# Numpy Log2 Returning NaN For Small Values

## 13 February 2022 - 1 answer

I've the following data frame with all positive values

``````    V1          V2          V3      V4          V5
0  F1H5N4S2  10.751263  0.216574  0.703209  10.674107
1    F2H4N7  12.131079  0.000004  1.883824   0.018118
2     H12N2  11.075072  0.214872  0.000004  10.674107
3      H3N7   1.091061  0.000004  3.503290   0.091797
4    F2H4N5   0.590545  0.000004  1.730215   0.223571
``````

When I'm trying to convert the numerical values to log2 using the following syntax in numpy(np)

``````log2df = df.apply(lambda x: np.log2(x) if np.issubdtype(x.dtype, np.float) else x)
``````

I'm getting the following data frame with NaNs in place of log2(0.000004). 0.000004 happens to be the smallest value in the dataframe which I imputed. Can anyone help me solve the problem? Thanks

``````    V1          V2          V3      V4          V5
0  F1H5N4S2  3.426434 -2.207070 -0.507974  3.416043
1    F2H4N7  3.600636       NaN  0.913664 -5.786433
2     H12N2  3.469244 -2.218451       NaN  3.416043
3      H3N7  0.125732       NaN  1.808710 -3.445414
4    F2H4N5 -0.759880       NaN  0.790951 -2.161198
``````

This works fine for me, but avoid using apply. Select the numeric type and apply a vectorial operation:

``````cols = df.select_dtypes('number').columns

df[cols] = np.log2(df[cols])
``````

output:

``````         V1        V2         V3         V4        V5
0  F1H5N4S2  3.426434  -2.207068  -0.507975  3.416043
1    F2H4N7  3.600636 -17.931569   0.913664 -5.786432
2     H12N2  3.469244  -2.218451 -17.931569  3.416043
3      H3N7  0.125732 -17.931569   1.808710 -3.445409
4    F2H4N5 -0.759881 -17.931569   0.790951 -2.161195
``````