CSC 5616 - ML for Signal Processing Applications

• Characterize systems and signals in terms of continuous versus discrete for dependent and independent variables, sampling rates and uniformity, dimensionality of dependent variables, seasonality (periodicity), and trends
• Convert signals between time and frequency domain using Fourier transforms
• Identify limitations and errors such as aliasing and spectral leakage that occur during spectral analysis due to inadequate sampling frequency, sample lengths, and windowing
• Design sampling schedules and choose windowing functions based on tradeoffs in generated artifacts as appropriate for a given problem
• Implement filters such as moving averages using the convolution operator
• Describe tradeoffs in an examples of time- and frequency-domain filtering techniques
• Perform time-series forecasting using engineering features that capture seasonality
• Perform anomaly detection and segmentation using windowing techniques and statistical tests as well as deep learning models
• Perform signal classification with deep learning models
• Apply and interpret metrics for machine learning on signal data
• Apply preprocessing techniques such as smoothing, trimming, scaling, and imputation to improve performance of learning tasks on signal data
• Motivate the use of convolutional neural network (CNN) architectures for deep learning on signal data
• Design sequential pipelines for complex tasks such as segmentation followed by classification

• Python programming experience including the use of data science and machine learning libraries
• Able to perform and interpret matrix and vector arithmetic including addition, dot products, and matrix-vector multiplication
• Able to prepare tabular data for machine learning tasks including feature extraction, scaling, and imputation
• Able to train and apply various classical machine learning models for classification and regression problems
• Able to design and execute experiments to evaluate machine learning models on tabular data sets
• Able to interpret metrics such as accuracy, precision, and recall to evaluate model prediction performance

• Terminology for describing signals such as continuous versus discrete for dependent and independent variables, sampling rates and uniformity, dimensionality of dependent variables, seasonality (periodicity), and trends
• Spectral analysis with Fourier transforms
• The convolutional operator and its application to signal filtering
• Feature engineering to capture seasonality (periodicity)
• Review of neural network concepts including neurons, activation functions, layers, tensor shapes, loss functions, and choices of optimization algorithms
• Dense (DNN) and convolutional neural network (CNN) architectures
• Statistical testing
• Signal processing related problems such as segmentation, anomaly detection, classification, and filtering
• Composing sequences of simple problems to solve complex signal processing tasks