Press "Enter" to skip to content

Category: Machine Learning

Preventing Overfitting in ML Models

Tom Jordan gives us four techniques to reduce the likelihood of overfitting in our models:

This technique is exclusively used within the training of neural networks, so isn’t applicable to all machine learning models, however can be used in the production of extremely effective neural network models. During the start of each step in the training process, each sub unit of the model, the neuron, has a probability of being included in that step or not. If it doesn’t make the cut, it is effectively deleted from the network for that step, and then reintroduced on the next step.

There are some good techniques here.

Leave a Comment

Dealing with NULLs in Java with SQL Server 2019

Niels Berglund covers changes in SQL Server Machine Learning Services around Java code execution:

In the null values post mentioned above, I mentioned that there are differences between SQL Server and Java in how they handle null. So, when we call into Java from SQL Server, we may want to treat null values the same way as we do in SQL Server.

I wrote about this in the SQL Server 2019 Extensibility Framework & Java – Null Values post mentioned above. However, that post was written before SQL Server 2019 CTP 2.5. In CTP 2.5 Microsoft introduced the Java SDK, and certain things changed. Amongst the things that changed is the way we handle nulls when we receive datasets from SQL Server in our Java code.

Read on to learn how it works today.

Leave a Comment

MLFlow on Databricks Community Edition

Jules Damji and Siddharth Murching have an interesting announcement:

Today, we are excited to extend Databricks Community Edition with hosted MLflow for free, as part of our ongoing commitment to help developers learn about machine learning lifecycle. With the Community Edition, you can try tutorials that demonstrate how to track results and experiments as you build machine learning models—a crucial stage in the machine learning model’s development lifecycle.

MLflow is an open-source platform for the machine learning lifecycle with four components: MLflow TrackingMLflow ProjectsMLflow Models, and MLflow Registry. MLflow is now included in Databricks Community Edition, meaning that you can utilize its Tracking and Model APIs within a notebook or from your laptop just as easily as you would with managed MLflow in Databricks Enterprise Edition.

I like showing off Databricks Community Edition, and I’m glad to see them extend it a bit.

Comments closed

Topic Modeling

Federico Pascual has an article on topic modeling and topic classification:

Topic modeling is an unsupervised machine learning technique that’s capable of scanning a set of documents, detecting word and phrase patterns within them, and automatically clustering word groups and similar expressions that best characterize a set of documents. It’s known as ‘unsupervised’ machine learning because it doesn’t require a predefined list of tags or training data that’s been previously classified by humans.

Since topic modeling doesn’t require training, it’s a quick and easy way to start analyzing your data. However, you can’t guarantee you’ll receive accurate results, which is why many businesses opt to invest time training a topic classification model.

The article is long but worth the read, with examples in Python and additional notes for R.

Comments closed

TensorFlow Changes

Ajit Jaokar summarizes key changes from TensorFlow from 1.x to 2.0:

The Data pipeline simplified:  TensorFlow2.0 has a separate module TensorFlow DataSets that can be used to operate with the model in more elegant way. Not only it has a large range of existing datasets, making your job of experimenting with a new architecture easier – it also has well defined way to add your data to it.
In TensorFlow 1.x for building a model we would first need to declare placeholders. These were the dummy variables which will later (in the session) used to feed data to the model. There were many built-in APIs for building the layers like tf.contrib, tf.layers and tf.keras, one could also build layers by defining the actual mathematical operations.
TensorFlow 2.0 you can build your model defining your own mathematical operations, as before you can use math module (tf.math) and linear algebra (tf.linalg) module. However, you can take advantage of the high level Keras API and tf.layers module. The important part is we do not need to define placeholders any more.

These look like some nice improvements.

Comments closed

K-Means Clustering with Python

Abhinav Choudhary walks us through k-means clustering using scikit-learn:

K Means Clustering tries to cluster your data into clusters based on their similarity. In this algorithm, we have to specify the number of clusters (which is a hyperparameter) we want the data to be grouped into. Hyperparameters are the variables whose value need to be set before applying value to the dataset. Hyperparameters are adjustable parameters you choose to train a model that carries out the training process itself.

Read on for a demo.

Comments closed

Secret Sharing and Neural Networks

Adrian Colyer reviews an interesting paper:

Take a system trained to make predictions on a language (word or character) model – an example you’re probably familiar with is Google Smart Compose. Now feed it a prefix such as “My social security number is “. Can you guess what happens next?

Read the whole thing. There’s a bit of discussion at the end around how you can stop this learning of secrets.

Comments closed

Key Concepts of Convolutional Neural Networks

Srinija Sirobhushanam takes us through some of the key concepts around convolutional neural networks:

How are convolution layer operations useful?
CNN helps us look for specific localized image features like the edges in the image that we can use later in the network Initial layers to detect simple patterns, such as horizontal and vertical edges in an image; and deeper layers detect complex patterns.

This is a rather high-level description of the topic, but it serves to give an understanding of what you need to know before building and using these networks.

Comments closed

LSTM in Databricks

Vedant Jain shows us an example of solving a multivariate time series forecasting problem using LSTM networks:

LSTM is a type of Recurrent Neural Network (RNN) that allows the network to retain long-term dependencies at a given time from many timesteps before. RNNs were designed to that effect using a simple feedback approach for neurons where the output sequence of data serves as one of the inputs. However, long term dependencies can make the network untrainable due to the vanishing gradient problem. LSTM is designed precisely to solve that problem.

Sometimes accurate time series predictions depend on a combination of both bits of old and recent data. We have to efficiently learn even what to pay attention to, accepting that there may be a long history of data to learn from. LSTMs combine simple DNN architectures with clever mechanisms to learn what parts of history to ‘remember’ and what to ‘forget’ over long periods. The ability of LSTM to learn patterns in data over long sequences makes them suitable for time series forecasting.

This is a nice overview and as a bonus, there’s a notebook as well where you can try it on your own.

Comments closed

Performance Tuning Neural Network Training

Sean Owen takes us through a few techniques for speeding up neural network model training:

Step #2: Use Early Stopping
Keras (and other frameworks) have built-in support for stopping when further training appears to be making the model worse. In Keras, it’s the EarlyStopping callback. Using it means passing the validation data to the training process for evaluation on every epoch. Training will stop after several epochs have passed with no improvement. restore_best_weights=True ensures that the final model’s weights are from its best epoch, not just the last one. This should be your default.

Sean focuses here on Keras + TensorFlow on Spark, but several of the tips are cross-product and generally applicable.

Comments closed