Whether you are in a startup looking to add some simple AI to your product, you are just getting started learning into AI in general or you just want to experiment with an idea — No-Code or Low-Code might be something worth checking into for you.
What are No-Code and Low-Code?
These terms are used so much these days, but I think it's worth just making sure we know what each means. If you are not writing any code, then that's a no-code solution. No-code usually has a nice easy to use interface and generates code behind the scenes. Typically the trade-off with no-code is that you have to rely on the platform that you build them on. In essence, locking you into using the product (Mito and BitRook are the exceptions) because you can’t export your model code and run it somewhere else. That being said you can build a working model is usually less than an hour if you have your data ready.
On the other hand, if you are writing a bit of code, but saving yourself the time and effort of writing a lot of code, then that is a low-code solution. Most of these are libraries in Python that are pretty easy to use and meant for someone not
Low-Code Cleaning Tools
Mito is a low-code way to be able to edit and transform your data in a Jupyter notebook and have it generate code for your changes. It's completely free for individuals and worth checking out.
If you love Jupyter notebooks but are not too confident with Python then this is a huge time saver. As you make changes to your dataset it will write the code beneath. This makes it so you possibly can make the changes once manually and then copy and paste the code to automate your changes in the future.
Installation is just a simple pip command and then creating a sheet with the instructions on their site.
python -m pip install mitoinstallerBitRook is a unique desktop app that is more like a Data Science swiss army knife. It uses ML to analyze and help clean your data — it even generates a python script to automate your cleaning. It helps you analyze your data, and instead of you searching for issues — it raises them to your attention and can tell you if a dataset is predictive. It handles large datasets and is much faster at loading data than Excel. Worth checking out and the free version is robust.
Features
- Generates a python script for you
- Predictive Data Detection (Correlation Matrix & Predictive Power Score)
- Handles large data
- Column Type Detection & Type Standardization
- Common Data Cleaning Functions Built-In
- Unique values, missing values
- Quantile & descriptive statistics
- Most frequent values (category, letter frequency & word frequency)
- Outlier handling
- PII Data Detection
- Viewing CSV is faster than Excel and EDA built-in
- Splitting data
- Data profiling script generation
Low-Code Model Building Libraries
H2O is a fully open-source, distributed in-memory machine learning platform with linear scalability. H2O supports the most widely used statistical & machine learning algorithms including gradient boosted machines, generalized linear models, deep learning, and more. H2O also has an industry-leading AutoML functionality that automatically runs through all the algorithms and their hyperparameters to produce a leaderboard of the best models.
Sound like a lot of marketing? Well, basically it's a way to make the most popular types of models in about 10 lines of code.
Installation
pip install h2o
Auto_ViML was designed for building High-Performance Interpretable Models with the fewest variables. The “V” in Auto_ViML stands for Variable because it tries multiple models with multiple features to find you the best performing model for your dataset. The “i” in Auto_ViML stands for “interpretable” since Auto_ViML selects the least number of features necessary to build a simpler, more interpretable model.
Features
- Helps you with data cleaning
- Assists you with variable classification
- Performs feature reduction automatically
- Produces model performance results as graphs automatically
- Handles text, date-time, structs, and more
- Allows you to use the featuretools library to do Feature Engineering
Installation
pip install autoviml
Usage is even easier and honestly, you can pretty much make multiple models with one line of code. Check out the usage docs here
If you need a library that gives you the ability to create a model with minimal code, then this is one to check out.
MLBox main package contains 3 sub-packages: preprocessing, optimization and prediction. Each one of them is respectively aimed at reading and preprocessing data, testing or optimizing a wide range of learners, and predicting the target on a test dataset.
Features
- Fast reading and distributed data preprocessing/cleaning/formatting
- Highly robust feature selection and leak detection
- Accurate hyper-parameter optimization in high-dimensional space
- State-of-the-art predictive models for classification and regression (Deep Learning, Stacking, LightGBM,…)
- Prediction with models interpretation
Installation
Installation is fairly simple, but a few things to make sure you have — check out their docs here.
pip install mlbox
PyCaret is an open-source, low-code machine learning library in Python that allows you to go from preparing your data to deploying your model within minutes in your choice of notebook environment. PyCaret is essentially a Python wrapper around several machine learning libraries and frameworks such as scikit-learn, XGBoost, Microsoft LightGBM, spaCy, and many more.
Features:
- Data Preparation
- Model Training
- Hyperparameter tuning
- Analysis & Interpretability
- Model Selection
- Experiment Logging
Installation
pip install pycaret[full]Developed by Salesforce TransmogrifAI (pronounced trăns-mŏgˈrə-fī) is an AutoML library written in Scala that runs on top of Apache Spark. Through automation, it achieves accuracies close to hand-tuned models with almost a 100x reduction in time.
Features:
- Build production-ready machine learning applications in hours, not months
- Build machine learning models without getting a Ph.D. in machine learning
- Build modular, reusable, strongly typed machine learning workflows
Couldn’t find a good tutorial I would recommend honestly, this one seems tougher to get started with than the other options listed here.
TPOT will automate the most tedious part of machine learning by intelligently exploring thousands of possible pipelines to find the best one for your data. TPOT is built on top of scikit-learn, so all of the code it generates should look familiar… if you’re familiar with scikit-learn, anyway.
