While mastering data science through the Flatiron school’s bootcamp, I started working part time at the clothing store lululemon. Upon starting my job in retail, I was just interested in the clothes and working out in the community. It wasn’t until my staff meeting last week that I realized how I can bridge the gap between my passion for data science and love for the cultural at lululemon. During our staff meeting we reviewed our employee pulse survey, which asks a series of question to judge employee satisfaction within the company. While the store manager was going over the results of our individual store, she compared them with the overall results of the company. In three categories we finished well above the company average and we were all very excited, especially the store manager, that we have such a positive experience working at this store; but it wasn’t until the store manager brought up some negative results that I started asking questions? Instinctively, my first question was “Are the store vs company stats for corporate as well or just store locations”. This could lead to why are store statistics were lower than the company average. For example, we fell below average with the question “Do you see yourself working in two years?”. If this question included corporate, it would be no surprise that we trended below average because being in retail, a lot of our employees are part time wether they are still in school studying finance or they are balancing retail with their teaching job. No matter what else our employees are pursuing outside of their job with lululemon, this could explain this negative trend. Another question popped into mind when our manager was showing us the stats for another question, calling this a second timer. She meant this was the second time in a row this question for our store had fallen below the company average. My immediate question to this was, “well is it better or worse than last year?”. The manager paused fo a moment and realized how my question could completely change how we viewed this review. Even though we may fall short of the company average, we should see how we as a store are progressing. Looking at the questions from this perspective can show us even though we may be below average, if we are trending upward it is still successful. After the meeting, I met with the store manager and explained how I could run data analysis on the survey results to better customize our stores answers. I will update my findings once I finish analyzing the data!
The concept of Artificial Intelligence is a new industrial revolution. Artificial intelligence will not just raise our standard of living, but change the way people live. AI will have the same economic and cultural impacts as inventions of the industrial revolution such as the railroad and electricity. These impacts will drastically change how we live our lives. With these thoughts of hope and possibilities of AI in the future we must embrace it and take it head on. We must see it as opportunity rather than a frustrating challenge and feel hope rather than fear. It is by having this mindset that AI will help us live and work smarter that we should approach the topic and dive in, like those at Stanford who have just launched an Institute for Human-Centered Artificial Intelligence. The Stanford HAI was created to show exactly how beneficial AI can be. It promotes and helps develop human-centered AI technologies and different applications that will enhance the human productivity and quality of life.
Batch normalization is a ground breaking development in the training and testing of data in deep learning. Batch normalization enables networks to train, but runs into problems with normalization. When using batch normalization errors in normalization increase dramatically when the batch size becomes smaller. This is caused by inaccurate batch statistics estimates. This normalization error limits the usage for training larger models using batch normalization. This paper will discuss alternatives to this, mainly group normalization. Group normalization works by dividing the channels into groups and runs within each group, making the computation independent of batch sizes. This means the accuracy is more stable for a wider range of batch sizes; outperforming other normalization variants.
A histogram is a widely used graph that visualizes the distribution of data over a continuous interval or a certain period. The data is collected on the outcome of the process. A histogram is composed of bars along either the x or y axis. Each bar represents the frequency at each interval, commonly referred to as a bin. Histograms show the spread of the data. It also shows the frequency of a specific category in a data distribution. Like anything, there are pros and cons to using it. Some of the positives of a histogram is that you can easily see the mean, median, mode and range of a given dataset. This is useful to know when using a new data set to start a new project. By quickly creating a histogram you have a better idea of what you are working with. Another pro when using a histogram is you can see which factor(s) has a relatively higher frequency than others. This can help you ask questions about why this may be happening.
Why did I want to start a career in data science? There is a lot of thought that went into this decision and there was no single contributing factor. I went to college to pursue a degree in mechanical engineering, mainly because I was loved math and science, but wasn’t sure exactly where I wanted it to take me. After graduating I moved to DC and started working in the construction industry for Hilti North America. I liked to consulting aspects of my job and I enjoyed running business analysis for my customers to see where I could improve their business and make our partnership better, but as a whole it was not what I wanted to be doing. I’ve been asked time and time again, like most of us are, “if I could be anything in the world, what would I choose?” I’ve always hated this question because I never had an exciting answer. I never wanted to be a singer or a professional athlete or an astronaut. My answer was always, if I could do anything I would want to be Jonah Hill’s character in the movie moneyball and leave it at that. It wasn’t until I started looking for new jobs that I started thinking this could be an option. I focused on what I like from my current job and looked for jobs that were more aligned with my goals. A lot of these jobs required some experience with Python or SQL so I thought I would teach myself the basics and see if I liked it. I did. It reminded me of the coding class I took in college, I loved the problem solving aspects of writing the code and seeing it run after figuring out exactly what you needed to type. It wasn’t until I started doing this that I realized there are schools and careers based off of these computer programs. I immediately started searching for careers and came across the title “Data Scientist”. It wasn’t something I have ever heard of, but I felt like it was what I was supposed to be doing. I started reading different job descriptions and responsibilities and it checked off the boxes for me. It also is a broad enough field that I could work in multiple industries and find my own path. Once I realized that I wanted a career as a data scientist, I started looking into how I go about it. That is when I discovered the flatiron school’s data science program. The flatiron school was the perfect fit for me and helped me take the first step on my journey of becoming a data scientist.