The purpose of this thesis is to present a novel method of learning to generate an image sequence from input a single image without sequentially aligned data. Given examples of a visual phenomena that can be divided into discrete time steps, the problem is to learn a model that takes an input from any such time and realizes this input at all other time steps in the sequence. For example, given a scenery picture in spring and output the corresponding pictures in sequence of summer, fall, and winter without changing overall layout and semantic information presented in the input picture. Furthermore, it is assumed that ground-truth aligned sequences is not provided. This broadens the real-world application of this method as it is often difficult to collect aligned sequential data for many problems. This task generalizes the unpaired image-to-image problem from generating pairs to generating sequences and associates a direction of time with the phenomena observed.

We show that this problem can be solved by incorporating Generative Adversarial Networks (GAN), a popular deep unsupervised learning technique, and a periodic assumption about the sequential visual phenomena modeled. The periodic assumption is enforced in model training by a novel Loop Consistency loss, inspired by the popular Cycle Consistency loss that has achieved huge success in unpaired image-to-image transformation. The two parts of the model, GAN and Loop Consistency, can be seen as two levels of constraints that together facilitate model training. The transformation unit itself is a neural network. We show the effects of different network architecture changes on generation quality and present the results of the model in comparison with several competitive baseline models.

With the physics of coherently upconvert ultrafast driving laser photon, the high harmonic generation process enables a tabletop laser source that is able to cover EUV to X-ray with a fully coherent attosecond burst. Its spectroscopic feature and extreme spatial and temporal resolution made it emerge as a novel tool for scientific research over the last three decades. The main limitation of this source to be is the brightness which is highly related to the phase-matching condition and the efficiency of optics after generation.The first observation of high harmonic generation is with a UV excimer laser with 248 nm over 30 years ago. This thesis presents works performed with the modern engineering design of high harmonic beamlines and a better understanding of phase-matching and physical models that were developed nearly thirty years. Chapter 3 is the selected experimental result of post-laser manipulation of laser parameters to a shorter pulse duration for more efficient high harmonic emission. Chapter 4 is a summary of our modern design of the beamline with hollow waveguide manufacturing and novel EUV to X-ray spectrometer design in both transmission and reflection mode with calculation and optimization where record-high efficiency spectrometer is designed and tested that can capture the most sensitive signal with physical significance that possibly ignored in the past. Chapter 5 presents an investigation of the feasibility of high harmonic spectrum tunability by shifting the spectrum of the driving laser in both directions to enable the spectroscopic study of metal resonance absorption edge. Finally, Chapters 6 and 7 present experimental results of the 400nm UV laser driving high harmonic with the robust parameter space for bright emission in the UV and X-ray regime for future application. The stable and robust harmonic emission is used as a platform for the spectroscopic study of the possible correlation effect between electrons and ions. For the first time, a double electron recombination process with an extended high harmonic cutoff in two-electron atoms has been experimentally observed which could pave the way for a better understanding of the few electron systems and contribute to further completing the theory of high harmonic generation.