Where it ends "how do we calculate the weights ?" is fairly simple.
Start completely randomly and compare output to known truth. When it's incorrect, you beat the model up pretty badly and repeat again. Eventually you get the correct answer pretty consistently.
... and by "beat it up" I mean tweak the weights - totally randomly will work but will take a long time (brute force), so we add a bit of intelligence to see which direction to tweak via some algorithms (backpropagation, gradient descent).
An incoming message to an IPv4 NAT router will not be forwarded to a LAN device unless it matches a known flow (typically continuation of a conversation, typically initiated by the LAN device, which is expected), or the user set up a DMZ forward to a particular destination. There is actually no reasonable way for non-DMZ LAN devices to be exposed to the noise.
For non-NAT IPv6, sure a firewall might be on by default, but it can be turned off - and therein lies the potential exposure to every LAN device to directed traffic.
In other words, the risky zone for IPv4 NAT tends to be setting up a DMZ exposing 1 device, while the risky zone for IPv6 non-firewalled tends to be exposing all of the devices behind the router.
Entirely unrelated to the topic, but my mind was blown when I recently learnt that clamp() can be implemented via .median(val, low, high), and the param order doesn't matter!
Hivestack by Perion | Montreal, QC (Hybrid Office/Remote) | Full Time
Industry: AdTech, specializing in Digital Out Of Home
I'm hiring for two engineering lead positions (senior/staff level):
* Data Discipline Lead - this includes driving the architecture and technical roadmap, and the implementation of various technologies to support adtech-scale ingestion, transport, processing, warehousing, and reporting.
* Edge Discipline Lead (Ad Bidding & Delivery): This position leads the team responsible for the high-volume low-latency ad bidding and delivery systems - a particularly good fit for someone who enjoys systems performance engineering, architecture, while still being hands-on
Where it ends "how do we calculate the weights ?" is fairly simple.
Start completely randomly and compare output to known truth. When it's incorrect, you beat the model up pretty badly and repeat again. Eventually you get the correct answer pretty consistently.
... and by "beat it up" I mean tweak the weights - totally randomly will work but will take a long time (brute force), so we add a bit of intelligence to see which direction to tweak via some algorithms (backpropagation, gradient descent).