class: center, middle # Machine Learning with TensorFlow JavaScript William Kaiser --- # Personal Introduction William Kaiser, a student from ~~Northern Virginia~~ Houston, Texas. - **Second-year Computer Science Major** in the School of Engineering and Applied Sciences - Long time hobbyist, first-time teacher - Worked with Professor **Stephen Baek** in the School of Data Science - TensorFlow Developer Certified ??? Getting certified is **not** worth it. --- # Personal Projects Built - [International Science and Engineering Fair Visualizer](https://wkaisertexas.github.io/all-isef-projects/) - [TikTok Uploader](https://github.com/wkaisertexas/tiktok-uploader) -  - [UVA Student Information System Google Sheets Connector](https://uvamisc.com) ??? Flex the fact that Dhar-Man style videos were uploaded because of me... --- # Course Breakdown - Machine Learning - TensorFlow - JavaScript --- # Machine Learning Finding patterns in data. $$y = m x + b \implies m, b \text{ are parameters to tune}$$ <img alt="Machine Learning is everywhere" src="https://s40424.pcdn.co/in/wp-content/uploads/2023/03/types-of-machine-learning.jpg.optimal.jpg" width="400px"/> --- # TensorFlow Machine learning is complicated, but really just linear algebra. Tensors are multi-dimensional arrays 1. Scalars (normal numbers) 2. Vectors (lists of number) 3. Arrays (lists of vectors) 4. Tensors (lists of lists of vectors) --- # TensorFlow (Continued) Operations (adding, subtracting, multiplication) are represented by nodes in a graph.  --- # JavaScript Invented in 1995 as a way to give web-pages interactivity. ```javascript window.click = alert; ``` Java-Inspired Syntax --- class: center, middle # Course Design Philosophy Why I am **not** crazy --- # Controversial choices Those versed in machine learning would find these choices objectionable. Why? - TensorFlow - JavaScript --- class: middle *Machine Learning with TensorFlow JS* is a beginner-friendly course designed to give students exposure to recent ai developments and a fundamental understanding of paradigms. Unlike traditional courses, a deep-dive on the underlying mathematics will **not** take place and the final projects will be a **communicable, portable project**. --- class: middle center # Summary: Learning what there is to Learn --- class: center # "communicable, portable project" > "Created a carbon emissions predictor for the *University of Virginia* to estimate emissions impact of construction projects on grounds. Achieved an F-1 score of 95%." > "Created a carbon emissions predictor and visualizer for the *University of Virginia* to estimate carbon emissions impact of construction projects on grounds. Achieved a F-1 score of 95%. [Check out the visualization here.](https://google.com) --- # JavaScript / Python Equivalence Unless you are doing something wonky, the defining machine learning models are conceptually identical. `JavaScript` is stylistically similar to `Java` which many people know because of JS ??? If you know one, you know the other. --- # Cloud-Computing Models are **typically** deployed on Servers in the Cloud in production. Furthermore, these are typically linked to data engineering pipelines to continuously update models. <!--  --> <img alt="jedi meme" src="meme2.jpeg" width="500px" /> ??? If you know this, get out of this class --- ### TensorFlowJS is fast Almost as fast as [CUDA](https://developer.nvidia.com/cuda-toolkit) for most applications. - Image detection application  ??? Benchmark compares MobileNet, image-detection: - In-Browser - Using Apple's Metal Compute library - OpenCompute Language --- - MandelBrot set  ??? Ran on a high-performance computing cluster with multiple multi-GPU servers --- class: center, middle # JavaScript is really good And has gotten way better in the past six months --- class: center, middle # Sync Point Questions? ??? Moving onto install node and writing your first javascript program --- class: center, middle ## Course Resources [Syllabus](./modified_syllabus) [JavaScript Reading](./javascriptIntro) --- ## JavaScript Demonstration --- ## Activity 1. Install [node](https://nodejs.org/en) 2. Fetch [https://api.devhub.virginia.edu/v1/courses](https://api.devhub.virginia.edu/v1/courses) 3. Transform the dataset into the following form: ```json { "headers": ["column1", "column2", "colum3"], "rows": [ { "items": ["item1","item2", "item3"] }, { "items": ["item1","item2", "item3"] } ] } ``` ---