Profile cover photo
Profile photo
Michael Haas
59 followers -
Code and Cat pictures.
Code and Cat pictures.

59 followers
About
Communities and Collections
View all
Posts

Post has attachment
What's your experience with the 'no calibration' deep sleep mode? The ESP documentation claims that it dramatically lowers power consumption on wakeup.

I am particularly interested in the impact it has on Wifi performance. I could not find any documentation on that.

I did find an interesting piece of information, however. With the default sleep mode, the calibration frequency is controlled by a setting stored in the flash. The value x of byte 108 in esp_init_data_default.bin indicates that every x-th wakeup should be performed with RF calibration. For my 2MB module, that section of flash starts at 0x1fc000 (+ 108 bytes for the value of interest).

What's your average time to connect to a Wifi in station mode with your esp8266?

I have been profiling my DHT22-based sensor node. Most of the time (~2s) is spent on connecting to the AP. Data acquisition itself takes around 275ms. Publishing the data (including the timing metrics) takes about 200ms as well.

What are your figures here? I already got it down from 4s to 2s by using a static IP.

Other things I have tried, which helped only marginally at best:

* setting the BSSID and the channel
* using Wifi.setAutoReconnect(true)

Most of these tips come from https://github.com/z2amiller/sensorboard/blob/master/PowerSaving.md

If you are curious, my current Arduino code is at https://github.com/mhaas/esp8266_sensor_node_code/blob/master/src/weather_station_mqtt.ino#L79 - still quite messy :)

Post has shared content
Do want!
Add a comment...

Post has attachment
Spaß des Tages: macvlan. Damit kann man virtuelle Interfaces mit eigenen MAC-Addressen definieren. Sehr nützlich, wenn man einen Server stresstesten möchte, der auf die MAC zur Identifikation setzt.

http://backreference.org/2014/03/20/some-notes-on-macvlanmacvtap/
Add a comment...

Fein, neues Wifidog Release. Das erste seit Jahren. Sogar mit ein bisschen Code von mir!

https://github.com/wifidog/wifidog-gateway/releases/tag/1.2.0

* Add domain whitelist in conf, and auto pass subdomains (#66)
* Add port range in firewall rules (#53)
* Add references to ipsets in firewall rules (#62)
* Add SSL support when talking to the auth server (#63)
* Add disconnect command (#82)
* Add -Wall -Wextra to CFLAGS (#69, #73)
* Add IP to login script URL parameters (#36)
* Add logging support to firewall (#4)
* Add check for valid MAC address (#42)
* Add support for transparent proxy
* Use semantic versioning instead of YYMMDD version scheme.
See http://semver.org/ (#101)
* Use tag for release, don't use master for release (#59)
* Fix incorrect usage of REJECT and DROP in NAT table (#58)
* Fix compiler warnings warnings (#64, #69, #73)
* Fix general code quality issues (#71, 74, #75, #86)
* Fix typo in ping_thread (pong -> PONG) (#46)
* Fix redirect by using HTTP 302 instead of 307 (#11, #14)
* Fix inconsistent indent, now uses spaces everywhere (#91)
* Upgrade libhttpd to 1.4 (#91)
* Remove incomplete and broken BSD support (#93)
* Update doc/README.developers.txt (#113)
* Update README (-> README.md) (#114)
Add a comment...

Fein, mein erstes Paket ist in OpenWrt gelandet: https://github.com/openwrt/packages/pull/1017
Add a comment...

Post has attachment

Post has attachment
A nice talk by Richard Socher on his work on recursive deep learning. Fascinating stuff. If you're super interested, there's also longer tutorials presented at both ACL 2012 and NAACL 2013.  The NAACL 2013 tutorial also talks about the RNTN they use for sentiment analysis.

I found some good review papers on Sentiment Analysis and figured I'd share with the community here. Perhaps you will find one of these interesting.

Feldman, Ronen. "Techniques and applications for sentiment analysis." Communications of the ACM 56.4 (2013): 82-89.
PDF: http://dl.acm.org/citation.cfm?id=2436274

A good, short overview.

Tang, Huifeng, Songbo Tan, and Xueqi Cheng. "A survey on sentiment detection of reviews." Expert Systems with Applications 36.7 (2009): 10760-10773.
PDF: http://www.sciencedirect.com/science/article/pii/S0957417409001626

A more in-depth survey focusing on product/movie reviews.

Bonus:

Taboada, Maite, et al. "Lexicon-based methods for sentiment analysis." Computational linguistics 37.2 (2011): 267-307.
PDF: http://www.mitpressjournals.org/doi/pdf/10.1162/COLI_a_00049

Everything you need to know about lexicon-based methods for sentiment analysis; i.e. if you don't just want to do the old train-SVM-and-go-home approach ;)

Bonus 2:

Wang, Sida, and Christopher D. Manning. "Baselines and bigrams: Simple, good sentiment and topic classification." Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics: Short Papers-Volume 2. Association for Computational Linguistics, 2012.
PDF: http://dl.acm.org/citation.cfm?id=2390688

There's the machine-learning based approach, kind of a "best practices" guide.

Post has attachment
A friend and I did an undergrad term project on sentiment analysis for German twitter data. We obtained noisily labeled data by checking for emoticons and common sentiment words and trained a classifier in Weka. We verified performance using existing annotated data sets containing German tweets. Code is available - if you ever wondered how you can get from the Twitter API to ARFF files for Weka, go check it out ;) All components are connected via JSON and HTTP, so it's easy to rip out anything you need.

Scroll down to "Twitter Sentiment Analysis"! The report is available on Github: https://github.com/mhaas/twitter-sentiment-analysis/blob/master/report.pdf?raw=true
Wait while more posts are being loaded