Very nice explanation of Recurrent Neural Networks (RNNs): The Unreasonable Effectiveness of Recurrent Neural Networks and LSTMs (Long Short-Term Memory): Understanding LSTM Networks
Some interesting and free learning courses here
Configuration file locations:
- Windows: C:\Program Files\PostgreSQL\x.x\data\postgresql.conf
- Linux: /etc/postgresql/x.x/main/postgresql.conf
Go to bottom of .conf file, and add this line:
Then create file ‘postgresql.custom.conf’ in the same directory and place your customised configuration settings in it. Any settings set in the custom file will override those in the main config.
Navigate to pgtune and enter the required information, and pgtune will generate custom settings based upon total RAM size and intended use etc:
Copy the generated settings into file ‘postgresql.custom.conf’:
max_connections = 100
shared_buffers = 8GB
effective_cache_size = 24GB
work_mem = 83886kB
maintenance_work_mem = 2GB
min_wal_size = 2GB
max_wal_size = 4GB
checkpoint_completion_target = 0.9
wal_buffers = 16MB
default_statistics_target = 100
random_page_cost = 1.1
Further reading on Postgres performance: http://www.craigkerstiens.com
If you’re not encrypting connections to SQL Azure (or any remote SQL Server instance), then you probably should.
Encrypted connections to SQL Server use SSL, and that is about as secure as you can get (currently).
[Remember: SSL protects only the connection, i.e. the data as it is transmitted ‘on the wire’ between the client and SQL Server. It says nothing about how the data is actually stored on the server].
When you open SSMS’s ‘Connect to Server’ dialog, click the bottom right ‘Options’ button, and make sure you tick the checkbox ‘Encrypt Connection’:
Ensure you add the -N command line option. The -N switch is used by the client to request an encrypted connection. This option is equivalent to the ADO.net option
ENCRYPT = true.
sqlcmd –N –U username –P password –S servername –d databasename –Q “SELECT * FROM myTable”
When creating a linked server to SQL Azure, the @provstr parameter must be set to ‘Encrypt=yes;’:
-- Create the linked server:
@server = 'LocalLinkedServername',
@srvproduct = N'Any',
@provider = 'SQLNCLI',
@datasrc = '???.database.windows.net', -- Azure server name
@location = '',
@provstr = N'Encrypt=yes;', -- <<-- Important!
@catalog = 'RemoteDatabaseName'; -- remote(Azure) database name
ADO.NET Connection strings
ENCRYPT = true” to your connection string, or set the SqlConnectionStringBuilder property to True.
[Remember: don’t distribute passwords by sending as plaintext over the Internet, i.e. don’t email passwords! ]
If you have a high end NVidia graphics card and you’re investigating data science with Keras+Tensorflow, then you obviously want Tensorflow to take advantage of your GPU (training times for deep neural networks can be 10 – 15 times faster even when compared to the latest CPUs).
Getting it all working can be tricky: I found this guide that explains the steps: Installing TensorFlow with GPU on Windows 10
Just came across this link at Packt: https://www.packtpub.com/packt/offers/free-learning
I’ve recently been learning Python with the goal of using it alongside R for data science. It’s got a lot going for it as a language and the package (library) support covers just about every domain you can think of.
Many of the ‘C’ like languages seem intent on creating too much complexity for no other reason than ‘you can’, but Python takes a more pragmatic approach.
I particularly like the Zen of Python (PEP 20):
The Zen of Python, by Tim Peters
Beautiful is better than ugly.
Explicit is better than implicit.
Simple is better than complex.
Complex is better than complicated.
Flat is better than nested.
Sparse is better than dense.
Special cases aren’t special enough to break the rules.
Although practicality beats purity.
Errors should never pass silently.
Unless explicitly silenced.
In the face of ambiguity, refuse the temptation to guess.
There should be one– and preferably only one –obvious way to do it.
Although that way may not be obvious at first unless you’re Dutch.
Now is better than never.
Although never is often better than *right* now.
If the implementation is hard to explain, it’s a bad idea.
If the implementation is easy to explain, it may be a good idea.
Namespaces are one honking great idea — let’s do more of those!