"Poverty is like creeping paralysis, slowly it destroys the desire for freedom, strips one of the ambition to enjoy better things of life and undermines personal initiatives.
I built `ripple_net`
A Python library for multimodal image search in image datasets, that is, searching images with text descriptions, or searching for similar images....with images
@tensor_kelechi
I'm a Python developer and Data Scientist. I can do the following:
Create algorithms using Python
Build data web apps using Streamlit
Work on Data Science and Machine Learning tasks using Python
Work on UML tasks
Please Retweet, my client could be on your timeline 🙏👍
DS by choice
ML by mistake
DL by mistake
But I'm currently loving ML and DL...
Currently working on 3 research and trying to find flaws on the VIT and recommendations
Would you like to get started with or transition into Machine Learning/ AI?
I just compiled a beginners friendly roadmap with (free efficient resourses) to take you from Basic Python Programming to intermediate ML in (about) 7 months.
Kindly reshare! 😁
1/
DataFestAfrica2024 Datathon is here!!!
Are you a data professional(data analyst, data scientist, data engineer, ML engineer) in East, West, North , and South Africa?!
Are you ready to showcase your data skills and solve real-world problems?
The
#DataFestAfrica2024
Datathon is
One of the Data Sources that people don’t talk about often
Web Scraping:
Extract relevant data from unstructured sources.
Also known as Screen scraping, web harvesting and web data Extraction.
Download specific data based on defined parameters.
Can extract text, contact
Code for the first question😮💨
product_sales = df.groupby('Product Name ')['Units Sold'].sum()
top_selling_products = product_sales.sort_values(ascending=False)
Hiring several backend/data/ML engineers for our new(ish) company, focused on building high-quality structured data from raw, noisy inputs. Have funding, revenue, users
Streamlit is an easy to use framework for building data apps... without relying heavily on JS/Frontend
I do use it a lot to build my ML models user interface
Finally deployed it on
@streamlit
cloud
Some images:
You can test it:
This was the first time I'm working with more than 2 datasets at the same time
Dataset was gotten from Maven Analytics recent challenge
Tried to convert a previous project of mine from its jupyter notebook to a web app using Streamlit.
This project implements Scipy, Pandas, Matplotlib, Sklearn and KMeans
There are some constraints to this, the web app only allows an image upload of less than 1mb
how to diagnose your machine learning models effectively!
if you want to know your ml model's health - plot a learning curve.
learning curves are very helpful in inferring if your model is overfitting, underfitting or doing just fine.
here's a pretty good hack i do use;
🧵🧵🧵
Start with a goal to understand the meaning of data, it's application, how data analytics, data science or data engineering could be implemented. Before you think about learning a tool
Dear Data Analysts,
Many people talk about the right path to be a data analyst. Start with this or start with that
I am here to tell you that there is no right path but I would like to use the word "recommended path"
🧵
I was going through the SUPERSTORE DATASET all over again and decided to dive deeper into the data to draw out insights and trust me I found some interesting stuffs 🧐
Not to bore you, I will divide the insights into phases.
This first phase will be based on PROFIT, SALES,
Trained a data with RF... accuracy was 0.90. did feature importance, and selected the best features with a threshold of >0.5
Used the selected features to train the data again and this time got a 0.99 accuracy and an improvement in F1 score.
Still doing my research into this
You learn better on the job... I started with Python for DS and ML, at work I learnt how to create backend APIs that could be used to deploy models, same with RAG, Langchain, llama index, Vector databases.
If you’re new in your DA career, my advice is don’t try to favour one tool over the other. Learn from both worlds. Chances are you might find a job that requires either Power BI or Tableau. I’ve had to use both even within the current company I work. Be flexible.
My contribution to this, your suggestions/criticisms is appreciated:
I'm currently working on a Medium article, but for now I need to watch Anime and close the PC😮💨
“Why is it that the Oil in the Niger Delta for the whole of Nigeria but the Gold in Zamfara belongs to Zamfara?”
The NBA conference gave rise to some important questions of national concern.
Some personal observations:
Using the T4 GPU to run this UNet hyperparameter tuning takes almost 2 hours
Using the Nvidia Tesla A100 GPU takes only 36 minutes on the same task
More GPU compute == A faster training process for Deep Learning models
Recently came across RAG (Retrieval Augmented Generation) and Its concept.
LLMs are trained on data fed into it. There are chances that such data might not be up to date, or the LLM might hallucinate when given a prompt that's not related to the data that it was trained on.
Yes, SQL is a domain specific language and not a general purpose programming language. You cannot create an application or build a webpage with SQL, but it definitely looks like programming when you use SQL to talk to your databases. More info:
Tbh I'm sorry to those that have messaged me for assistance with their projects or models on X... I'm in my finals and I've work (Startup, freelance, open source)that's choked up... trying to combine both hasn't been easy
@raqibcodes
I'm at the stage of building an App that uses the model, I have wrapped the model into a flask API, tested it on Postman, it works and can make predictions
So I've been up trying to replicate the entire flask backend process using nodejs
Will ship it out soon
Thanks to a friend, we were able to fix the nodejs aspect and now I can successfully deploy and test a ML model via nodejs
Next: Deploying and testing ML models via Rust
Trained a data with RF... accuracy was 0.90. did feature importance, and selected the best features with a threshold of >0.5
Used the selected features to train the data again and this time got a 0.99 accuracy and an improvement in F1 score.
Still doing my research into this
@AdoraNwodo
I went for some night parties in my first year 2nd semester, taught Python remotely, worked on data science gigs remotely and still had A and B in some courses.
Yes I agree that this is a myopic thinking