Features
- Code or command-line driven
- Explores thousands of possible pipelines to find the best one for your data.
- Provides you with the Python code for the best pipeline it found so you can tinker with the pipeline from there
AutoGluon enables easy-to-use and easy-to-extend AutoML with a focus on automated stack ensembling, deep learning, and real-world applications spanning text, image, and tabular data. Intended for both ML beginners and experts.
Features
- Quickly prototype deep learning and classical ML solutions for your raw data with a few lines of code.
- Automatically utilize state-of-the-art techniques (where appropriate) without expert knowledge.
- Leverage automatic hyperparameter tuning, model selection/ensembling, architecture search, and data processing.
- Easily improve/tune your bespoke models and data pipelines, or customize AutoGluon for your use case.
No-Code Model Building
CreateML is Apple’s iOS tool set to create models to run on iOS and Mac devices. It is made to be easy and require almost no coding to generate the model and some coding to implement the model. Can be used for audio, video motion, sound, text or tabular data models.
Features
- Build and train powerful on-device models with an easy-to-use app interface.
- Train multiple models using different datasets, all in a single project.
- Preview your model performance using Continuity with your iPhone camera and microphone on your Mac, or drop in sample data.
- Pause, save, resume, and extend your training process.
- Train models blazingly fast right on your Mac while taking advantage of CPU and GPU.
- Use an external graphics processing unit with your Mac for even better model training performance.
Installation
No installation, Create ML is included in MacOS XCode
Google Cloud Auto ML Vertex AI
Build, deploy, and scale ML models faster, with pre-trained and custom tooling within a unified AI platform. If you are using GCP, consider this one.
Features
- Train models without code, minimal expertise required
- Build advanced ML models with custom tooling
- Manage your models with confidence
- A unified UI for the entire ML workflow
- Pre-trained APIs for vision, video, natural language, and more
- End-to-end integration for data and AI
- Support for all open source frameworks
Hard to describe this platform, but if you just want to worry about data-prep and not building or deploying a model then this is a platform you should check out. I think the demo video above really shows whats possible with this platform.
Features
- Shape your training and testing data
- Automatically clean your data and select recommended features
- Develop multiple algorithms and measure results to find the best preforming algorithms
- Get access to the model through a simple python library or through REST API
Akkio tries to make building models with Auto ML easy for “data-savvy operators & engineers”. Most of the implementation doesn’t require code, but more just access to training data. Worth checking out and trying.
Feature
- Integrations with many common data sources (Snowflake, CRMs and Google Sheets)
- Lots of the most common models for a business for easy implementation
- Deploy and easily call models through a simple REST API
Pre-Made AI Models
Its hard to describe HuggingFace entirely, but I still agree with my tweet a while ago to just sum up where HuggingFace sits in the world. It is an AI community that specializes in Transformer based models. It allows you to access and use some of the most sophisticated models and even deploy it in popular environments like Amazon’s Sagemaker. While they do have many different kinds of models (over 20k as of writing this), they do seem to have mostly NLP models.
Features
- Access to one of the largest lists of models that specialize in various tasks and constantly updating as well
- Easily train the model with your own data in your AWS cloud or with HuggingFace
- Deploy models in seconds and use they practically right out of the box with a simple API call
- Great community support and tooling
A large curated catalog of pre-trained models in various frameworks for several kinds of tasks and most importantly links to their source code.
Features
- Free pre-trained model code (but that also means you need to deploy them yourself)
- Large list of models and direct access to Github source code
- Brings in the github repo’s markdown file to quickly show how to implement the model code
Fairly large list of pre-trained models from a variety of companies that helps to bridge the gap between companies offering pre-trained models and the infrastructure that these models need to run on.
Features
- Each model shows cost from business and the necessary infrastructure in AWS to run the model
- Since these are commercial pre-trained models you can purchase support from these companies
- Geared toward companies already using AWS and need a quick model in AWS Sagemaker
A community for pre-trained models created by customers and companies that support a variety of tasks and data types. Google leans towards its amazing Tensorflow framework with many of these models. Sadly some of these models are still a bit complicated to deploy on Google Cloud’s Vertex AI.
Features
- If you are a fan of Tensorflow framework, then this community might be for you
- Many of them deploy in Google Colab for notebook users
- Great documentation on many of the models
- Easy to find models based of the data type and task
What can I say about this one? It’s just a large Github list of pre-trained models focusing on the task of Computer Vision. Each is separated by the framework it is using and an easy to read description and license of each model. Great one to give a Github Star to for later.
Another great list by the same awesome guy, except this one is about NLP. Again it separates each of the models by the framework its using under the hood. Great one to give a Github Star to for later.
The final list focused on Audio and Speech pre-trained models. Each separated by framework and ready for a git clone command.
Last thing I will leave you with is that there are honestly many more I could have listed here. There is an expanding landscape of services that are democratizing AI and there are many communities around specific frameworks (like TensorFlow Hub ).
Hit me up on Twitter if you know of others I should have included! Always love to hear from any of you.
Final Thought
The trend for AI being easier to create and deploy is going to continue and sophisticated custom models will still continue as well pushing the boundaries and paving a way for the no-code platforms to make deploying that model just a couple of clicks away.
