cst383 - week 7

    This week focused on encoding categorical variables, logistic regression, and overfitting. While the lectures introduced several important machine learning concepts, I think the homework was what helped reinforce them the most. Unlike many previous assignments, this one was much more open ended and required me to make my own decisions about how to approach the problem. We were given a dataset and told to predict a target variable using machine learning practices that we've learned, so how to go about preprocessing the data and tuning the model was left to me.

    Initially the open endedness of the assignment made it a bit more challenging with not having a clear step by step process to follow, but it certainly made the assignment feel much more realistic seeing as in an actual career, problems arent going to be presented with exact instructions. Being able to evaluate different approaches is an important skill to train. I felt that I had made good decisions in the end and certainly helped me in feeling like I can properly apply what weve learned.

    On the topic of the lectures, one thing that stood out to me was overfitting. It was interesting to learn that creating a model that performs extremely well on training data is not always a good thing. I certainly would have assumed that higher accuracy automatically meant a better model, but learning about overfitting made m realize that a model also needs to generalize well to new data.

    The final exam is next week and I find myself feeling both confident and nervous. There has been a large amount of material covered and I have some worries about being able to retain it all. I plan to spend plenty of time reviewing during what remains of this week.

Comments

Post a Comment

Popular posts from this blog

CST334 Week 2

    This week has been a great introduction to how CPU process selection works in operating systems. We covered a range of scheduling algorithms, including FIFO (First In, First Out), LIFO (Last In, First Out), Shortest Job First (SJF), and Round Robin (RR). I learned how each algorithm has its own strategy for picking which process should run next, and how these choices affect system performance and metrics like response times.      I also learned to calculating those performance metrics such as response time, turnaround time, average arrival time, and average turnaround time. For Round Robin scheduling in particular, I learned how to calculate how long a process will stay in the system based on the time slice and the number of other processes in the queue. These exercises helped me understand the mechanics of how schedulers work and why the design choices made are so important.      One of the most interesting parts was comparing the pros and ...
Read more

CST334 Week 5

     This week I focused on concurrency and how it allows processes to run multiple threads in parallel, allowing tasks to be handled in the background while the main application continues. While I’ve previously worked with multithreading in Java, learning how the operating system manages concurrency gave me a better understanding of how multithreading works in the CPU. I learned that threads can share certain resources, such as virtual address space which makes communication between them efficient, but also introduces its own challenges.      One of the key concepts I learned was nondeterminism in multithreaded programs. Even if two threads call the same function, the order in which memory is accessed and modified can change which can lead to different outcomes every time the program runs. This unpredictability can cause bugs that are hard to reproduce and debug.      To address this, we use thread synchronization APIs like locks, or mutexes...
Read more

CST334 Week 6

       This week we continued concurrent programming and synchronization concepts and built on the ideas from last week. One of the first things I learned about was binder buffer coding, which was about how data can be shared and communicated between different parts of a system or between different threads. It's a used when dealing with message passing or structured data flow. Systems with modular or layered designs can make use of this to communicate with other parts of the system.    After  that, I learned about the concept of semaphores. Semaphores are a core component in concurrent programming as they act as signaling mechanisms used to control access to shared resources. The main idea of semaphores are integer variables protected by atomic operations like wait() and signal(). A semaphore can be used to limit the number of threads that access a resource such as a database connection pool or a bounded buffer to ensure that the variables remain safe ...
Read more