		Assignment2( To be released at https://blackboard.cuhk.edu.hk )
Ch9: Convolution Neural Network CNN	7(18/10)	Convolution Neural Network CNN	Tensorflow tutorials 5707_tf_keras.pptx https://colab.research.google.com/
Ch 10. Auto-encoder	8(25/10)	Auto-encoder	5707_0math_probability.pptx		Code examples / Generative Deep Learning / Variational AutoEncoder
Ch11. RNN_LSTM	9(1/11)	Recurrent Neural Net, Long short-Term Memory			Code examples / Natural Language Processing / Bidirectional LSTM on IMDB
Ch12. Word representation and seq2seq	10(8/11)	Word embedding, Machine translation, Transformer			Assignment3 (To be released at https://blackboard.cuhk.edu.hk ) Tutorial_LSTM_music_genre_classification.docx
Ch13. Decision tree	11(15/11)	Decision tree	https://scikit-learn.org/stable/
Ch14. Fuzzy	12(22/11)	Fuzzy inference systems
	13(29/11)	No lecture
		Final Exam, time/venue: see above

The followings are other advanced topics (not to be covered in the final Exam)

Regression (include logistic regression)
SVM (optional) Support vector machine

(Logistic_khw2.zip) code

Expectation Maximization (draft)
Gan (Generative Adversarial Networks) preliminary draft
Music genre classification

preliminary draft

CTE

Marking scheme and Exams

Part 1: Exercises and programming assignments 40% of the overall assessment.

Assignment 1,2,3. Each assignment carries equal marks.

Part 2: Final exam 60% of the overall assessment.

A student must gain at least 40% of the full marks in each part in order to pass the course.

The exam will start from 6:45pm (please come to the classroom at 6:30pm). See classroom_info and cuhk-campus-map.

The exam will last for 2 hours, it is a closed-book exam and no cheat sheet allowed. However, the CMSC5707_formulas.docx (draft will be updated later) will be printed and attached to your exam question paper. We will provide each student an answer book, supplementary sheets and scrap paper for calculation. A non-programmable calculator can be used. We accept calculators that are (or similar to that) in the list
Closed-book exam, no cheat sheet. A non-programmable calculator can be used. We accept calculators that are in (or similar to that) the list https://www.hkeaa.edu.hk/en/IPE/hkia/index.html.

Sample exam paper.

Important information