Posts

CST383 - Week 6

CST383 Week 6 This week, we are learning about how to optimize machine learning algorithms. Hyperparameters are values that are not able to be learned by the algorithm, so must be set before the training of data begins. To find the best values for hyperparameters, tuning tests on the existing data needs to be ran. With hyperparameter tuning, we can use three strategies: grid search, random search, or weighted KNN. Grid search goes through every combination in a data set to find the best value for the hyperparameters. Random search quickly calculates a best value from the minimum amount of tests, but can be inaccurate. Weighted KNN makes inverse distance of points matter more than the mode of the nearest # neighbors. The steps for hyperparameter tuning best practices is:  Test/train split Cross validation with training data Train best model on full training set Compute test score with test set For supervised learning, classification predicts a category while regression predicts a qu...
Post a Comment
Read more

CST383 - Week 5

Image
CST383 - Week 5 This week we learned about how to prepare data for intro machine learning techniques. This week was rather difficult for me to understand how to properly create KNN classifiers, so I went online to find some supplementary material to help me. One resource that made it a lot easier for me to understand was this video:  The homework was good practice for me to work with, but I'll have to practice more to get the hang of creating classifiers! 
Post a Comment
Read more

CST383 - Week 4

CST383 - Week 4      This week, we worked more on plotting data, this time creating more complicated plots involving two or more variables. We used pd.crosstab() to create grouped bar plots for two variables, while for three or more we use sns.barplot() .       I had some trouble understanding why some of my plots are not plotting at all (loading infinitely) while others are able to plot. I tried a few ways to plot them, including normalizing the data, using different aggregations like sum() , and limiting the amount of data that was plotted using top 10 values [:10] . I will have to review the homework once it is graded to understand how best to work through these problems.     I also learned more about P notation for joint probability: use "and", or a ",", or a "^". Such as, what is the probability for a and b: P(a and b), or P(a, b), or P(a ^ b). For OR, use "or" or a "v".       For conditional probabilities like "for ...
Post a Comment
Read more

CST383 - Week 3

CST383 - Week 3      This week we learned more about how to work in Colab/Jupyter notebook. I personally used Jupyter because I thought self-hosted material is more appropriate for my use case. I think this notebook style of coding is very useful, because I can test only a single block of code at a time rather than the full file.      We started to learn how to code Pyplot visualizations of data, which is very nice. I am a visual learner so these exercises were very helpful for me to understand these concepts better. I didn't have much difficulty understanding these concepts to complete the homework, though I know I will need to read further into graphing two continuous variables in the coming weeks. A few things stood out to me to remember: Use a semicolon ( ; ) after the last plotting statement or use plt.show() at the end of the code block to prevent matplotlib from automatically adding text before a graph. Use plt.xlabel() or plt.ylabel() to label th...
Post a Comment
Read more

CST383 - Week 2

CST383 Week 2 This week we learned about Pandas, a data analysis Python library that makes working with data tables much easier. It can make use of NumPy if needed, specifically for functions like mean().  I have had some trouble with remembering Pandas syntax. I know to use df['x'] to obtain a Series named 'x', then df[df[x]] to obtain the whole Dataframe named 'x'. Once we got to aggregation, I was also unsure when to use 'groupby' and when to use 'aggregate'. One more thing that was difficult for me to get right was how to use conditions correctly in selecting data. One of the most helpful articles to understand this better was the  user guide to Pandas. Along with this, I also found a nice cheat sheet on the Pandas home page for remembering specific syntax. Here is the link to it. I think I would need a lot more specific examples in the lecture to understand the syntax better and when to apply them.
Post a Comment
Read more

CST383 - Week 1

CST383 Week 1 New semester! This is the Intro to Data Science course.      The reading started us off with the Iris dataset, and how to manipulate the data from it to predict future iris data (at a 97% success rate). This style of machine learning is supervised learning, where we train it on data that we know to be true. This allows us to measure the accuracy of the model in identifying irises.     For unsupervised learning, we feed data into the algorithm but do not provide any known output data that we know to be true. An example would be to create a summary of reviews of a product. There is no known output data because the reviews must be fed into the algorithm first and compare it to existing reviews in the database, and find commonalities with them.     Before creating a supervised model, the best practice is to look at your data (visually helps, like a scatter plot) and find commonalities with each feature of the samples. This can help identify i...
Post a Comment
Read more

CST370 - Week 7

CST370 Week 7 (& 8?)      This week we learned about dynamic programming and how to use backtracking to find optimal solution paths. Instead of brute-forcing through all solution paths, dynamic programming uses a solution array that keeps track of already-calculated sub-problems in a table. Once we hit the end, we can use that table to backtrack and find the solution to the problem.      We also learned about Warshall's algorithm, which uses transitive closure to find paths from one vertex to a non-adjacent vertex in a directed graph, and Floyd's algorithm to further apply this to find the all-pairs shortest-paths matrix solution of a connected cyclic graph. The solution to Warshall's algorithm is a matrix of 0s and 1s showing if a valid path from one vertex to another is possible. If the path (row) has all 1s across, a path from that vertex (row index) can be traced to all vertices, including itself if it is part of a cycle. The solution to Floyd's alg...
Post a Comment
Read more