CILQ: An Evolving Modern Curriculum
The Communications, Information, Learning and Quantum Group (CILQ) is made up researchers in the Electrical and Computer Engineering Department who are working on a diverse set of challenges. Our work is at the forefront of many important emerging technologies and our curriculum is constantly evolving to give our students the skills necessary to succeed in these fields today.
CILQ continuously modernizes its teaching to be up to date with the developments that impact the careers of our students. A typical example is the recent revamping of the ECE 535 (Wireless Communications) and ECE 635 (Advanced Wireless Communications). These courses form the basis for the jobs many of our students get at manufacturers such as Samsung, Nokia, and Qualcomm, as well as operators like Verizon and AT&T, and many startups. Thus, ECE 535 recently moved the emphasis from the traditional "cellphones" approach to those aspects (such as multi-user scheduling, smart antennas, etc.) that are relevant for internet of things and high-speed communications with smartphones. We also present up-to-date information on the 5G standards (3GPP NR, WiFi 6), and cutting-edge research topics such as massive MIMO and mm-wave communications.
The scale of data collection has reached unprecedented levels and it will only increase in intensity and scope. For example, in Facebook alone, more than 4 petabytes of data is generated/collected from the users each day. This *personal* data is often shared (or sold) with first and third-party entities to run machine learning algorithms that learn the preference/behavior/interests of the users.
There is now an increasing awareness about the privacy, security, transparency, and fairness for individuals and the society as a whole. The new CILQ course: "EE599: Foundations of Secure and Private Computing: from Machine Learning to Blockchains” has been created to address this trend. This course provides a foundational view of the principles and algorithms for enabling secure and private computing, in particular for applications in machine learning and blockchains. The course consists of five main parts: (1) Information theoretic measures of security and privacy and basic cyphers for secure communication; (2) basic key exchange and public-key encryption; (3) secure and private multi-party computing; (4) secure and privacy-preserving machine learning; and (5) bitcoin and cryptocurrencies.
An EE599 Special Topics class on Deep Learning was offered during the Spring 2019 semester and is similarly scheduled for Spring 2020. It is planned to convert the class to a regular course as part of the ECE department’s new MS degree in Machine Learning and Data Science. This class is intended to complement and build on our more analytical graduate level courses and utilizes a “learning by doing” educational paradigm.
With only a minimal programming background and no prior knowledge of machine learning, students developed software for increasing realistic problems as the class progressed. This included classification of facial expressions using convolutional neural networks (CNNs) and language classification of speech using recurrent neural networks (RNNs). This was part of the process of learning to develop and work with large datasets. Students programmed the training of neural networks using just the standard NumPy Python library before learning and utilizing the industry-standard deep learning toolkits (Keras, TensorFlow, and PyTorch).
As the scope of the assignments increased, students moved from training on their personal computers to training in the cloud using GPUs. Amazon Educate generously supported these cloud computing assignments with $15,000 in AWS cloud computing credits provided directly to enrolled students. The class culminated with a Deep Learning Symposium in which student teams presented the results of their self-defined projects to the USC community. In addition to their presentations, students also produced final reports for their projects and YouTube videos.
More Information Coming Soon