Deep Discovery: Artificial intelligence for fighting global corruption.
"Three things cannot long stay hidden: the sun, the moon and the truth."
--Buddha
@marktenenholtz Global metrics for accuracy are best ignored completely in favor of slice based learning: does it perform well on the data that matters?
📣Attn software engineers! @MarcoPoloApp, @deepdiscoveryai, @join_honeycomb and many more companies are hiring at https://t.co/zZ89DzVT3s! Check them out 👀
Treasury sanctions crypto exchange Chatex over accusations of association with ransomware scams. KYC requirements for crypto service providers is INTENSE right now.
https://t.co/LrgcSMYxKZ
Hey
@SnorkelAI
fans -
@deepdiscoveryai
is hiring a Weakly Supervised Learning Engineer to help us label data via weak and distant supervision, semi-supervised learning and to optimize experimental design via active learning.
You could be the world's 1st official WSL Engineer!
Agile Data Scienc 2.0 has received a MASSIVE code update. The book now runs on Docker for PySpark, Mongo, Elasticsearch and Kafka. It is VERY easy to go through in the notebooks. The libraries are all updated to modern choices and versions. Enjoy!
https://t.co/AHu6CKfuuI
Hey @cstross ... I thought you should see this. How can we setup some kind of containment field? Your books are excellent historiography but have so little applied information.
I wrote a GEXFWriter class to export graphs with attributes in Python to @Gephi that takes a pair of node/edge pyspark.sql.DataFrames or pandas.DataFrames and produces valid GEXF with attributes locally or on S3.
Not sure where to contribute this to FOSS?
@dryifanhu your paper on the Yifan Hu graph layout isn't really searchable. I can't get anything to really parse it's encoding (Acrobat, Mendeley, Chrome) due to an encoding problem, so I can't search it. Do you have another copy other than the one at https://t.co/T4m0Zy5VcE?
There is great responsibility when automating manual tasks of import with machine learning that you hold yourself to the same burden of evidence that you would evaluating the manual process. Otherwise it isn’t the important task you have automated, but a new and ill defined one.
Message passing in GraphFrames presents a Pregel like interface that is incredibly powerful for implementing graph algorithms. We use it for our risk scoring algorithms at @deepdiscoveryai :)
Like large scale distributed graph ML? We’re hiring!
https://t.co/pY28GUzyfG
I need a parallel or distributed layout algorithm for subgraphs of our multi billion node business graph with millions of nodes and edges. Yes I need beautiful hairballs. OpenORD isn’t a parallel algorithm and is too slow on Gephi.